v15.5.0 (#2850) (#2851)

* update

* Revert "v15.4.0"

This reverts commit 74d4f20858.

* Revert "prepare v15.4.0 versuch 2"

This reverts commit e790a14caa.

* revert submnodules (#2768)



* Update submodules (#2769)

- Update `esp-nn` to v1.0.2
- Update `esp32-camera` to v2.0.6
- Updated `tflite-micro-esp-examples` (The repo got renamed to `esp-tflite-micro ` and the folder structure got cleaned up).

* fix(unity-test): Run unity tests in dedicated task
Avoid running out of heap

* Add delay

* Update

* Update

* Enhance busy notification (#2774)

* on marker updating, show message and retry until round got completed

* same for contrast enhancement

---------



* Fix negatives on extended resolution false #2744 (#2772)

* not extended resolution allows -1 on the lowest digit

* not extended resolution allows -1 on the lowest number

* negatives on last value digit with -1 will set to prevalue and is not an error  #2744

---------



* Implemented late analog / digital transition (#2778)

* Implemented late transition

Complete rewrite of analog / digital transition

Two tests is still failing, which need to be discussed.

* Allow wider range of transition values to support late transition

* Added documentation

* Fix testings (#2783)

* fix all tests and more description

* The decimal point offset. -3 corresponds to x.yyy

---------



* Bugfix InfluxDB shifting times (#2785)

* Modify time_sntp

* Update time_sntp.cpp

* Update time_sntp.cpp

* update

* Update interface_influxdb.cpp

* Update interface_influxdb.cpp

* Update time_sntp.cpp

* Upload

* Update interface_influxdb.cpp

* Update interface_influxdb.cpp

* Update interface_influxdb.cpp

* Remove Time Convert vom influx Interface

* Update platform-espressif32 to 6.5.0 (#2770)

* Update ESP IDF to 6.5.0

* Migration to new IDF

* Correct smtp vor v5.1

---------




* ATA-Trim support (#2781)

* Add files via upload

* Update main.cpp

* Update main.cpp

* Update main.cpp

* Update Helper.cpp

* Update Helper.h

* Update CMakeLists.txt

* Update CMakeLists.txt

* Update diskio_sdmmc_mh.c

* Update diskio_sdmmc_mh.h

* Update ff_mh.c

* Update vfs_fat_sdmmc_mh.c

* Update sdmmc_common_mh.h

* Update sdmmc_common_mh.c

* Update Helper.cpp

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update ff_mh.c

---------



* Update Changelog.md

* Add device info info page (#2789)



* fix chip information REST API

* Update pipelines to fix warnings (#2841)

Update actions

* Move param doc (#2843)

moved param doc from docs repo to here

* fix(tflite): Fix memory leaks in tflite integration (#2842)

* Update Changelog.md (#2849)

---------

Signed-off-by: Florian Grabmeier <flo.grabmeier@gmail.com>
Co-authored-by: jomjol <30766535+jomjol@users.noreply.github.com>
Co-authored-by: CaCO3 <caco@ruinelli.ch>
Co-authored-by: Slider0007 <jobbelle@gmx.net>
Co-authored-by: Frank Haverland <fspapaping@googlemail.com>
Co-authored-by: Martin Siggel <martin.siggel@dlr.de>
Co-authored-by: michael <Heinrich-Tuning@web.de>
Co-authored-by: flox_x <93255373+flooxo@users.noreply.github.com>
Co-authored-by: Slider0007 <115730895+Slider0007@users.noreply.github.com>

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -0,0 +1,67 @@
+name: 🐞 Bug Report
+description: Use this form if you think you found a bug / Verwende dieses Formular, wenn Du denkst, dass Du einen Fehler gefunden hast.
+labels: bug
+body:
+
+
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for taking your time to report a bug.
+
+        Before you proceed, please check:
+        - [ ] Do you use the [latest version](https://github.com/jomjol/AI-on-the-edge-device/releases)?
+        - [ ] Is there already another similar issue? Check for open and closed [Issue](https://github.com/jomjol/AI-on-the-edge-device/issues) and [Discussions](https://github.com/jomjol/AI-on-the-edge-device/discussions).
+        - [ ] Are instructions in the [README](https://github.com/jomjol/AI-on-the-edge-device/tree/master#readme) solving your issue?
+        - [ ] Are instructions in the [Wiki](https://github.com/jomjol/AI-on-the-edge-device/wiki) solving your issue?
+        
+        Du darfst gerne auch in Deutsch schreiben!
+
+
+
+  - type: textarea
+    validations:
+      required: true
+    attributes:
+      label: The Problem
+      description: A clear and concise description of what the bug is.
+
+
+
+  - type: input
+    validations:
+      required: true
+    attributes:
+      label: Version
+      description: Which version are you using? (See menu `System > Info`).    
+        
+        
+
+  - type: textarea
+    validations:
+      required: true
+    attributes:
+      label: Logfile
+      description: Add the logfile (See menu `System > Log Viewer`) to help explain your problem. This will be automatically formatted into code, so no need to format it on your side.
+      render: shell
+
+
+
+  - type: textarea
+    attributes:
+      label: Expected Behavior
+      description: A clear and concise description of what you expected to happen.
+        
+
+
+  - type: textarea
+    attributes:
+      label: Screenshots
+      description: If applicable, add screenshots to help explain your problem.
+        
+
+
+  - type: textarea
+    attributes:
+      label: Additional Context
+      description: Add any other context about the problem here.

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,5 @@
+blank_issues_enabled: false
+contact_links:
+  - name: 💬 Open a new Discussion
+    url: https://github.com/jomjol/AI-on-the-edge-device/discussions/categories/q-a
+    about: Use this form if you have a question or need help setting your AI-on-the-edge-device it up

.github/ISSUE_TEMPLATE/feature.yaml

@@ -0,0 +1,23 @@
+name: 💡 Feature Request
+description: Use this form if you have an idea or wish for a new feature
+labels: enhancement
+body:
+
+
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for taking your time to document your idea.
+
+        Before you proceed, please check:
+        - [ ] Is there already another similar issue open or closed? Check for open and closed [Issue](https://github.com/jomjol/AI-on-the-edge-device/issues) and [Discussions](https://github.com/jomjol/AI-on-the-edge-device/discussions).
+        - [ ] Is there already another similar [discussion](https://github.com/jomjol/AI-on-the-edge-device/discussions) ongoing?
+
+        Please be aware that we might decline your request if we have a reason for it!
+
+  - type: textarea
+    validations:
+      required: true
+    attributes:
+      label: The Feature
+      description: A clear and concise description of what the new feature should do. If possible, refer to other projects or add a mockup.

.github/ISSUE_TEMPLATE/training.yaml

@@ -0,0 +1,19 @@
+name: 📖 New Training Data
+description: Use this form if you collected images and want to provide them for training the model
+title: New Training Data
+labels: new training data
+assignees: jomjol
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Before you proceed, please check:
+        - [ ] Did you make sure we don't have similar data yet in the training data? (see [Digital Counters](https://jomjol.github.io/neural-network-digital-counter-readout/) resp. [Analog Needles](https://jomjol.github.io/neural-network-analog-needle-readout))
+        - [ ] Did you follow the guideline as documented in the [Wiki](https://github.com/jomjol/AI-on-the-edge-device/wiki/ROI-Configuration)? We will only be able to accept the files if they fulfill the rules!
+
+  - type: textarea
+    validations:
+      required: true
+    attributes:
+      label: Files
+      description: You can drag & drop your **zipped** images here

.github/ISSUE_TEMPLATE/x_plain.yaml

@@ -0,0 +1,8 @@
+name: 🗒️ Blank Form
+description: Use this form if none of the others fit. Please only use none of the others fit!
+
+body:
+  - type: textarea
+    attributes:
+      label: Issue
+  

.github/label-commenter-config.yml

@@ -0,0 +1,123 @@
+# Reply Bot Configuration
+# See https://github.com/peaceiris/actions-label-commenter
+# Make sure to also add the response to .github/workflows/reply-bot.yml!
+# Due to the way it works, you have to add each response twice, once for the issue, once for the discussions!
+  
+ labels:
+  #######################################################################
+  # Bot Response: Documentation
+  #######################################################################
+  - name: bot-reply Documentation
+    labeled:
+      issue:
+          body: |
+            Please have a look on our documentation: https://jomjol.github.io/AI-on-the-edge-device-docs
+      discussion:
+          body: |
+            Please have a look on our documentation: https://jomjol.github.io/AI-on-the-edge-device-docs
+
+  #######################################################################
+  # Bot Response: ROI setup
+  #######################################################################
+  - name: bot-reply ROI Setup
+    labeled:
+      issue:
+          body: |
+            Make sure to setup your ROIs properly. Have a look on our documentation: https://jomjol.github.io/AI-on-the-edge-device-docs/ROI-Configuration/#how-to-setup-the-digit-rois-perfectly
+      discussion:
+          body: |
+            Make sure to setup your ROIs properly. Have a look on our documentation: https://jomjol.github.io/AI-on-the-edge-device-docs/ROI-Configuration/#how-to-setup-the-digit-rois-perfectly
+
+
+  #######################################################################
+  # Bot Response: Logfile
+  #######################################################################
+  - name: bot-reply Logfile
+    labeled:
+      issue:
+          body: |
+            Please provide a logfile!
+            Make sure to first enable the `DEBUG` level in `Settings->Configuration->Debug->Logfile Log Level`!
+            Then wait until the issue arises again.
+            When you copy the log into here, please make sure to use **Fenced code blocks** by wrapping it into separate lines with ` ``` `, see https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks#fenced-code-blocks
+      discussion:
+          body: |
+            Please provide a logfile!
+            Make sure to first enable the `DEBUG` level in `Settings->Configuration->Debug->Logfile Log Level`!
+            Then wait until the issue arises again.
+            When you copy the log into here, please make sure to use **Fenced code blocks** by wrapping it into separate lines with ` ``` `, see https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks#fenced-code-blocks
+
+
+  #######################################################################
+  # Bot Response: Web Console
+  #######################################################################
+  - name: bot-reply Web Console
+    labeled:
+      issue:
+          body: |
+            You can use the [Web Console](https://jomjol.github.io/AI-on-the-edge-device/index.html) to get USB log from the device.
+            The USB log contains more information about the startup and operation of the device than the normal Web UI log
+            When you copy the log into herm, please make sure to use **Fenced code blocks** by wrapping it into separate lines with ` ``` `, see https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks#fenced-code-blocks
+      discussion:
+          body: |
+            You can use the [Web Console](https://jomjol.github.io/AI-on-the-edge-device/index.html) to get USB log from the device.
+            The USB log contains more information about the startup and operation of the device than the normal Web UI log
+            When you copy the log into herm, please make sure to use **Fenced code blocks** by wrapping it into separate lines with ` ``` `, see https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks#fenced-code-blocks
+
+
+  #######################################################################
+  # Bot Response: Properly Format Code
+  #######################################################################
+  - name: bot-reply Properly Format Code
+    labeled:
+      issue:
+          body: |
+            Please make sure to use **Fenced code blocks** by wrapping it into separate lines with ` ``` `, see https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks#fenced-code-blocks
+            This makes your code or log much easier to read!
+      discussion:
+          body: |
+            Please make sure to use **Fenced code blocks** by wrapping it into separate lines with ` ``` `, see https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks#fenced-code-blocks
+            This makes your code or log much easier to read!
+
+
+  #######################################################################
+  # Bot Response: Web Installer
+  #######################################################################
+  - name: bot-reply Web Installer
+    labeled:
+      issue:
+          body: |
+            You can use the [Web Installer](https://jomjol.github.io/AI-on-the-edge-device/index.html) install the firmware onto the ESP32.
+      discussion:
+          body: |
+            You can use the [Web Installer](https://jomjol.github.io/AI-on-the-edge-device/index.html) install the firmware onto the ESP32.
+
+
+  #######################################################################
+  # Bot Response: Rolling Build
+  #######################################################################
+  - name: bot-reply Rolling Build
+    labeled:
+      issue:
+          body: |
+            You can try the latest [Automatic Build](https://github.com/jomjol/AI-on-the-edge-device/actions/workflows/build.yaml?query=branch%3Arolling+event%3Apush) of the the `rolling` or any other branch. It might already contain a fix for your issue.
+            See the [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/rolling-installation) for additional information.
+      discussion:
+          body: |
+            You can try the latest [Automatic Build](https://github.com/jomjol/AI-on-the-edge-device/actions/workflows/build.yaml?query=branch%3Arolling+event%3Apush) of the the `rolling` or any other branch. It might already contain a fix for your issue.
+            See the [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/rolling-installation) for additional information.
+            
+            
+  #######################################################################
+  # Bot Response: Show Trained Digits/Pointers
+  #######################################################################
+  - name: bot-reply Show Trained Digits/Pointers
+    labeled:
+      issue:
+          body: |
+            See [Digital Digits](https://jomjol.github.io/neural-network-digital-counter-readout) resp. [Analogue Pointers](https://jomjol.github.io/neural-network-analog-needle-readout) for an overview of all trained data.
+            If your type is not contained it can be added to our training material, see [here](https://jomjol.github.io/AI-on-the-edge-device-docs/collect-new-images/).
+      discussion:
+          body: |
+            See [Digital Digits](https://jomjol.github.io/neural-network-digital-counter-readout) resp. [Analogue Pointers](https://jomjol.github.io/neural-network-analog-needle-readout) for an overview of all trained data.
+            If your type is not contained it can be added to our training material, see [here](https://jomjol.github.io/AI-on-the-edge-device-docs/collect-new-images/).

.github/workflows/build.yml

@@ -0,0 +1,429 @@
+name: Build and Pack
+
+on: [push, pull_request]
+
+jobs:
+#########################################################################################
+## Build Firmware
+#########################################################################################
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - id: skip_check
+      uses: fkirc/skip-duplicate-actions@v5
+      with:
+        concurrent_skipping: same_content_newer
+    
+    - uses: actions/checkout@v4
+      with:
+        submodules: recursive
+
+    - name: Set Variables
+      id: vars
+      run: |
+        echo "sha_short=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
+
+    - name: Update PIP cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: ~/.cache/pip
+        key: pip-${{ github.run_id }}
+        restore-keys: pip # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+
+    - name: Update PlatformIO cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: ~/.platformio
+        key: platformio-${{ github.run_id }}
+        restore-keys: platformio # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+          
+    - name: Update Build cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: ./code/.pio/
+        key: build-${{ github.run_id }}
+        restore-keys: build # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+          
+    - name: Update generated-files cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: |
+          ./code/.pio/build/esp32cam/firmware.bin
+          ./code/.pio/build/esp32cam/partitions.bin
+          ./code/.pio/build/esp32cam/bootloader.bin
+          ./html/*
+        key: generated-files-${{ github.run_id }}
+        restore-keys: generated-files # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache          
+
+    - name: Set up Python
+      uses: actions/setup-python@v5
+      with:
+        python-version: '3.10'
+
+    - name: Install PlatformIO
+      run: |
+        python -m pip install --upgrade pip
+        pip install --upgrade platformio
+
+    - name: Build Firmware
+      #run: echo "Testing... ${{ github.ref_name }}, ${{ steps.vars.outputs.sha_short }}" > ./sd-card/html/version.txt; mkdir -p ./code/.pio/build/esp32cam/; cd ./code/.pio/build/esp32cam/; echo "${{ steps.vars.outputs.sha_short }}" > firmware.bin; cp firmware.bin partitions.bin; cp firmware.bin bootloader.bin # Testing
+      run: cd code; platformio run --environment esp32cam
+
+    - name: Prepare Web UI (generate tooltip pages and update hashes in all files)
+      run: |
+        rm -rf ./html
+        mkdir html
+        cp -r ./sd-card/html/* ./html/
+
+        python -m pip install markdown
+        mkdir html/param-tooltips
+        cd tools/parameter-tooltip-generator
+        python generate-param-doc-tooltips.py
+        cd ../..
+
+        cp -r ./sd-card/html/* ./html/
+
+        echo "Replacing variables..."
+        cd html; find . -type f -exec sed -i 's/$COMMIT_HASH/${{ steps.vars.outputs.sha_short }}/g' {} \;
+        
+
+#########################################################################################
+## Pack for Update
+#########################################################################################
+  pack-for-update:
+    # New OTA concept
+    # update__version.zip file with following content:
+    #  - /firmware.bin
+    #  - (optional) /html/* (inkl. subfolders)
+    #  - (optional) /config/*.tfl
+    runs-on: ubuntu-latest
+    needs: build
+
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: Update generated-files cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: |
+          ./code/.pio/build/esp32cam/firmware.bin
+          ./code/.pio/build/esp32cam/partitions.bin
+          ./code/.pio/build/esp32cam/bootloader.bin
+          ./html/*
+        key: generated-files-${{ github.run_id }}
+        restore-keys: generated-files # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+
+    - name: Update update cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: update
+        key: update-${{ github.run_id }}
+        restore-keys: update # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+    
+    - name: Set Variables
+      id: vars
+      run: |
+        echo "sha_short=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
+        echo "branch=$(echo ${{ github.ref_name }} | tr / __)" >> $GITHUB_OUTPUT
+
+    - name: Prepare update__*.zip artifact
+      run: |
+        rm -rf ./update
+        mkdir -p ./update
+        cp "./code/.pio/build/esp32cam/firmware.bin" "update/firmware.bin"
+
+    - name: Add Web UI to update
+      run: cp -r ./html ./update/
+
+    - name: Add CNN to update
+      run: |
+        rm -rf ./update/config/
+        mkdir -p ./update/config/
+        cp ./sd-card/config/*.tfl ./update/config/ 2>/dev/null || true
+        cp ./sd-card/config/*.tflite ./update/config/ 2>/dev/null || true
+
+    - name: Upload update as update.zip artifact (Firmware + Web UI + CNN)
+      uses: actions/upload-artifact@v4
+      with:
+        name: "AI-on-the-edge-device__update__${{ steps.vars.outputs.branch }}_(${{ steps.vars.outputs.sha_short }})"
+        path: ./update/*
+
+
+
+#########################################################################################
+## Pack for Remote Setup 
+#########################################################################################
+  pack-for-remote_setup:
+    # New Remote Setup concept
+    # remote_setup__version.zip file with following content:
+    #  - /firmware.bin
+    #  - /html/* (inkl. subfolders)
+    #  - /config/*
+    runs-on: ubuntu-latest
+    needs: build
+
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: Update generated-files cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: |
+          ./code/.pio/build/esp32cam/firmware.bin
+          ./code/.pio/build/esp32cam/partitions.bin
+          ./code/.pio/build/esp32cam/bootloader.bin
+          ./html/*
+        key: generated-files-${{ github.run_id }}
+        restore-keys: generated-files # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+          
+    - name: Update remote_setup cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: remote_setup
+        key: remote_setup-${{ github.run_id }}
+        restore-keys: remote_setup # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+
+    - name: Set Variables
+      id: vars
+      run: |
+        echo "sha_short=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
+        echo "branch=$(echo ${{ github.ref_name }} | tr / __)" >> $GITHUB_OUTPUT
+
+    - name: Prepare remote_setup__*.zip artifact
+      run: |
+        rm -rf ./remote_setup
+        mkdir -p ./remote_setup
+
+    - name: Add Web UI to remote_setup
+      run: cp -r ./html ./remote_setup/
+
+    - name: Add whole config folder to remote_setup
+      run: |
+        rm -rf ./remote_setup/config/
+        mkdir -p ./remote_setup/config/
+        cp ./sd-card/config/* ./remote_setup/config/ 2>/dev/null || true
+
+    - name: Upload remote_setup as remote_setup.zip artifact (Firmware + Web UI + Config)
+      uses: actions/upload-artifact@v4
+      with:
+        name: "AI-on-the-edge-device__remote-setup__${{ steps.vars.outputs.branch }}_(${{ steps.vars.outputs.sha_short }})"
+        path: ./remote_setup/*
+
+
+#########################################################################################
+## Pack for a fresh install (USB flashing) (manual_setup)
+#########################################################################################
+  pack-for-manual_setup:
+    # creates old style binaries for fresh installation (backward compatible to wiki)
+    runs-on: ubuntu-latest
+    needs: build
+
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: Update generated-files cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: |
+          ./code/.pio/build/esp32cam/firmware.bin
+          ./code/.pio/build/esp32cam/partitions.bin
+          ./code/.pio/build/esp32cam/bootloader.bin
+          ./html/*
+        key: generated-files-${{ github.run_id }}
+        restore-keys: generated-files # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+
+    - name: Update manual_setup cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: manual_setup
+        key: manual_setup-${{ github.run_id }}
+        restore-keys: manual_setup # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+
+    - name: Set Variables
+      id: vars
+      run: |
+        echo "sha_short=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
+        echo "branch=$(echo ${{ github.ref_name }} | tr / __)" >> $GITHUB_OUTPUT
+
+    - name: Prepare manual_setup__*.zip artifact
+      run: |
+        rm -rf manual_setup
+        mkdir -p manual_setup
+        rm -rf manual_setup/*.zip
+        rm -rf release
+        mkdir -p release
+        # copy builds to manual_setup folder
+        cp -f "./code/.pio/build/esp32cam/firmware.bin" "manual_setup/firmware.bin"
+        cp -f "./code/.pio/build/esp32cam/bootloader.bin" "manual_setup/bootloader.bin"
+        cp -f "./code/.pio/build/esp32cam/partitions.bin" "manual_setup/partitions.bin"
+        rm -rf ./sd-card/html
+        cp -r ./html ./sd-card/ # Overwrite the Web UI with the preprocessed files
+        cd sd-card; zip -r ../manual_setup/sd-card.zip *; cd ..
+        cd ./manual_setup
+  
+    - name: Upload manual_setup.zip artifact (Firmware + Bootloader + Partitions + Web UI)
+      uses: actions/upload-artifact@v4
+      with:
+        name: "AI-on-the-edge-device__manual-setup__${{ steps.vars.outputs.branch }}_(${{ steps.vars.outputs.sha_short }})"
+        path: ./manual_setup
+
+
+#########################################################################################
+## Prepare and create release
+#########################################################################################
+  release: 
+    runs-on: ubuntu-latest
+    needs: [pack-for-update, pack-for-manual_setup, pack-for-remote_setup]
+    if: startsWith(github.ref, 'refs/tags/') 
+
+    # Sets permissions of the GITHUB_TOKEN to allow updating the branches
+    permissions:
+      contents: write
+      pages: write
+      id-token: write
+
+    steps:
+    - uses: actions/checkout@v4
+    
+    - name: Update update cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: update
+        key: update-${{ github.run_id }}
+        restore-keys: update # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+    
+    - name: Update remote_setup cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: remote_setup
+        key: remote_setup-${{ github.run_id }}
+        restore-keys: remote_setup # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+    
+    - name: Update manual_setup cache on every commit
+      uses: actions/cache@v4
+      with:
+        path: manual_setup
+        key: manual_setup-${{ github.run_id }}
+        restore-keys: manual_setup # This matches above key as it is only used as a prefix. it the restores the nearest cache, see https://github.com/restore-keys:/blob/main/tips-and-workarounds.md#update-a-cache
+
+    - name: Set Variables
+      id: vars
+      run: |
+        echo "sha_short=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
+        echo "branch=$(echo ${{ github.ref_name }} | tr / __)" >> $GITHUB_OUTPUT
+
+    - name: Prepare artifacts for release
+      run: |
+        rm -rf release
+        mkdir -p release
+
+        # create AI-on-the-edge-device__update__*.zip like "AI-on-the-edge-device__update__v13.0.5.zip"
+        cd ./update
+        zip -r ../release/AI-on-the-edge-device__update__${{ steps.vars.outputs.branch }}.zip .
+
+        # create AI-on-the-edge-device__manual-setup__*.zip like "AI-on-the-edge-device__manual-setup__v13.0.5.zip"
+        cd ../manual_setup
+        zip -r ../release/AI-on-the-edge-device__manual-setup__${{ steps.vars.outputs.branch }}.zip .
+
+        # create AI-on-the-edge-device__remote-setup__*.zip like "AI-on-the-edge-device__remote-setup__v13.0.5.zip"
+        cd ../remote_setup
+        zip -r ../release/AI-on-the-edge-device__remote-setup__${{ steps.vars.outputs.branch }}.zip .
+
+    # extract the version used in next step
+    - id: get_version
+      if: startsWith(github.ref, 'refs/tags/') 
+      uses: Simply007/get-version-action@v2
+
+#    # the changelog [unreleased] will now be changed to the release version
+#    - name: Update changelog
+#      uses: thomaseizinger/keep-a-changelog-new-release@v1
+#      if: startsWith(github.ref, 'refs/tags/') 
+#      with:
+#        changelogPath: Changelog.md
+#        version: ${{ steps.get_version.outputs.version-without-v }}
+            
+#    # the release notes will be extracted from changelog 
+#    - name: Extract release notes
+#      id: extract-release-notes
+#      if: startsWith(github.ref, 'refs/tags/') 
+#      uses: ffurrer2/extract-release-notes@v1
+#      with:
+#          changelog_file: Changelog.md
+
+    # Releases should only be created on master by tagging the last commit.
+    # all artifacts in firmware folder pushed to the release
+    - name: Release
+      uses: softprops/action-gh-release@v1
+      # Note:
+      # If you get the error "Resource not accessible by integration",
+      # The access rights are not sufficient, see
+      # https://github.com/softprops/action-gh-release/issues/232#issuecomment-1131379440
+      if: startsWith(github.ref, 'refs/tags/')
+      with:
+        name:  ${{ steps.get_version.outputs.version-without-v }}
+        body: ${{ steps.extract-release-notes.outputs.release_notes }}
+        files: |
+          release/*
+
+#    # Commit&Push Changelog to master branch. Must be manually merged back to rolling
+#    - name: Commit changes and push changes
+#      if: startsWith(github.ref, 'refs/tags/') 
+#      run: |
+#        git config user.name github-actions
+#        git config user.email github-actions@github.com
+#        git add Changelog.md
+#        git commit Changelog.md -m "Update Changelog.md for ${{github.event.inputs.versionIncrement}} release"
+#        git push origin HEAD:master
+
+
+#########################################################################################
+## Update the Web Installer on a release
+#########################################################################################
+# Make sure to also update update-webinstaller.yml!
+  update-web-installer:    
+    needs: [release]
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    runs-on: ubuntu-latest
+
+    # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
+    permissions:
+      contents: read
+      pages: write
+      id-token: write
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+      
+    - name: Get version of last release
+      id: last_release
+      uses: mindojo/get-latest-release@0b8ef1434d7468d6bffcc8263baff5c777f72321
+      with:
+        myToken: ${{ github.token }}
+        exclude_types: "draft|prerelease"
+        view_top: 1     
+      
+    - name: Add binary to Web Installer and update manifest
+      run: |
+        echo "Updating Web installer to use firmware from ${{ steps.last_release.outputs.tag_name }}..."
+        rm -f docs/binary/firmware.bin
+        wget https://github.com/jomjol/AI-on-the-edge-device/releases/download/${{ steps.last_release.outputs.tag_name }}/AI-on-the-edge-device__update__${{ steps.last_release.outputs.tag_name }}.zip
+        unzip AI-on-the-edge-device__update__${{ steps.last_release.outputs.tag_name }}.zip
+        cp -f firmware.bin docs/binary/firmware.bin
+        echo "Updating index and manifest file..."
+        sed -i 's/$VERSION/${{ steps.last_release.outputs.tag_name }}/g' docs/index.html
+        sed -i 's/$VERSION/${{ steps.last_release.outputs.tag_name }}/g' docs/manifest.json
+    - name: Setup Pages
+      uses: actions/configure-pages@v4
+
+    - name: Upload artifact
+      uses: actions/upload-pages-artifact@v2
+      with:
+        path: 'docs'
+
+    - name: Deploy to GitHub Pages
+      id: deployment
+      uses: actions/deploy-pages@v4

.github/workflows/clear_cache.yml

@@ -0,0 +1,30 @@
+name: Clear cache
+
+on:
+  workflow_dispatch:
+
+permissions:
+  actions: write
+
+jobs:
+  clear-cache:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clear cache
+        uses: actions/github-script@v6
+        with:
+          script: |
+            console.log("About to clear")
+            const caches = await github.rest.actions.getActionsCacheList({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+            })
+            for (const cache of caches.data.actions_caches) {
+              console.log(cache)
+              github.rest.actions.deleteActionsCacheById({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                cache_id: cache.id,
+              })
+            }
+            console.log("Clear completed")

.github/workflows/manual-update-webinstaller.yml

@@ -0,0 +1,63 @@
+# This updates the Web Installer with the files from the docs folder and the binary of the latest release
+# it only gets run on:
+#  - Manually triggered
+# Make sure to also update the lower part of build.yml!
+
+name: Manual Web Installer Update
+
+on:
+  workflow_dispatch: # Run on manual trigger
+#  push:
+#    branches:
+#    - rolling
+#    paths:
+#    - docs # The path filter somehow does not work, so lets run it on every change to rolling
+      
+jobs:
+  manually-update-web-installer:    
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    runs-on: ubuntu-latest
+
+    # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
+    permissions:
+      contents: read
+      pages: write
+      id-token: write
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v3
+      
+    - name: Get version of last release
+      id: last_release
+      uses: mindojo/get-latest-release@0b8ef1434d7468d6bffcc8263baff5c777f72321
+      with:
+        myToken: ${{ github.token }}
+        exclude_types: "draft|prerelease"
+        view_top: 1     
+      
+    - name: Add binary to Web Installer and update manifest
+      run: |
+        echo "Updating Web installer to use firmware from ${{ steps.last_release.outputs.tag_name }}..."
+        rm -f docs/binary/firmware.bin
+        wget https://github.com/jomjol/AI-on-the-edge-device/releases/download/${{ steps.last_release.outputs.tag_name }}/AI-on-the-edge-device__update__${{ steps.last_release.outputs.tag_name }}.zip
+        unzip AI-on-the-edge-device__update__${{ steps.last_release.outputs.tag_name }}.zip
+        cp -f firmware.bin docs/binary/firmware.bin
+        echo "Updating index and manifest file..."
+        sed -i 's/$VERSION/${{ steps.last_release.outputs.tag_name }}/g' docs/index.html
+        sed -i 's/$VERSION/${{ steps.last_release.outputs.tag_name }}/g' docs/manifest.json
+
+    - name: Setup Pages
+      uses: actions/configure-pages@v2
+
+    - name: Upload artifact
+      uses: actions/upload-pages-artifact@v1
+      with:
+        path: 'docs'
+
+    - name: Deploy to GitHub Pages
+      id: deployment
+      uses: actions/deploy-pages@v1   
+  

.github/workflows/reply-bot.yml

@@ -0,0 +1,79 @@
+# Reply Bot
+# It uses the configuration in .github/label-commenter-config.yml
+# See https://github.com/peaceiris/actions-label-commenter
+
+name: Reply-Bot
+
+on:
+  issues:
+    types: [labeled]
+  discussion:
+    types: [labeled]
+
+permissions:
+  contents: read
+  issues: write
+  pull-requests: write
+  discussions: write
+
+jobs:
+  comment:
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions/checkout@v3
+      
+      
+      ####################################################################
+      ## Remove labels again (issues only)
+      ## Make sure to also add the reply message to .github/label-commenter-config.yml!
+      ## This currently seems no longer to work due to changes on the actions-cool/issues-helper!
+      ####################################################################
+#      - name: Remove 'Logfile' label again (issues only)
+#        if: github.event.label.name == 'bot-reply Logfile'
+#        uses: actions-cool/issues-helper@v3
+#        with:
+#          actions: 'remove-labels'
+#          labels: 'bot-reply Logfile'
+#          
+#      - name: Remove 'Web Console' label again (issues only)
+#        if: github.event.label.name == 'bot-reply Web Console'
+#        uses: actions-cool/issues-helper@v3
+#        with:
+#          actions: 'remove-labels'
+#          labels: 'bot-reply Web Console'
+#
+#      - name: Remove 'Properly Format Code' label again (issues only)
+#        if: github.event.label.name == 'bot-reply Properly Format Code'
+#        uses: actions-cool/issues-helper@v3
+#        with:
+#          actions: 'remove-labels'
+#          labels: 'bot-reply Properly Format Code'
+#          
+#      - name: Remove 'Web Installer' label again (issues only)
+#        if: github.event.label.name == 'bot-reply Web Installer'
+#        uses: actions-cool/issues-helper@v3
+#        with:
+#          actions: 'remove-labels'
+#          labels: 'bot-reply Web Installer'
+#
+#      - name: Remove 'Rolling Build' label again (issues only)
+#        if: github.event.label.name == 'bot-reply Rolling Build'
+#        uses: actions-cool/issues-helper@v3
+#        with:
+#          actions: 'remove-labels'
+#          labels: 'bot-reply Rolling Build'
+#          
+#      - name: Remove 'Show Trained Digits/Pointers' label again (issues only)
+#        if: github.event.label.name == 'bot-reply Show Trained Digits/Pointers'
+#        uses: actions-cool/issues-helper@v3
+#        with:
+#          actions: 'remove-labels'
+#          labels: 'bot-reply Show Trained Digits/Pointers'
+      
+      ####################################################################
+      ## Write the response
+      ####################################################################
+      - name: Write Response
+        uses: peaceiris/actions-label-commenter@c2d00660c86f2b9ed0fb35b372c451558eba85b3
+        with:
+          repo-token: "${{ secrets.GITHUB_TOKEN }}"

.gitignore

@@ -4,8 +4,13 @@
 .code-workspace
 /sd-card/htm./.vscode/
 /code/build
+/code/.helper
 /sd-card/html/debug/
-
+/firmware/
+version.txt
+/dist/
+/dist_release/
+/dist_old_ota
 CMakeLists.txt.user
 CMakeCache.txt
 CMakeFiles
@@ -17,3 +22,6 @@ compile_commands.json
 CTestTestfile.cmake
 _deps
 code/edgeAI.code-workspace
+.DS_Store
+tools/parameter-tooltip-generator/html
+tools/parameter-tooltip-generator/AI-on-the-edge-device-docs

.gitmodules

@@ -0,0 +1,12 @@
+[submodule "code/components/esp32-camera"]
+	path = code/components/esp32-camera
+	url = https://github.com/espressif/esp32-camera.git
+[submodule "code/components/esp-nn"]
+	path = code/components/esp-nn
+	url = https://github.com/espressif/esp-nn.git
+[submodule "code/components/esp-tflite-micro"]
+	path = code/components/esp-tflite-micro
+	url = https://github.com/espressif/esp-tflite-micro.git
+[submodule "code/components/stb"]
+	path = code/components/stb
+	url = https://github.com/nothings/stb.git

FeatureRequest.md

@@ -2,15 +2,88 @@
 
 **There are a lot of ideas for further improvements, but only limited capacity on side of the developer.** Therefore I have created this page as a collection of ideas. 
 
-1. Who ever has a new idea can put it here, so it that it is not forgotten. 
+1. Whoever has a new idea can put it here, so that it is not forgotten. 
 
-2. Who ever has time, capacity and passion to support, can take any of the ideas and implement them. 
-   I will support and help where ever I can!
+2. Whoever has the time, capacity and passion to support the project can take any of the ideas and implement them. I will provide support and help wherever I can!
    
    
 
 ____
 
+
+#### #40 Trigger with cron like exact time slot
+
+* https://github.com/jomjol/AI-on-the-edge-device/issues/2470
+
+
+
+#### #39 upnp implementation to auto detect the device
+
+* https://github.com/jomjol/AI-on-the-edge-device/issues/2481
+
+  
+
+#### #38 Energy Saving
+
+* Deep sleep between recognition
+* https://github.com/jomjol/AI-on-the-edge-device/issues/2486
+
+
+
+#### #37 Auto init SD card
+
+* Fully implement the SD card handling (including formatting) into the firmware
+* https://github.com/jomjol/AI-on-the-edge-device/issues/2488Demo 
+
+#### #36 Run demo without camera
+
+Demo mode requires a working camera (if not, one receives a 'Cam bad' error). Would be nice to demo or play around on other ESP32 boards (or on ESP32-CAM boards when you broke the camera cable...).
+
+#### #35 Use the same model, but provide the image from a Smartphone Camera
+as reading the Electricity or Water meter every few minutues only delivers apparent accuracy (DE: "Scheingenauigkeit") you could just as well take a picture with your Smartphone evey so often (e.g. once a week when you are in the Basement anyway), then with some "semi clever" tricks pass this image to the model developed here, and the values then on to who ever needs them e.g. via MQTT.
+IMO: It is not needed to have that many readings (datapoints) as our behaviour (Use of electricity or water) doesn't vary that much, say, over a weeks time. The interpolation between weekly readings will give sufficient information on the power and/or water usage. 
+
+
+#### #34 implement state and Roi for water leak detection
+for example see Roi on the next picture..
+![grafik](https://user-images.githubusercontent.com/38385805/207858812-2a6ba41d-1a8c-4fa1-9b6a-53cdd113c106.png)
+in case of position change between the measurments set this state to true, if there is no change set it back to false.
+In a defined time window this movement can lead into an alarm state / water leak..
+haveing this state in the mqtt broker can trigger functions like closing the ater pipe walve and so on...
+
+
+
+#### #33 Implement MATTER protocoll
+
+* see [#1404](https://github.com/jomjol/AI-on-the-edge-device/issues/1404)
+
+#### #32 Add feature to correct misinterpreted value
+
+* If a value is misinterpreted, the user can manually correct the value.
+* The misinterpreted ROIs would be saved in a "training data" -folder on the SD-card
+* Stretch goal: make sending of saved training data as easy as pushing a button =)
+
+#### #31 Implement InfluxDB v2.x interface
+
+* Currently only InfluxDB v1.x is supportet, extend to v2.x
+* Remark: interface has changed
+* see [#1160](https://github.com/jomjol/AI-on-the-edge-device/issues/1160)
+
+#### #30 Support meter clock over
+
+* In case of meter clocking over, that is, reaching its max. value and starting over from 0,
+  accept the new value and calculate correctly the difference.
+  (see line 739 onwards in ClassFlowPostProcessing.cpp)
+
+#### ~~#29 Add favicon and use the hostname for the website~~- implemented v11.3.1
+
+* ~~https://github.com/jomjol/AI-on-the-edge-device/issues/927~~
+
+#### #28 Improved error handling for ROIs
+
+* In case a ROI is out of the image, there is no error message, but a non sense image is used
+* Implement a error message for wrong configuratioin of ROI
+
 #### #27 Use Homie Spec for Mqtt binding
 
 * Use the standardized Home Protocol for the Mqtt binding 
@@ -40,9 +113,9 @@ ____
 
   
 
-#### #22 Direct hint to the different neural network files in the other repositories
+#### ~~#22 Direct hint to the different neural network files in the other repositories~~- implemented >v11.3.1
 
-* https://github.com/jomjol/AI-on-the-edge-device/issues/644
+* ~~https://github.com/jomjol/AI-on-the-edge-device/issues/644~~
 
   
 
@@ -55,7 +128,10 @@ ____
 #### #20 Deep sleep and push mode
 
 * Let the device be normally in deep sleep state, and wake it up periodically to collect data and push it via MQTT or HTTP post.
+* Support ESP-NOW to reduce the overhead of connecting to wifi and mqtt 
+* the above should enable battery powered applications
 
+* An other way to set deep sleep would be to enable it in a specific period (at night).
   
 
 #### #19 Extended log informations
@@ -96,13 +172,13 @@ ____
 
 
 
-#### #14 Backup and restore option for configuration
+#### ~~#14 Backup and restore option for configuration~~- implemented v11.3.1
 
-* https://github.com/jomjol/AI-on-the-edge-device/issues/459
+* ~~https://github.com/jomjol/AI-on-the-edge-device/issues/459~~
 
-* Implement a zip file compression for store and restore
+* ~~Implement a zip file compression for store and restore~~
 
-* Update the html to handle it
+* ~~Update the html to handle it~~
 
   
 

README.md

@@ -1,107 +1,108 @@
-# AI-on-the-edge-device
+# Welcome to the AI-on-the-edge-device
+<img src="images/icon/watermeter.svg" width="100px">
 
-This is an example of Artificial Intelligence (AI) calculations on a very cheap hardware.
+Artificial intelligence based systems have become established in our everyday lives. Just think of speech or image recognition. Most of the systems rely on either powerful processors or a direct connection to the cloud for doing the calculations there. With the increasing power of modern processors, the AI systems are coming closer to the end user – which is usually called **edge computing**.
+Here, this edge computing is put into a practically oriented example, where an AI network is implemented on an ESP32 device so: **AI on the edge**.
 
-### Details on **function**, **installation** and **configuration** can be found on the **[Wiki Page](https://github.com/jomjol/AI-on-the-edge-device/wiki)**
+This project allows you to digitize your **analog** water, gas, power and other meters using cheap and easily available hardware.
 
-A 3d-printable housing can be found here:
-  - https://www.thingiverse.com/thing:4573481 (Water Meter)
-  - https://www.thingiverse.com/thing:5028229 (Power Meter)
-  - https://www.thingiverse.com/thing:4571627 (ESP32-Cam housing only)
+All you need is an [ESP32 board with a supported camera](https://jomjol.github.io/AI-on-the-edge-device-docs/Hardware-Compatibility/) and something of a practical hand.
 
+<img src="images/esp32-cam.png" width="200px">
+
+## Key features
+- Tensorflow Lite (TFlite) integration – including easy-to-use wrapper
+- Inline image processing (feature detection, alignment, ROI extraction)
+- **Small** and **cheap** device (3 x 4.5 x 2 cm³, < 10 EUR)
+- Integrated camera and illumination
+- Web interface for administration and control
+- OTA interface for updating directly via the web interface
+- Full integration into Homeassistant
+- Support for Influx DB 1 and 2
+- MQTT
+- REST API
+
+## Workflow
+The device takes a photo of your meter at a defined interval. It then extracts the Regions of Interest (ROIs) from the image and runs them through artificial intelligence. As a result, you get the digitized value of your meter.
+
+There are several options for what to do with that value. Either send it to an MQTT broker, write it to an InfluxDb or simply provide access to it via a REST API.
+
+<img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/idea.jpg" width="600"> 
+
+## Impressions
+### AI-on-the-edge-device on a Water Meter
 <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/watermeter_all.jpg" width="200"><img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/main.jpg" width="200"><img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/size.png" width="200"> 
 
+### Web Interface (Water Meter)
 <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/watermeter.jpg" width="600"> 
 
+### AI-on-the-edge-device on a Electrical Power Meter
 <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/powermeter.jpg" width="600"> 
 
 
+## Setup
+There is growing [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/) which provides you with a lot of information. Head there to get a start, set it up and configure it.
 
+There are also articles in the German Heise magazine "make:" about the setup and technical background (behind a paywall): [DIY - Setup](https://www.heise.de/select/make/2021/2/2103513300897420296)
+
+A lot of people created useful Youtube videos which might help you getting started.
+Here a small selection:
+
+- [youtube.com/watch?v=HKBofb1cnNc](https://www.youtube.com/watch?v=HKBofb1cnNc)
+- [youtube.com/watch?v=yyf0ORNLCk4](https://www.youtube.com/watch?v=yyf0ORNLCk4)
+- [youtube.com/watch?v=XxmTubGek6M](https://www.youtube.com/watch?v=XxmTubGek6M)
+- [youtube.com/watch?v=mDIJEyElkAU](https://www.youtube.com/watch?v=mDIJEyElkAU)
+- [youtube.com/watch?v=SssiPkyKVVs](https://www.youtube.com/watch?v=SssiPkyKVVs)
+- [youtube.com/watch?v=MAHE_QyHZFQ](https://www.youtube.com/watch?v=MAHE_QyHZFQ)
+- [youtube.com/watch?v=Uap_6bwtILQ](https://www.youtube.com/watch?v=Uap_6bwtILQ)
+
+For further background information, head to [Neural Networks](https://www.heise.de/select/make/2021/6/2126410443385102621), [Training Neural Networks](https://www.heise.de/select/make/2022/1/2134114065999161585) and [Programming on the ESP32](https://www.heise.de/select/make/2022/2/2204010051597422030).
+
+### Download
+The latest available version can be found on the [Releases page](https://github.com/jomjol/AI-on-the-edge-device/releases).
+
+### Flashing the ESP32
+Initially you will have to flash the ESP32 via a USB connection. Later updates are possible directly over the air (OTA using WIFI).
+
+There are different ways to flash your ESP32:
+- The prefered way is the [Web Installer and Console](https://jomjol.github.io/AI-on-the-edge-device/index.html) which is a browser-based tool to flash the ESP32 and extract the log over USB:
+![](images/web-installer.png)
+- Flash Tool from Espressif
+- ESPtool (command-line tool)
+
+See the [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/Installation/) for more information.
+
+### Flashing the SD Card
+The SD card can be setup automatically after the firmware got installed. See the [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/Installation/#remote-setup-using-the-built-in-access-point) for details. For this to work, the SD card must be FAT formated (which is the default on a new SD card).
+Alternatively the SD card still can be setup manually, see the [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/Installation/#3-sd-card) for details!
+
+## Casing
+Various 3D-printable housing can be found here:
+  - https://www.thingiverse.com/thing:4573481 (Water Meter)
+  - https://www.thingiverse.com/thing:5028229 (Power Meter)
+  - https://www.thingiverse.com/thing:5224101 (Gas Meter)
+  - https://www.thingiverse.com/thing:4571627 (ESP32-cam housing only)
 
 ## Donate
+If you would like to support the developer with a cup of coffee, you can do that via [PayPal](https://www.paypal.com/donate?hosted_button_id=8TRSVYNYKDSWL).
 
-------
+<a href="https://www.paypal.com/donate?hosted_button_id=8TRSVYNYKDSWL"><img border="0" src="images/paypal.png" width="200px" target="_blank"></a>
 
-If you would like to support the developer with a cup of coffee you can do that via [Paypal](https://www.paypal.com/donate?hosted_button_id=8TRSVYNYKDSWL).
+## Support
+If you have any technical problems please search the [discussions](https://github.com/jomjol/AI-on-the-edge-device/discussions). In case you found a ug or have a feature request, please open an [issue](https://github.com/jomjol/AI-on-the-edge-device/issues).
 
-<form action="https://www.paypal.com/donate" method="post" target="_top">
-<input type="hidden" name="hosted_button_id" value="8TRSVYNYKDSWL" />
-<input type="image" src="https://www.paypalobjects.com/en_US/DK/i/btn/btn_donateCC_LG.gif" border="0" name="submit" title="PayPal - The safer, easier way to pay online!" alt="Donate with PayPal button" />
-<img alt="" border="0" src="https://www.paypal.com/en_DE/i/scr/pixel.gif" width="1" height="1" />
-</form>
-If you have any technical topics, you can file a issue in this repository. 
+In other cases you can contact the developer via email: <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/mail.jpg" height="25">
 
-In other cases you can contact the developer via email: <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/mail.jpg" height="25"> 
+## Changes and History 
+See [Changelog](Changelog.md).
 
-
-------
-## Change log
-**General remark:** Besides the file `firmware.bin`, typically the content of `/html` will need to be updated!
-
-------
-
-##### 11.0.0 - Intermediate Digits
-
-- Implementation of new CNN types to detect intermediate values of digits with rolling numbers
-
-  - By default the old algo (0, 1, ..., 9, "N") is active (due to the limited types of digits trained so far)
-  - Activation can be done by selection a tflite file with the new trained model in the 'config.ini'
-  - **Details can be found in the [wiki](https://github.com/jomjol/AI-on-the-edge-device/wiki/Neural-Network-Types)** (different types, trained image types, naming convention)
-
-- Updated  neural network files (and adaption to new naming convention)
-
-- Published a tool to download and combine log files - **Thanks to **
-
-  - Files see ['/tools/logfile-tool'](tbd), How-to see [wiki](https://github.com/jomjol/AI-on-the-edge-device/wiki/Gasmeter-Log-Downloader)
-
-- Bug Fix: InfluxDB enabling in grahic configuration
-
-  
+## Build It Yourself
+See [Build Instructions](code/README.md).
 
 ## Tools
-
 * Logfile downloader and combiner (Thx to [reserve85](https://github.com/reserve85))
-  * Files see ['/tools/logfile-tool'](tbd), How-to see [wiki](https://github.com/jomjol/AI-on-the-edge-device/wiki/Gasmeter-Log-Downloader)
-
-
+  * Files see ['/tools/logfile-tool'](tbd), how-to see [documentation](https://jomjol.github.io/AI-on-the-edge-device-docs/outdated--Gasmeter-Log-Downloader/)
 
 ## Additional Ideas
-
-There are some ideas and feature requests which are not followed currently - mainly due to capacity reasons on side of the developer. They are collected here: [FeatureRequest.md](FeatureRequest.md)
-
-
-
-------
-
-## History
-
-##### 10.6.2 - Stability Increase (2022-07-24)
-
-##### 9.2.0 - External Illumination (2021-12-02)
-
-##### 8.5.0 Multi Meter Support (2021-10-07)
-
-##### 7.1.2 MQTT-Update - (2021-06-17)
-
-##### 6.7.2 Image Processing in Memory - (2021-05-01)
-
-##### 5.0.0 Setup Modus - (2020-12-06)
-
-##### 4.1.1 Configuration editor - (2020-12-02)
-
-##### 4.0.0 Tflite Core - (2020-11-15)
-##### 3.1.0 MQTT-Client - (2020-10-26)
-
-##### 2.2.1 Version Control - (2020-09-27)
-
-
-##### 2.1.0 Decimal Shift, Chrome & Edge - (2020-09-25)
-
-
-##### 2.0.0 Layout update - (2020-09-12)
-
-##### 1.1.3 Initial Version - (2020-09-09)
-
-
-#### [Full Changelog](Changelog.md)
-
+There are some ideas and feature requests which are not currently being pursued – mainly due to capacity reasons on the part of the developers.
+They features are collected in the [issues](https://github.com/jomjol/AI-on-the-edge-device/issues) and in [FeatureRequest.md](FeatureRequest.md).

code/.gitignore

@@ -1,5 +1,16 @@
 .pio
+.idea
 .vscode/.browse.c_cpp.db*
 .vscode/c_cpp_properties.json
 .vscode/launch.json
 .vscode/ipch
+version.cpp
+sdkconfig.esp32cam
+sdkconfig.esp32cam-dev
+sdkconfig.esp32cam-debug
+sdkconfig.esp32cam-board-rev3
+sdkconfig.esp32cam-cpu-freq-240
+sdkconfig.esp32cam-board-rev3-cpu-freq-240
+sdkconfig.esp32cam-dev-himem
+sdkconfig.esp32cam-dev-task-analysis
+sdkconfig.esp32cam-no-softap

code/.helper/copy-esp-idf.bat

@@ -1,3 +0,0 @@
-copy "..\..\code\build\esp32cam-server-only.bin" "..\..\firmware\firmware.bin"
-copy "..\..\code\build\bootloader\bootloader.bin" "..\..\firmware\bootloader.bin"
-copy "..\..\code\build\partition_table\partition-table.bin" "..\..\firmware\partitions.bin"

code/.helper/makezip.bat

@@ -1 +1 @@
-powershell Compress-Archive "..\..\sd-card\html\*.*" "..\..\firmware\html.zip"
+powershell Compress-Archive -Path "..\..\sd-card\html\*.*" -DestinationPath "..\..\firmware\html.zip"

code/CMakeLists.txt

@@ -1,8 +1,6 @@
-cmake_minimum_required(VERSION 3.13.4)
+cmake_minimum_required(VERSION 3.16.0)
 
-list(APPEND EXTRA_COMPONENT_DIRS $ENV{IDF_PATH}/examples/common_components/protocol_examples_common)
-
-set(PROJECT_VER "0.0.9.3")
+list(APPEND EXTRA_COMPONENT_DIRS $ENV{IDF_PATH}/examples/common_components/protocol_examples_common components/esp-tflite-micro)
 
 ADD_CUSTOM_COMMAND(
     OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/version.cpp
@@ -10,6 +8,15 @@ ADD_CUSTOM_COMMAND(
     COMMAND ${CMAKE_COMMAND} -P
             ${CMAKE_CURRENT_SOURCE_DIR}/version.cmake)
 
+if(EXISTS "${SDKCONFIG}.defaults")
+    if(EXISTS "sdkconfig.defaults")
+        set(SDKCONFIG_DEFAULTS "${SDKCONFIG}.defaults;sdkconfig.defaults")
+        message(STATUS "-- Using defaults: ${SDKCONFIG_DEFAULTS} + sdkconfig.defaults")
+    else()
+        set(SDKCONFIG_DEFAULTS "${SDKCONFIG}.defaults")
+    endif()
+    message(STATUS "-- Using defaults: ${SDKCONFIG_DEFAULTS}")
+endif()
 
 include($ENV{IDF_PATH}/tools/cmake/project.cmake)
-project(esp32cam-server-only)
+project(AI-on-the-edge)

code/README.md

@@ -0,0 +1,74 @@
+# Build
+
+## Preparations
+```
+git clone https://github.com/jomjol/AI-on-the-edge-device.git
+cd AI-on-the-edge-device
+git checkout rolling
+git submodule update --init
+```
+
+## Update Submodules
+```
+cd /components/submodule-name (e.g. tflite-micro-example)
+git checkout VERSION (e.g. HASH of latest tflite-micro-example build)
+cd ../../ (auf Ebene von code)
+git submodule update --init
+```
+Evt. muss man vorher noch einige Verzeichnisse in compenents von Hand löschen, da sie beim checkout nicht gelöscht wurden (vor update -- init)
+
+## Build and Flash within terminal
+See further down to build it within an IDE.
+### Compile
+```
+cd code
+platformio run --environment esp32cam
+```
+
+### Upload
+```
+pio run --target upload --upload-port /dev/ttyUSB0
+```
+
+Alternatively you also can set the UART device in `platformio.ini`, eg. `upload_port = /dev/ttyUSB0`
+
+### Monitor UART Log
+```
+pio device monitor -p /dev/ttyUSB0
+```
+
+## Build and Flash with Visual Code IDE
+
+- Download and install VS Code
+  - https://code.visualstudio.com/Download
+- Install the VS Code platform io plugin
+  - <img src="https://raw.githubusercontent.com/jomjol/ai-on-the-edge-device/master/images/platformio_plugin.jpg" width="200" align="middle">
+  - Check for error messages, maybe you need to manually add some python libraries
+    - e.g. in my Ubuntu a python3-env was missing: `sudo apt-get install python3-venv`
+- git clone this project
+  - in Linux: 
+
+    ```
+    git clone https://github.com/jomjol/AI-on-the-edge-device.git
+    cd AI-on-the-edge-device
+    git checkout rolling
+    git submodule update --init
+    ```
+
+- in VS code, open the `AI-on-the-edge-device/code` 
+	- from terminal: `cd AI-on-the-edge-device/code && code .`
+- open a pio terminal (click on the terminal sign in the bottom menu bar)
+- make sure you are in the `code` directory
+- To build, type  `platformio run --environment esp32cam`
+  - or use the graphical interface:
+    <img src="https://raw.githubusercontent.com/jomjol/ai-on-the-edge-device/master/images/platformio_build.jpg" width="200" align="middle">
+  - the build artifacts are stored in  `code/.pio/build/esp32cam/`
+- Connect the device and type `pio device monitor`. There you will see your device and can copy the name to the next instruction
+- Add `upload_port = you_device_port` to the `platformio.ini` file
+- make sure an sd card with the contents of the `sd_card` folder is inserted and you have changed the wifi details
+- `pio run --target erase` to erase the flash
+- `pio run --target upload` this will upload the `bootloader.bin, partitions.bin,firmware.bin` from the `code/.pio/build/esp32cam/` folder. 
+- `pio device monitor` to observe the logs via uart
+
+# Update Parameters
+If you create or rename a parameter, make sure to update its documentation in `../param-docs/parameter-pages`! Check the `../param-docs/README.md` for more information.

code/components/esp-fatfs/CMakeLists.txt

@@ -0,0 +1,18 @@
+set(srcs "diskio/diskio_mh.c"
+         "diskio/diskio_rawflash_mh.c"
+         "diskio/diskio_sdmmc_mh.c"
+         "diskio/diskio_wl_mh.c"
+         "src/ff_mh.c"
+         "port/freertos/ffsystem_mh.c"
+         "src/ffunicode_mh.c"
+         "vfs/vfs_fat_mh.c"
+         "vfs/vfs_fat_sdmmc_mh.c"
+         "vfs/vfs_fat_spiflash_mh.c")
+
+idf_component_register(SRCS ${srcs}
+                       INCLUDE_DIRS "." diskio vfs src
+                       REQUIRES wear_levelling esp-sdmmc
+                       PRIV_REQUIRES vfs spi_flash
+                      )
+
+target_compile_options(${COMPONENT_LIB} PRIVATE "-Wno-format")

code/components/esp-fatfs/README.md

@@ -0,0 +1,278 @@
+# AIOTED related changes, see https://github.com/jomjol/AI-on-the-edge-device/pull/2781
+These files/folders were copied from `framework-espidf@3.50002.230601/components/` and adapted to our own needs.
+Since not every SD/MMC was recognized and this was due to the implementation of ATA trim support, this was revised.
+Furthermore, files that we don't need were deleted from it.
+
+## The most relevant changes are:
+### fatfs/diskio/diskio_sdmmc.c
+DRESULT ff_sdmmc_ioctl (BYTE pdrv, BYTE cmd, void* buff), at lines 106 to 110 changed from:
+```c
+#if FF_USE_TRIM
+        case CTRL_TRIM:
+            return ff_sdmmc_trim (pdrv, *((DWORD*)buff), //start_sector
+                    (*((DWORD*)buff + 1) - *((DWORD*)buff) + 1)); //sector_count
+#endif //FF_USE_TRIM
+```
+to:
+```c
+#if (FF_USE_TRIM)
+        case CTRL_TRIM:
+            if(FF_CAN_TRIM){
+                return ff_sdmmc_trim (pdrv, *((DWORD*)buff), //start_sector
+                    (*((DWORD*)buff + 1) - *((DWORD*)buff) + 1)); //sector_count
+            }
+            else{
+                return RES_ERROR;
+            }
+#endif //FF_USE_TRIM
+```
+
+### fatfs/src/ff.c
+added:
+```c
+#include "sdmmc_cmd.h"
+```
+
+static FRESULT remove_chain(FFOBJID* obj, DWORD clst, DWORD pclst), at lines 1437 to 1454 changed from:
+```c
+#if FF_FS_EXFAT || FF_USE_TRIM
+		if (ecl + 1 == nxt) {	/* Is next cluster contiguous? */
+			ecl = nxt;
+		} else {				/* End of contiguous cluster block */
+#if FF_FS_EXFAT
+			if (fs->fs_type == FS_EXFAT) {
+				res = change_bitmap(fs, scl, ecl - scl + 1, 0);	/* Mark the cluster block 'free' on the bitmap */
+				if (res != FR_OK) return res;
+			}
+#endif
+#if FF_USE_TRIM
+			rt[0] = clst2sect(fs, scl);					/* Start of data area to be freed */
+			rt[1] = clst2sect(fs, ecl) + fs->csize - 1;	/* End of data area to be freed */
+			disk_ioctl(fs->pdrv, CTRL_TRIM, rt);		/* Inform storage device that the data in the block may be erased */
+#endif
+			scl = ecl = nxt;
+		}
+#endif
+```
+
+to:
+```c
+#if FF_FS_EXFAT || FF_USE_TRIM
+		if(FF_FS_EXFAT || FF_CAN_TRIM){
+			if (ecl + 1 == nxt) {	/* Is next cluster contiguous? */
+				ecl = nxt;
+			} 
+			else {				/* End of contiguous cluster block */
+#if FF_FS_EXFAT
+				if (fs->fs_type == FS_EXFAT) {
+					res = change_bitmap(fs, scl, ecl - scl + 1, 0);	/* Mark the cluster block 'free' on the bitmap */
+					if (res != FR_OK) return res;
+				}
+#endif
+#if FF_USE_TRIM
+				if(FF_CAN_TRIM){
+					rt[0] = clst2sect(fs, scl);					/* Start of data area to be freed */
+					rt[1] = clst2sect(fs, ecl) + fs->csize - 1;	/* End of data area to be freed */
+					disk_ioctl(fs->pdrv, CTRL_TRIM, rt);		/* Inform storage device that the data in the block may be erased */
+				}
+#endif
+				scl = ecl = nxt;
+			}
+		}
+#endif
+```
+
+FRESULT f_mkfs(const TCHAR* path, const MKFS_PARM* opt, void* work, UINT len), at lines 5946 to 5949 changed from:
+```c
+#if FF_USE_TRIM
+		lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+		disk_ioctl(pdrv, CTRL_TRIM, lba);
+#endif
+```
+to:
+```c
+#if FF_USE_TRIM
+		if(FF_CAN_TRIM){
+			lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+			disk_ioctl(pdrv, CTRL_TRIM, lba);
+		}
+#endif
+```
+
+FRESULT f_mkfs(const TCHAR* path, const MKFS_PARM* opt, void* work, UINT len), at lines 6175 to 6178 changed from:
+```c
+#if FF_USE_TRIM
+		lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+		disk_ioctl(pdrv, CTRL_TRIM, lba);
+#endif
+```
+to:
+```c
+#if FF_USE_TRIM
+		if(FF_CAN_TRIM){
+			lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+			disk_ioctl(pdrv, CTRL_TRIM, lba);
+		}
+#endif
+```
+
+### sdmmc/sdmmc_cmd.c
+added:
+```c
+int FF_CAN_TRIM = 0;
+```
+
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card), at lines 630 to 636 changed from:
+```c
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.sec_feature & EXT_CSD_SEC_GB_CL_EN)) {
+        return ESP_OK;
+    }
+    return ESP_FAIL;
+}
+```
+to:
+```c
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.sec_feature & EXT_CSD_SEC_GB_CL_EN)) {
+        FF_CAN_TRIM = 1;
+        return ESP_OK;
+    }
+    FF_CAN_TRIM = 0;
+    return ESP_FAIL;
+}
+```
+
+### sdmmc/include/sdmmc_cmd.h
+
+added:
+```c
+extern int FF_CAN_TRIM;
+```
+
+# Espressif IoT Development Framework
+
+* [中文版](./README_CN.md)
+
+ESP-IDF is the development framework for Espressif SoCs supported on Windows, Linux and macOS.
+
+# ESP-IDF Release Support Schedule
+
+![Support Schedule](https://dl.espressif.com/dl/esp-idf/support-periods.svg)
+
+- Please read [the support policy](SUPPORT_POLICY.md) and [the documentation](https://docs.espressif.com/projects/esp-idf/en/latest/esp32/versions.html) for more information about ESP-IDF versions.
+- Please see the [End-of-Life Advisories](https://www.espressif.com/en/support/documents/advisories?keys=&field_type_of_advisory_tid%5B%5D=817) for information about ESP-IDF releases with discontinued support.
+
+# ESP-IDF Release and SoC Compatibility
+
+The following table shows ESP-IDF support of Espressif SoCs where ![alt text][preview] and ![alt text][supported] denote preview status and support, respectively. The preview support is usually limited in time and intended for beta versions of chips. Please use an ESP-IDF release where the desired SoC is already supported.
+
+|Chip         |          v4.1         |          v4.2          |         v4.3           |          v4.4          |          v5.0          |                                                                                      |
+|:----------- |:---------------------:| :---------------------:| :---------------------:| :---------------------:| :---------------------:|:------------------------------------------------------------------------------------ |
+|ESP32        |![alt text][supported] | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] |                                                                                      |
+|ESP32-S2     |                       | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] |                                                                                      |
+|ESP32-C3     |                       |                        | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] |                                                                                      |
+|ESP32-S3     |                       |                        |                        | ![alt text][supported] | ![alt text][supported] | [Announcement](https://www.espressif.com/en/news/ESP32_S3)                           |
+|ESP32-C2     |                       |                        |                        |                        | ![alt text][supported] | [Announcement](https://blog.espressif.com/esp32-c2-and-why-it-matter-s-bcf4d7d0b2c6) |
+|ESP32-H2     |                       |                        |                        | ![alt text][preview]   | ![alt text][preview]   | [Announcement](https://www.espressif.com/en/news/ESP32_H2)                           |
+
+[supported]: https://img.shields.io/badge/-supported-green "supported"
+[preview]: https://img.shields.io/badge/-preview-orange "preview"
+
+Espressif SoCs released before 2016 (ESP8266 and ESP8285) are supported by [RTOS SDK](https://github.com/espressif/ESP8266_RTOS_SDK) instead.
+
+# Developing With ESP-IDF
+
+## Setting Up ESP-IDF
+
+See https://idf.espressif.com/ for links to detailed instructions on how to set up the ESP-IDF depending on chip you use.
+
+**Note:** Each SoC series and each ESP-IDF release has its own documentation. Please see Section [Versions](https://docs.espressif.com/projects/esp-idf/en/latest/esp32/versions.html) on how to find documentation and how to checkout specific release of ESP-IDF.
+
+### Non-GitHub forks
+
+ESP-IDF uses relative locations as its submodules URLs ([.gitmodules](.gitmodules)). So they link to GitHub. If ESP-IDF is forked to a Git repository which is not on GitHub, you will need to run the script [tools/set-submodules-to-github.sh](tools/set-submodules-to-github.sh) after git clone.
+
+The script sets absolute URLs for all submodules, allowing `git submodule update --init --recursive` to complete. If cloning ESP-IDF from GitHub, this step is not needed.
+
+## Finding a Project
+
+As well as the [esp-idf-template](https://github.com/espressif/esp-idf-template) project mentioned in Getting Started, ESP-IDF comes with some example projects in the [examples](examples) directory.
+
+Once you've found the project you want to work with, change to its directory and you can configure and build it.
+
+To start your own project based on an example, copy the example project directory outside of the ESP-IDF directory.
+
+# Quick Reference
+
+See the Getting Started guide links above for a detailed setup guide. This is a quick reference for common commands when working with ESP-IDF projects:
+
+## Setup Build Environment
+
+(See the Getting Started guide listed above for a full list of required steps with more details.)
+
+* Install host build dependencies mentioned in the Getting Started guide.
+* Run the install script to set up the build environment. The options include `install.bat` or `install.ps1` for Windows, and `install.sh` or `install.fish` for Unix shells.
+* Run the export script on Windows (`export.bat`) or source it on Unix (`source export.sh`) in every shell environment before using ESP-IDF.
+
+## Configuring the Project
+
+* `idf.py set-target <chip_name>` sets the target of the project to `<chip_name>`. Run `idf.py set-target` without any arguments to see a list of supported targets.
+* `idf.py menuconfig` opens a text-based configuration menu where you can configure the project.
+
+## Compiling the Project
+
+`idf.py build`
+
+... will compile app, bootloader and generate a partition table based on the config.
+
+## Flashing the Project
+
+When the build finishes, it will print a command line to use esptool.py to flash the chip. However you can also do this automatically by running:
+
+`idf.py -p PORT flash`
+
+Replace PORT with the name of your serial port (like `COM3` on Windows, `/dev/ttyUSB0` on Linux, or `/dev/cu.usbserial-X` on MacOS. If the `-p` option is left out, `idf.py flash` will try to flash the first available serial port.
+
+This will flash the entire project (app, bootloader and partition table) to a new chip. The settings for serial port flashing can be configured with `idf.py menuconfig`.
+
+You don't need to run `idf.py build` before running `idf.py flash`, `idf.py flash` will automatically rebuild anything which needs it.
+
+## Viewing Serial Output
+
+The `idf.py monitor` target uses the [idf_monitor tool](https://docs.espressif.com/projects/esp-idf/en/latest/get-started/idf-monitor.html) to display serial output from Espressif SoCs. idf_monitor also has a range of features to decode crash output and interact with the device. [Check the documentation page for details](https://docs.espressif.com/projects/esp-idf/en/latest/get-started/idf-monitor.html).
+
+Exit the monitor by typing Ctrl-].
+
+To build, flash and monitor output in one pass, you can run:
+
+`idf.py flash monitor`
+
+## Compiling & Flashing Only the App
+
+After the initial flash, you may just want to build and flash just your app, not the bootloader and partition table:
+
+* `idf.py app` - build just the app.
+* `idf.py app-flash` - flash just the app.
+
+`idf.py app-flash` will automatically rebuild the app if any source files have changed.
+
+(In normal development there's no downside to reflashing the bootloader and partition table each time, if they haven't changed.)
+
+## Erasing Flash
+
+The `idf.py flash` target does not erase the entire flash contents. However it is sometimes useful to set the device back to a totally erased state, particularly when making partition table changes or OTA app updates. To erase the entire flash, run `idf.py erase-flash`.
+
+This can be combined with other targets, ie `idf.py -p PORT erase-flash flash` will erase everything and then re-flash the new app, bootloader and partition table.
+
+# Resources
+
+* Documentation for the latest version: https://docs.espressif.com/projects/esp-idf/. This documentation is built from the [docs directory](docs) of this repository.
+
+* The [esp32.com forum](https://esp32.com/) is a place to ask questions and find community resources.
+
+* [Check the Issues section on github](https://github.com/espressif/esp-idf/issues) if you find a bug or have a feature request. Please check existing Issues before opening a new one.
+
+* If you're interested in contributing to ESP-IDF, please check the [Contributions Guide](https://docs.espressif.com/projects/esp-idf/en/latest/contribute/index.html).

code/components/esp-fatfs/diskio/diskio_impl_mh.h

@@ -0,0 +1,72 @@
+// Copyright 2017-2019 Espressif Systems (Shanghai) PTE LTD
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdint.h>
+typedef unsigned int UINT;
+typedef unsigned char BYTE;
+typedef uint32_t DWORD;
+
+#define FF_DRV_NOT_USED 0xFF
+
+#include "diskio_mh.h"
+#include "esp_err.h"
+
+/**
+ * Structure of pointers to disk IO driver functions.
+ *
+ * See FatFs documentation for details about these functions
+ */
+typedef struct {
+    DSTATUS (*init) (unsigned char pdrv);    /*!< disk initialization function */
+    DSTATUS (*status) (unsigned char pdrv);  /*!< disk status check function */
+    DRESULT (*read) (unsigned char pdrv, unsigned char* buff, uint32_t sector, unsigned count);  /*!< sector read function */
+    DRESULT (*write) (unsigned char pdrv, const unsigned char* buff, uint32_t sector, unsigned count);   /*!< sector write function */
+    DRESULT (*ioctl) (unsigned char pdrv, unsigned char cmd, void* buff); /*!< function to get info about disk and do some misc operations */
+} ff_diskio_impl_t;
+
+/**
+ * Register or unregister diskio driver for given drive number.
+ *
+ * When FATFS library calls one of disk_xxx functions for driver number pdrv,
+ * corresponding function in discio_impl for given pdrv will be called.
+ *
+ * @param pdrv drive number
+ * @param discio_impl   pointer to ff_diskio_impl_t structure with diskio functions
+ *                      or NULL to unregister and free previously registered drive
+ */
+void ff_diskio_register(BYTE pdrv, const ff_diskio_impl_t* discio_impl);
+
+#define ff_diskio_unregister(pdrv_) ff_diskio_register(pdrv_, NULL)
+
+
+/**
+ * Get next available drive number
+ *
+ * @param   out_pdrv            pointer to the byte to set if successful
+ *
+ * @return  ESP_OK              on success
+ *          ESP_ERR_NOT_FOUND   if all drives are attached
+ */
+esp_err_t ff_diskio_get_drive(BYTE* out_pdrv);
+
+
+#ifdef __cplusplus
+}
+#endif

code/components/esp-fatfs/diskio/diskio_mh.c

@@ -0,0 +1,117 @@
+/*-----------------------------------------------------------------------*/
+/* Low level disk I/O module skeleton for FatFs     (C)ChaN, 2016        */
+/* ESP-IDF port Copyright 2016 Espressif Systems (Shanghai) PTE LTD      */
+/*-----------------------------------------------------------------------*/
+/* If a working storage control module is available, it should be        */
+/* attached to the FatFs via a glue function rather than modifying it.   */
+/* This is an example of glue functions to attach various exsisting      */
+/* storage control modules to the FatFs module with a defined API.       */
+/*-----------------------------------------------------------------------*/
+
+#include <string.h>
+#include <time.h>
+#include <stdlib.h>
+#include <sys/time.h>
+#include "diskio_impl_mh.h"
+#include "ffconf_mh.h"
+#include "ff_mh.h"
+
+static ff_diskio_impl_t * s_impls[FF_VOLUMES] = { NULL };
+
+#if FF_MULTI_PARTITION		/* Multiple partition configuration */
+const PARTITION VolToPart[FF_VOLUMES] = {
+    {0, 0},    /* Logical drive 0 ==> Physical drive 0, auto detection */
+    {1, 0},    /* Logical drive 1 ==> Physical drive 1, auto detection */
+#if FF_VOLUMES > 2
+    {2, 0},     /* Logical drive 2 ==> Physical drive 2, auto detection */
+#endif
+#if FF_VOLUMES > 3
+    {3, 0},     /* Logical drive 3 ==> Physical drive 3, auto detection */
+#endif
+#if FF_VOLUMES > 4
+    {4, 0},     /* Logical drive 4 ==> Physical drive 4, auto detection */
+#endif
+#if FF_VOLUMES > 5
+    {5, 0},     /* Logical drive 5 ==> Physical drive 5, auto detection */
+#endif
+#if FF_VOLUMES > 6
+    {6, 0},     /* Logical drive 6 ==> Physical drive 6, auto detection */
+#endif
+#if FF_VOLUMES > 7
+    {7, 0},     /* Logical drive 7 ==> Physical drive 7, auto detection */
+#endif
+#if FF_VOLUMES > 8
+    {8, 0},     /* Logical drive 8 ==> Physical drive 8, auto detection */
+#endif
+#if FF_VOLUMES > 9
+    {9, 0},     /* Logical drive 9 ==> Physical drive 9, auto detection */
+#endif
+};
+#endif
+
+esp_err_t ff_diskio_get_drive(BYTE* out_pdrv)
+{
+    BYTE i;
+    for(i=0; i<FF_VOLUMES; i++) {
+        if (!s_impls[i]) {
+            *out_pdrv = i;
+            return ESP_OK;
+        }
+    }
+    return ESP_ERR_NOT_FOUND;
+}
+
+void ff_diskio_register(BYTE pdrv, const ff_diskio_impl_t* discio_impl)
+{
+    assert(pdrv < FF_VOLUMES);
+
+    if (s_impls[pdrv]) {
+        ff_diskio_impl_t* im = s_impls[pdrv];
+        s_impls[pdrv] = NULL;
+        free(im);
+    }
+
+    if (!discio_impl) {
+        return;
+    }
+
+    ff_diskio_impl_t * impl = (ff_diskio_impl_t *)malloc(sizeof(ff_diskio_impl_t));
+    assert(impl != NULL);
+    memcpy(impl, discio_impl, sizeof(ff_diskio_impl_t));
+    s_impls[pdrv] = impl;
+}
+
+DSTATUS ff_disk_initialize (BYTE pdrv)
+{
+    return s_impls[pdrv]->init(pdrv);
+}
+DSTATUS ff_disk_status (BYTE pdrv)
+{
+    return s_impls[pdrv]->status(pdrv);
+}
+DRESULT ff_disk_read (BYTE pdrv, BYTE* buff, LBA_t sector, UINT count)
+{
+    return s_impls[pdrv]->read(pdrv, buff, sector, count);
+}
+DRESULT ff_disk_write (BYTE pdrv, const BYTE* buff, LBA_t sector, UINT count)
+{
+    return s_impls[pdrv]->write(pdrv, buff, sector, count);
+}
+DRESULT ff_disk_ioctl (BYTE pdrv, BYTE cmd, void* buff)
+{
+    return s_impls[pdrv]->ioctl(pdrv, cmd, buff);
+}
+
+DWORD get_fattime(void)
+{
+    time_t t = time(NULL);
+    struct tm tmr;
+    localtime_r(&t, &tmr);
+    int year = tmr.tm_year < 80 ? 0 : tmr.tm_year - 80;
+    return    ((DWORD)(year) << 25)
+            | ((DWORD)(tmr.tm_mon + 1) << 21)
+            | ((DWORD)tmr.tm_mday << 16)
+            | (WORD)(tmr.tm_hour << 11)
+            | (WORD)(tmr.tm_min << 5)
+            | (WORD)(tmr.tm_sec >> 1);
+}

code/components/esp-fatfs/diskio/diskio_rawflash_mh.c

@@ -0,0 +1,99 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2022 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <string.h>
+#include "diskio_impl_mh.h"
+#include "ffconf_mh.h"
+#include "ff_mh.h"
+#include "esp_log.h"
+#include "diskio_rawflash_mh.h"
+#include "esp_compiler.h"
+#include "spi_flash_mmap.h"
+
+
+static const char* TAG = "diskio_rawflash";
+
+const esp_partition_t* ff_raw_handles[FF_VOLUMES];
+
+
+DSTATUS ff_raw_initialize (BYTE pdrv)
+{
+    return 0;
+}
+
+DSTATUS ff_raw_status (BYTE pdrv)
+{
+    return 0;
+}
+
+DRESULT ff_raw_read (BYTE pdrv, BYTE *buff, DWORD sector, UINT count)
+{
+    ESP_LOGV(TAG, "ff_raw_read - pdrv=%i, sector=%i, count=%in", (unsigned int)pdrv, (unsigned int)sector, (unsigned int)count);
+    const esp_partition_t* part = ff_raw_handles[pdrv];
+    assert(part);
+    esp_err_t err = esp_partition_read(part, sector * SPI_FLASH_SEC_SIZE, buff, count * SPI_FLASH_SEC_SIZE);
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "esp_partition_read failed (0x%x)", err);
+        return RES_ERROR;
+    }
+    return RES_OK;
+}
+
+
+DRESULT ff_raw_write (BYTE pdrv, const BYTE *buff, DWORD sector, UINT count)
+{
+    return RES_ERROR;
+}
+
+DRESULT ff_raw_ioctl (BYTE pdrv, BYTE cmd, void *buff)
+{
+    const esp_partition_t* part = ff_raw_handles[pdrv];
+    ESP_LOGV(TAG, "ff_raw_ioctl: cmd=%in", cmd);
+    assert(part);
+    switch (cmd) {
+        case CTRL_SYNC:
+            return RES_OK;
+        case GET_SECTOR_COUNT:
+            *((DWORD *) buff) = part->size / SPI_FLASH_SEC_SIZE;
+            return RES_OK;
+        case GET_SECTOR_SIZE:
+            *((WORD *) buff) = SPI_FLASH_SEC_SIZE;
+            return RES_OK;
+        case GET_BLOCK_SIZE:
+            return RES_ERROR;
+    }
+    return RES_ERROR;
+}
+
+
+esp_err_t ff_diskio_register_raw_partition(BYTE pdrv, const esp_partition_t* part_handle)
+{
+    if (pdrv >= FF_VOLUMES) {
+        return ESP_ERR_INVALID_ARG;
+    }
+    static const ff_diskio_impl_t raw_impl = {
+        .init = &ff_raw_initialize,
+        .status = &ff_raw_status,
+        .read = &ff_raw_read,
+        .write = &ff_raw_write,
+        .ioctl = &ff_raw_ioctl
+    };
+    ff_diskio_register(pdrv, &raw_impl);
+    ff_raw_handles[pdrv] = part_handle;
+    return ESP_OK;
+
+}
+
+
+BYTE ff_diskio_get_pdrv_raw(const esp_partition_t* part_handle)
+{
+    for (int i = 0; i < FF_VOLUMES; i++) {
+        if (part_handle == ff_raw_handles[i]) {
+            return i;
+        }
+    }
+    return 0xff;
+}

code/components/esp-fatfs/diskio/diskio_rawflash_mh.h

@@ -1,9 +1,9 @@
-// Copyright 2015-2016 Espressif Systems (Shanghai) PTE LTD
+// Copyright 2015-2018 Espressif Systems (Shanghai) PTE LTD
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
-//
+
 //     http://www.apache.org/licenses/LICENSE-2.0
 //
 // Unless required by applicable law or agreed to in writing, software
@@ -11,19 +11,27 @@
 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#ifndef _CONVERSIONS_YUV_H_
-#define _CONVERSIONS_YUV_H_
+
+#ifndef _DISKIO_RAWFLASH_DEFINED
+#define _DISKIO_RAWFLASH_DEFINED
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-#include <stdint.h>
+#include "esp_partition.h"
 
-void yuv2rgb(uint8_t y, uint8_t u, uint8_t v, uint8_t *r, uint8_t *g, uint8_t *b);
+/**
+ * Register spi flash partition
+ *
+ * @param pdrv  drive number
+ * @param part_handle  pointer to raw flash partition.
+ */
+esp_err_t ff_diskio_register_raw_partition(unsigned char pdrv, const esp_partition_t* part_handle);
+unsigned char ff_diskio_get_pdrv_raw(const esp_partition_t* part_handle);
 
 #ifdef __cplusplus
 }
 #endif
 
-#endif /* _CONVERSIONS_YUV_H_ */
+#endif // _DISKIO_RAWFLASH_DEFINED

code/components/esp-fatfs/diskio/diskio_sdmmc_mh.c

@@ -0,0 +1,147 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2021 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "diskio_impl_mh.h"
+#include "ffconf_mh.h"
+#include "ff_mh.h"
+#include "sdmmc_cmd_mh.h"
+#include "esp_log.h"
+#include "esp_compiler.h"
+
+static sdmmc_card_t* s_cards[FF_VOLUMES] = { NULL };
+static bool s_disk_status_check_en[FF_VOLUMES] = { };
+
+static const char* TAG = "diskio_sdmmc";
+
+//Check if SD/MMC card is present
+static DSTATUS ff_sdmmc_card_available(BYTE pdrv)
+{
+    sdmmc_card_t* card = s_cards[pdrv];
+    assert(card);
+    esp_err_t err = sdmmc_get_status(card);
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "Check status failed (0x%x)", err);
+        return STA_NOINIT;
+    }
+    return 0;
+}
+
+/**
+*   ff_sdmmc_status() and ff_sdmmc_initialize() return STA_NOINIT when sdmmc_get_status()
+*   fails. This error value is checked throughout the FATFS code.
+*   Both functions return 0 on success.
+*/
+DSTATUS ff_sdmmc_initialize (BYTE pdrv)
+{
+    return ff_sdmmc_card_available(pdrv);
+}
+
+DSTATUS ff_sdmmc_status(BYTE pdrv)
+{
+    if (s_disk_status_check_en[pdrv]) {
+        return ff_sdmmc_card_available(pdrv);
+    }
+    return 0;
+}
+
+DRESULT ff_sdmmc_read (BYTE pdrv, BYTE* buff, DWORD sector, UINT count)
+{
+    sdmmc_card_t* card = s_cards[pdrv];
+    assert(card);
+    esp_err_t err = sdmmc_read_sectors(card, buff, sector, count);
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "sdmmc_read_blocks failed (%d)", err);
+        return RES_ERROR;
+    }
+    return RES_OK;
+}
+
+DRESULT ff_sdmmc_write (BYTE pdrv, const BYTE* buff, DWORD sector, UINT count)
+{
+    sdmmc_card_t* card = s_cards[pdrv];
+    assert(card);
+    esp_err_t err = sdmmc_write_sectors(card, buff, sector, count);
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "sdmmc_write_blocks failed (%d)", err);
+        return RES_ERROR;
+    }
+    return RES_OK;
+}
+
+#if (FF_USE_TRIM)
+DRESULT ff_sdmmc_trim (BYTE pdrv, DWORD start_sector, DWORD sector_count)
+{
+    sdmmc_card_t* card = s_cards[pdrv];
+    assert(card);
+    sdmmc_erase_arg_t arg;
+
+    arg = sdmmc_can_discard(card) == ESP_OK ? SDMMC_DISCARD_ARG : SDMMC_ERASE_ARG;
+    esp_err_t err = sdmmc_erase_sectors(card, start_sector, sector_count, arg);
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "sdmmc_erase_sectors failed (%d)", err);
+        return RES_ERROR;
+    }
+    return RES_OK;
+}
+#endif //FF_USE_TRIM
+
+DRESULT ff_sdmmc_ioctl (BYTE pdrv, BYTE cmd, void* buff)
+{
+    sdmmc_card_t* card = s_cards[pdrv];
+    assert(card);
+    switch(cmd) {
+        case CTRL_SYNC:
+            return RES_OK;
+        case GET_SECTOR_COUNT:
+            *((DWORD*) buff) = card->csd.capacity;
+            return RES_OK;
+        case GET_SECTOR_SIZE:
+            *((WORD*) buff) = card->csd.sector_size;
+            return RES_OK;
+        case GET_BLOCK_SIZE:
+            return RES_ERROR;
+#if (FF_USE_TRIM)
+        case CTRL_TRIM:
+            if(FF_CAN_TRIM){
+                return ff_sdmmc_trim (pdrv, *((DWORD*)buff), //start_sector
+                    (*((DWORD*)buff + 1) - *((DWORD*)buff) + 1)); //sector_count
+            }
+            else{
+                return RES_ERROR;
+            }
+#endif //FF_USE_TRIM
+    }
+    return RES_ERROR;
+}
+
+void ff_sdmmc_set_disk_status_check(BYTE pdrv, bool enable)
+{
+    s_disk_status_check_en[pdrv] = enable;
+}
+
+void ff_diskio_register_sdmmc(BYTE pdrv, sdmmc_card_t* card)
+{
+    static const ff_diskio_impl_t sdmmc_impl = {
+        .init = &ff_sdmmc_initialize,
+        .status = &ff_sdmmc_status,
+        .read = &ff_sdmmc_read,
+        .write = &ff_sdmmc_write,
+        .ioctl = &ff_sdmmc_ioctl
+    };
+    s_cards[pdrv] = card;
+    s_disk_status_check_en[pdrv] = false;
+    ff_diskio_register(pdrv, &sdmmc_impl);
+}
+
+BYTE ff_diskio_get_pdrv_card(const sdmmc_card_t* card)
+{
+    for (int i = 0; i < FF_VOLUMES; i++) {
+        if (card == s_cards[i]) {
+            return i;
+        }
+    }
+    return 0xff;
+}

code/components/esp-fatfs/diskio/diskio_sdmmc_mh.h

@@ -0,0 +1,43 @@
+/*
+ * SPDX-FileCopyrightText: 2017-2022 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#pragma once
+
+#include "sdmmc_cmd_mh.h"
+
+#include "driver/sdmmc_defs.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @brief Enable/disable SD card status checking
+ *
+ * @param pdrv   drive number
+ * @param enable mock ff_sdmmc_status function (return 0)
+ */
+void ff_sdmmc_set_disk_status_check(BYTE pdrv, bool enable);
+
+/**
+ * Register SD/MMC diskio driver
+ *
+ * @param pdrv  drive number
+ * @param card  pointer to sdmmc_card_t structure describing a card; card should be initialized before calling f_mount.
+ */
+void ff_diskio_register_sdmmc(unsigned char pdrv, sdmmc_card_t* card);
+
+/**
+ * @brief Get the driver number corresponding to a card
+ *
+ * @param card The card to get its driver
+ * @return Driver number to the card
+ */
+BYTE ff_diskio_get_pdrv_card(const sdmmc_card_t* card);
+
+#ifdef __cplusplus
+}
+#endif

code/components/esp-fatfs/diskio/diskio_wl_mh.c

@@ -0,0 +1,118 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2021 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <string.h>
+#include "diskio_impl_mh.h"
+#include "ffconf_mh.h"
+#include "ff_mh.h"
+#include "esp_log.h"
+#include "diskio_wl_mh.h"
+#include "wear_levelling.h"
+#include "esp_compiler.h"
+
+static const char* TAG = "ff_diskio_spiflash";
+
+wl_handle_t ff_wl_handles[FF_VOLUMES] = {
+        [0 ... FF_VOLUMES - 1] = WL_INVALID_HANDLE
+};
+
+DSTATUS ff_wl_initialize (BYTE pdrv)
+{
+    return 0;
+}
+
+DSTATUS ff_wl_status (BYTE pdrv)
+{
+    return 0;
+}
+
+DRESULT ff_wl_read (BYTE pdrv, BYTE *buff, DWORD sector, UINT count)
+{
+    ESP_LOGV(TAG, "ff_wl_read - pdrv=%i, sector=%i, count=%i\n", (unsigned int)pdrv, (unsigned int)sector, (unsigned int)count);
+    wl_handle_t wl_handle = ff_wl_handles[pdrv];
+    assert(wl_handle + 1);
+    esp_err_t err = wl_read(wl_handle, sector * wl_sector_size(wl_handle), buff, count * wl_sector_size(wl_handle));
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "wl_read failed (%d)", err);
+        return RES_ERROR;
+    }
+    return RES_OK;
+}
+
+DRESULT ff_wl_write (BYTE pdrv, const BYTE *buff, DWORD sector, UINT count)
+{
+    ESP_LOGV(TAG, "ff_wl_write - pdrv=%i, sector=%i, count=%i\n", (unsigned int)pdrv, (unsigned int)sector, (unsigned int)count);
+    wl_handle_t wl_handle = ff_wl_handles[pdrv];
+    assert(wl_handle + 1);
+    esp_err_t err = wl_erase_range(wl_handle, sector * wl_sector_size(wl_handle), count * wl_sector_size(wl_handle));
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "wl_erase_range failed (%d)", err);
+        return RES_ERROR;
+    }
+    err = wl_write(wl_handle, sector * wl_sector_size(wl_handle), buff, count * wl_sector_size(wl_handle));
+    if (unlikely(err != ESP_OK)) {
+        ESP_LOGE(TAG, "wl_write failed (%d)", err);
+        return RES_ERROR;
+    }
+    return RES_OK;
+}
+
+DRESULT ff_wl_ioctl (BYTE pdrv, BYTE cmd, void *buff)
+{
+    wl_handle_t wl_handle = ff_wl_handles[pdrv];
+    ESP_LOGV(TAG, "ff_wl_ioctl: cmd=%i\n", cmd);
+    assert(wl_handle + 1);
+    switch (cmd) {
+    case CTRL_SYNC:
+        return RES_OK;
+    case GET_SECTOR_COUNT:
+        *((DWORD *) buff) = wl_size(wl_handle) / wl_sector_size(wl_handle);
+        return RES_OK;
+    case GET_SECTOR_SIZE:
+        *((WORD *) buff) = wl_sector_size(wl_handle);
+        return RES_OK;
+    case GET_BLOCK_SIZE:
+        return RES_ERROR;
+    }
+    return RES_ERROR;
+}
+
+
+esp_err_t ff_diskio_register_wl_partition(BYTE pdrv, wl_handle_t flash_handle)
+{
+    if (pdrv >= FF_VOLUMES) {
+        return ESP_ERR_INVALID_ARG;
+    }
+    static const ff_diskio_impl_t wl_impl = {
+        .init = &ff_wl_initialize,
+        .status = &ff_wl_status,
+        .read = &ff_wl_read,
+        .write = &ff_wl_write,
+        .ioctl = &ff_wl_ioctl
+    };
+    ff_wl_handles[pdrv] = flash_handle;
+    ff_diskio_register(pdrv, &wl_impl);
+    return ESP_OK;
+}
+
+BYTE ff_diskio_get_pdrv_wl(wl_handle_t flash_handle)
+{
+    for (int i = 0; i < FF_VOLUMES; i++) {
+        if (flash_handle == ff_wl_handles[i]) {
+            return i;
+        }
+    }
+    return 0xff;
+}
+
+void ff_diskio_clear_pdrv_wl(wl_handle_t flash_handle)
+{
+    for (int i = 0; i < FF_VOLUMES; i++) {
+        if (flash_handle == ff_wl_handles[i]) {
+            ff_wl_handles[i] = WL_INVALID_HANDLE;
+        }
+    }
+}

code/components/esp-fatfs/diskio/diskio_wl_mh.h

@@ -1,9 +1,9 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
+// Copyright 2015-2017 Espressif Systems (Shanghai) PTE LTD
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
-//
+
 //     http://www.apache.org/licenses/LICENSE-2.0
 //
 // Unless required by applicable law or agreed to in writing, software
@@ -12,35 +12,28 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#pragma once
-
-#if defined(CONFIG_NN_OPTIMIZED)
-#ifdef CONFIG_IDF_TARGET_ESP32S3
-#define ARCH_ESP32_S3 1
-#endif
-#ifdef CONFIG_IDF_TARGET_ESP32
-#define ARCH_ESP32 1
-#endif
-#endif
+#ifndef _DISKIO_WL_DEFINED
+#define _DISKIO_WL_DEFINED
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-/* reference kernels included by default */
-#include "esp_nn_ansi_headers.h"
+#include "wear_levelling.h"
 
-#if defined(CONFIG_NN_OPTIMIZED)
-#ifdef ARCH_ESP32_S3
-#include "esp_nn_esp32s3.h"
-#endif
-#ifdef ARCH_ESP32
-#include "esp_nn_esp32.h"
-#endif
-#else
-#include "esp_nn_ansi_c.h"
-#endif
+
+/**
+ * Register spi flash partition
+ *
+ * @param pdrv  drive number
+ * @param flash_handle  handle of the wear levelling partition.
+ */
+esp_err_t ff_diskio_register_wl_partition(unsigned char pdrv, wl_handle_t flash_handle);
+unsigned char ff_diskio_get_pdrv_wl(wl_handle_t flash_handle);
+void ff_diskio_clear_pdrv_wl(wl_handle_t flash_handle);
 
 #ifdef __cplusplus
 }
 #endif
+
+#endif // _DISKIO_WL_DEFINED

code/components/esp-fatfs/port/freertos/ffsystem_mh.c

@@ -0,0 +1,108 @@
+/*------------------------------------------------------------------------*/
+/* Sample Code of OS Dependent Functions for FatFs                        */
+/* (C)ChaN, 2017                                                          */
+/*------------------------------------------------------------------------*/
+
+
+#include <string.h>
+#include <stdlib.h>
+#include "ff_mh.h"
+#include "sdkconfig.h"
+#ifdef CONFIG_FATFS_ALLOC_PREFER_EXTRAM
+#include "esp_heap_caps.h"
+#endif
+
+void* ff_memalloc (    /* Returns pointer to the allocated memory block (null on not enough core) */
+    unsigned msize     /* Number of bytes to allocate */
+)
+{
+#ifdef CONFIG_FATFS_ALLOC_PREFER_EXTRAM
+    return heap_caps_malloc_prefer(msize, 2, MALLOC_CAP_DEFAULT | MALLOC_CAP_SPIRAM,
+                                            MALLOC_CAP_DEFAULT | MALLOC_CAP_INTERNAL);
+#else
+    return malloc(msize);
+#endif
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Free a memory block                                                    */
+/*------------------------------------------------------------------------*/
+
+void ff_memfree (
+    void* mblock    /* Pointer to the memory block to free (nothing to do for null) */
+)
+{
+    free(mblock);   /* Free the memory block with POSIX API */
+}
+
+
+
+
+#if FF_FS_REENTRANT    /* Mutal exclusion */
+
+/*------------------------------------------------------------------------*/
+/* Create a Synchronization Object                                        */
+/*------------------------------------------------------------------------*/
+/* This function is called in f_mount() function to create a new
+/  synchronization object for the volume, such as semaphore and mutex.
+/  When a 0 is returned, the f_mount() function fails with FR_INT_ERR.
+*/
+
+
+int ff_cre_syncobj (    /* 1:Function succeeded, 0:Could not create the sync object */
+    BYTE vol,           /* Corresponding volume (logical drive number) */
+    FF_SYNC_t *sobj     /* Pointer to return the created sync object */
+)
+{
+    *sobj = xSemaphoreCreateMutex();
+    return (*sobj != NULL) ? 1 : 0;
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Delete a Synchronization Object                                        */
+/*------------------------------------------------------------------------*/
+/* This function is called in f_mount() function to delete a synchronization
+/  object that created with ff_cre_syncobj() function. When a 0 is returned,
+/  the f_mount() function fails with FR_INT_ERR.
+*/
+
+int ff_del_syncobj (    /* 1:Function succeeded, 0:Could not delete due to an error */
+    FF_SYNC_t sobj      /* Sync object tied to the logical drive to be deleted */
+)
+{
+    vSemaphoreDelete(sobj);
+    return 1;
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Request Grant to Access the Volume                                     */
+/*------------------------------------------------------------------------*/
+/* This function is called on entering file functions to lock the volume.
+/  When a 0 is returned, the file function fails with FR_TIMEOUT.
+*/
+
+int ff_req_grant (    /* 1:Got a grant to access the volume, 0:Could not get a grant */
+    FF_SYNC_t sobj    /* Sync object to wait */
+)
+{
+    return (xSemaphoreTake(sobj, FF_FS_TIMEOUT) == pdTRUE) ? 1 : 0;
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Release Grant to Access the Volume                                     */
+/*------------------------------------------------------------------------*/
+/* This function is called on leaving file functions to unlock the volume.
+*/
+
+void ff_rel_grant (
+    FF_SYNC_t sobj    /* Sync object to be signaled */
+)
+{
+    xSemaphoreGive(sobj);
+}
+
+#endif // FF_FS_REENTRANT

code/components/esp-fatfs/src/00history.txt

@@ -0,0 +1,329 @@
+----------------------------------------------------------------------------
+  Revision history of FatFs module
+----------------------------------------------------------------------------
+
+R0.00 (February 26, 2006)
+
+  Prototype.
+
+
+
+R0.01 (April 29, 2006)
+
+  The first release.
+
+
+
+R0.02 (June 01, 2006)
+
+  Added FAT12 support.
+  Removed unbuffered mode.
+  Fixed a problem on small (<32M) partition.
+
+
+
+R0.02a (June 10, 2006)
+
+  Added a configuration option (_FS_MINIMUM).
+
+
+
+R0.03 (September 22, 2006)
+
+  Added f_rename().
+  Changed option _FS_MINIMUM to _FS_MINIMIZE.
+
+
+
+R0.03a (December 11, 2006)
+
+  Improved cluster scan algorithm to write files fast.
+  Fixed f_mkdir() creates incorrect directory on FAT32.
+
+
+
+R0.04 (February 04, 2007)
+
+  Added f_mkfs().
+  Supported multiple drive system.
+  Changed some interfaces for multiple drive system.
+  Changed f_mountdrv() to f_mount().
+
+
+
+R0.04a (April 01, 2007)
+
+  Supported multiple partitions on a physical drive.
+  Added a capability of extending file size to f_lseek().
+  Added minimization level 3.
+  Fixed an endian sensitive code in f_mkfs().
+
+
+
+R0.04b (May 05, 2007)
+
+  Added a configuration option _USE_NTFLAG.
+  Added FSINFO support.
+  Fixed DBCS name can result FR_INVALID_NAME.
+  Fixed short seek (<= csize) collapses the file object.
+
+
+
+R0.05 (August 25, 2007)
+
+  Changed arguments of f_read(), f_write() and f_mkfs().
+  Fixed f_mkfs() on FAT32 creates incorrect FSINFO.
+  Fixed f_mkdir() on FAT32 creates incorrect directory.
+
+
+
+R0.05a (February 03, 2008)
+
+  Added f_truncate() and f_utime().
+  Fixed off by one error at FAT sub-type determination.
+  Fixed btr in f_read() can be mistruncated.
+  Fixed cached sector is not flushed when create and close without write.
+
+
+
+R0.06 (April 01, 2008)
+
+  Added fputc(), fputs(), fprintf() and fgets().
+  Improved performance of f_lseek() on moving to the same or following cluster.
+
+
+
+R0.07 (April 01, 2009)
+
+  Merged Tiny-FatFs as a configuration option. (_FS_TINY)
+  Added long file name feature. (_USE_LFN)
+  Added multiple code page feature. (_CODE_PAGE)
+  Added re-entrancy for multitask operation. (_FS_REENTRANT)
+  Added auto cluster size selection to f_mkfs().
+  Added rewind option to f_readdir().
+  Changed result code of critical errors.
+  Renamed string functions to avoid name collision.
+
+
+
+R0.07a (April 14, 2009)
+
+  Septemberarated out OS dependent code on reentrant cfg.
+  Added multiple sector size feature.
+
+
+
+R0.07c (June 21, 2009)
+
+  Fixed f_unlink() can return FR_OK on error.
+  Fixed wrong cache control in f_lseek().
+  Added relative path feature.
+  Added f_chdir() and f_chdrive().
+  Added proper case conversion to extended character.
+
+
+
+R0.07e (November 03, 2009)
+
+  Septemberarated out configuration options from ff.h to ffconf.h.
+  Fixed f_unlink() fails to remove a sub-directory on _FS_RPATH.
+  Fixed name matching error on the 13 character boundary.
+  Added a configuration option, _LFN_UNICODE.
+  Changed f_readdir() to return the SFN with always upper case on non-LFN cfg.
+
+
+
+R0.08 (May 15, 2010)
+
+  Added a memory configuration option. (_USE_LFN = 3)
+  Added file lock feature. (_FS_SHARE)
+  Added fast seek feature. (_USE_FASTSEEK)
+  Changed some types on the API, XCHAR->TCHAR.
+  Changed .fname in the FILINFO structure on Unicode cfg.
+  String functions support UTF-8 encoding files on Unicode cfg.
+
+
+
+R0.08a (August 16, 2010)
+
+  Added f_getcwd(). (_FS_RPATH = 2)
+  Added sector erase feature. (_USE_ERASE)
+  Moved file lock semaphore table from fs object to the bss.
+  Fixed f_mkfs() creates wrong FAT32 volume.
+
+
+
+R0.08b (January 15, 2011)
+
+  Fast seek feature is also applied to f_read() and f_write().
+  f_lseek() reports required table size on creating CLMP.
+  Extended format syntax of f_printf().
+  Ignores duplicated directory separators in given path name.
+
+
+
+R0.09 (September 06, 2011)
+
+  f_mkfs() supports multiple partition to complete the multiple partition feature.
+  Added f_fdisk().
+
+
+
+R0.09a (August 27, 2012)
+
+  Changed f_open() and f_opendir() reject null object pointer to avoid crash.
+  Changed option name _FS_SHARE to _FS_LOCK.
+  Fixed assertion failure due to OS/2 EA on FAT12/16 volume.
+
+
+
+R0.09b (January 24, 2013)
+
+  Added f_setlabel() and f_getlabel().
+
+
+
+R0.10 (October 02, 2013)
+
+  Added selection of character encoding on the file. (_STRF_ENCODE)
+  Added f_closedir().
+  Added forced full FAT scan for f_getfree(). (_FS_NOFSINFO)
+  Added forced mount feature with changes of f_mount().
+  Improved behavior of volume auto detection.
+  Improved write throughput of f_puts() and f_printf().
+  Changed argument of f_chdrive(), f_mkfs(), disk_read() and disk_write().
+  Fixed f_write() can be truncated when the file size is close to 4GB.
+  Fixed f_open(), f_mkdir() and f_setlabel() can return incorrect value on error.
+
+
+
+R0.10a (January 15, 2014)
+
+  Added arbitrary strings as drive number in the path name. (_STR_VOLUME_ID)
+  Added a configuration option of minimum sector size. (_MIN_SS)
+  2nd argument of f_rename() can have a drive number and it will be ignored.
+  Fixed f_mount() with forced mount fails when drive number is >= 1. (appeared at R0.10)
+  Fixed f_close() invalidates the file object without volume lock.
+  Fixed f_closedir() returns but the volume lock is left acquired. (appeared at R0.10)
+  Fixed creation of an entry with LFN fails on too many SFN collisions. (appeared at R0.07)
+
+
+
+R0.10b (May 19, 2014)
+
+  Fixed a hard error in the disk I/O layer can collapse the directory entry.
+  Fixed LFN entry is not deleted when delete/rename an object with lossy converted SFN. (appeared at R0.07)
+
+
+
+R0.10c (November 09, 2014)
+
+  Added a configuration option for the platforms without RTC. (_FS_NORTC)
+  Changed option name _USE_ERASE to _USE_TRIM.
+  Fixed volume label created by Mac OS X cannot be retrieved with f_getlabel(). (appeared at R0.09b)
+  Fixed a potential problem of FAT access that can appear on disk error.
+  Fixed null pointer dereference on attempting to delete the root direcotry. (appeared at R0.08)
+
+
+
+R0.11 (February 09, 2015)
+
+  Added f_findfirst(), f_findnext() and f_findclose(). (_USE_FIND)
+  Fixed f_unlink() does not remove cluster chain of the file. (appeared at R0.10c)
+  Fixed _FS_NORTC option does not work properly. (appeared at R0.10c)
+
+
+
+R0.11a (September 05, 2015)
+
+  Fixed wrong media change can lead a deadlock at thread-safe configuration.
+  Added code page 771, 860, 861, 863, 864, 865 and 869. (_CODE_PAGE)
+  Removed some code pages actually not exist on the standard systems. (_CODE_PAGE)
+  Fixed errors in the case conversion teble of code page 437 and 850 (ff.c).
+  Fixed errors in the case conversion teble of Unicode (cc*.c).
+
+
+
+R0.12 (April 12, 2016)
+
+  Added support for exFAT file system. (_FS_EXFAT)
+  Added f_expand(). (_USE_EXPAND)
+  Changed some members in FINFO structure and behavior of f_readdir().
+  Added an option _USE_CHMOD.
+  Removed an option _WORD_ACCESS.
+  Fixed errors in the case conversion table of Unicode (cc*.c).
+
+
+
+R0.12a (July 10, 2016)
+
+  Added support for creating exFAT volume with some changes of f_mkfs().
+  Added a file open method FA_OPEN_APPEND. An f_lseek() following f_open() is no longer needed.
+  f_forward() is available regardless of _FS_TINY.
+  Fixed f_mkfs() creates wrong volume. (appeared at R0.12)
+  Fixed wrong memory read in create_name(). (appeared at R0.12)
+  Fixed compilation fails at some configurations, _USE_FASTSEEK and _USE_FORWARD.
+
+
+
+R0.12b (September 04, 2016)
+
+  Made f_rename() be able to rename objects with the same name but case.
+  Fixed an error in the case conversion teble of code page 866. (ff.c)
+  Fixed writing data is truncated at the file offset 4GiB on the exFAT volume. (appeared at R0.12)
+  Fixed creating a file in the root directory of exFAT volume can fail. (appeared at R0.12)
+  Fixed f_mkfs() creating exFAT volume with too small cluster size can collapse unallocated memory. (appeared at R0.12)
+  Fixed wrong object name can be returned when read directory at Unicode cfg. (appeared at R0.12)
+  Fixed large file allocation/removing on the exFAT volume collapses allocation bitmap. (appeared at R0.12)
+  Fixed some internal errors in f_expand() and f_lseek(). (appeared at R0.12)
+
+
+
+R0.12c (March 04, 2017)
+
+  Improved write throughput at the fragmented file on the exFAT volume.
+  Made memory usage for exFAT be able to be reduced as decreasing _MAX_LFN.
+  Fixed successive f_getfree() can return wrong count on the FAT12/16 volume. (appeared at R0.12)
+  Fixed configuration option _VOLUMES cannot be set 10. (appeared at R0.10c)
+
+
+
+R0.13 (May 21, 2017)
+
+  Changed heading character of configuration keywords "_" to "FF_".
+  Removed ASCII-only configuration, FF_CODE_PAGE = 1. Use FF_CODE_PAGE = 437 instead.
+  Added f_setcp(), run-time code page configuration. (FF_CODE_PAGE = 0)
+  Improved cluster allocation time on stretch a deep buried cluster chain.
+  Improved processing time of f_mkdir() with large cluster size by using FF_USE_LFN = 3.
+  Improved NoFatChain flag of the fragmented file to be set after it is truncated and got contiguous.
+  Fixed archive attribute is left not set when a file on the exFAT volume is renamed. (appeared at R0.12)
+  Fixed exFAT FAT entry can be collapsed when write or lseek operation to the existing file is done. (appeared at R0.12c)
+  Fixed creating a file can fail when a new cluster allocation to the exFAT directory occures. (appeared at R0.12c)
+
+
+
+R0.13a (October 14, 2017)
+
+  Added support for UTF-8 encoding on the API. (FF_LFN_UNICODE = 2)
+  Added options for file name output buffer. (FF_LFN_BUF, FF_SFN_BUF).
+  Added dynamic memory allocation option for working buffer of f_mkfs() and f_fdisk().
+  Fixed f_fdisk() and f_mkfs() create the partition table with wrong CHS parameters. (appeared at R0.09)
+  Fixed f_unlink() can cause lost clusters at fragmented file on the exFAT volume. (appeared at R0.12c)
+  Fixed f_setlabel() rejects some valid characters for exFAT volume. (appeared at R0.12)
+
+
+
+R0.13b (April 07, 2018)
+
+  Added support for UTF-32 encoding on the API. (FF_LFN_UNICODE = 3)
+  Added support for Unix style volume ID. (FF_STR_VOLUME_ID = 2)
+  Fixed accesing any object on the exFAT root directory beyond the cluster boundary can fail. (appeared at R0.12c)
+  Fixed f_setlabel() does not reject some invalid characters. (appeared at R0.09b)
+
+
+
+R0.13c (October 14, 2018)
+  Supported stdint.h for C99 and later. (integer.h was included in ff.h)
+  Fixed reading a directory gets infinite loop when the last directory entry is not empty. (appeared at R0.12)
+  Fixed creating a sub-directory in the fragmented sub-directory on the exFAT volume collapses FAT chain of the parent directory. (appeared at R0.12)
+  Fixed f_getcwd() cause output buffer overrun when the buffer has a valid drive number. (appeared at R0.13b)

code/components/esp-fatfs/src/00readme.txt

@@ -0,0 +1,20 @@
+FatFs Module Source Files R0.13c
+
+
+FILES
+
+  00readme.txt   This file.
+  00history.txt  Revision history.
+  ff.c           FatFs module.
+  ffconf.h       Configuration file of FatFs module.
+  ff.h           Common include file for FatFs and application module.
+  diskio.h       Common include file for FatFs and disk I/O module.
+  diskio.c       An example of glue function to attach existing disk I/O module to FatFs.
+  ffunicode.c    Optional Unicode utility functions.
+  ffsystem.c     An example of optional O/S related functions.
+
+
+  Low level disk I/O module is not included in this archive because the FatFs
+  module is only a generic file system layer and it does not depend on any specific
+  storage device. You need to provide a low level disk I/O module written to
+  control the storage device that attached to the target system.

code/components/esp-fatfs/src/diskio_mh.c

@@ -0,0 +1,228 @@
+/*-----------------------------------------------------------------------*/
+/* Low level disk I/O module SKELETON for FatFs     (C)ChaN, 2019        */
+/*-----------------------------------------------------------------------*/
+/* If a working storage control module is available, it should be        */
+/* attached to the FatFs via a glue function rather than modifying it.   */
+/* This is an example of glue functions to attach various exsisting      */
+/* storage control modules to the FatFs module with a defined API.       */
+/*-----------------------------------------------------------------------*/
+
+#include "ff_mh.h"			/* Obtains integer types */
+#include "diskio_mh.h"		/* Declarations of disk functions */
+
+/* Definitions of physical drive number for each drive */
+#define DEV_RAM		0	/* Example: Map Ramdisk to physical drive 0 */
+#define DEV_MMC		1	/* Example: Map MMC/SD card to physical drive 1 */
+#define DEV_USB		2	/* Example: Map USB MSD to physical drive 2 */
+
+
+/*-----------------------------------------------------------------------*/
+/* Get Drive Status                                                      */
+/*-----------------------------------------------------------------------*/
+
+DSTATUS disk_status (
+	BYTE pdrv		/* Physical drive nmuber to identify the drive */
+)
+{
+	DSTATUS stat;
+	int result;
+
+	switch (pdrv) {
+	case DEV_RAM :
+		result = RAM_disk_status();
+
+		// translate the reslut code here
+
+		return stat;
+
+	case DEV_MMC :
+		result = MMC_disk_status();
+
+		// translate the reslut code here
+
+		return stat;
+
+	case DEV_USB :
+		result = USB_disk_status();
+
+		// translate the reslut code here
+
+		return stat;
+	}
+	return STA_NOINIT;
+}
+
+
+
+/*-----------------------------------------------------------------------*/
+/* Inidialize a Drive                                                    */
+/*-----------------------------------------------------------------------*/
+
+DSTATUS disk_initialize (
+	BYTE pdrv				/* Physical drive nmuber to identify the drive */
+)
+{
+	DSTATUS stat;
+	int result;
+
+	switch (pdrv) {
+	case DEV_RAM :
+		result = RAM_disk_initialize();
+
+		// translate the reslut code here
+
+		return stat;
+
+	case DEV_MMC :
+		result = MMC_disk_initialize();
+
+		// translate the reslut code here
+
+		return stat;
+
+	case DEV_USB :
+		result = USB_disk_initialize();
+
+		// translate the reslut code here
+
+		return stat;
+	}
+	return STA_NOINIT;
+}
+
+
+
+/*-----------------------------------------------------------------------*/
+/* Read Sector(s)                                                        */
+/*-----------------------------------------------------------------------*/
+
+DRESULT disk_read (
+	BYTE pdrv,		/* Physical drive nmuber to identify the drive */
+	BYTE *buff,		/* Data buffer to store read data */
+	LBA_t sector,	/* Start sector in LBA */
+	UINT count		/* Number of sectors to read */
+)
+{
+	DRESULT res;
+	int result;
+
+	switch (pdrv) {
+	case DEV_RAM :
+		// translate the arguments here
+
+		result = RAM_disk_read(buff, sector, count);
+
+		// translate the reslut code here
+
+		return res;
+
+	case DEV_MMC :
+		// translate the arguments here
+
+		result = MMC_disk_read(buff, sector, count);
+
+		// translate the reslut code here
+
+		return res;
+
+	case DEV_USB :
+		// translate the arguments here
+
+		result = USB_disk_read(buff, sector, count);
+
+		// translate the reslut code here
+
+		return res;
+	}
+
+	return RES_PARERR;
+}
+
+
+
+/*-----------------------------------------------------------------------*/
+/* Write Sector(s)                                                       */
+/*-----------------------------------------------------------------------*/
+
+#if FF_FS_READONLY == 0
+
+DRESULT disk_write (
+	BYTE pdrv,			/* Physical drive nmuber to identify the drive */
+	const BYTE *buff,	/* Data to be written */
+	LBA_t sector,		/* Start sector in LBA */
+	UINT count			/* Number of sectors to write */
+)
+{
+	DRESULT res;
+	int result;
+
+	switch (pdrv) {
+	case DEV_RAM :
+		// translate the arguments here
+
+		result = RAM_disk_write(buff, sector, count);
+
+		// translate the reslut code here
+
+		return res;
+
+	case DEV_MMC :
+		// translate the arguments here
+
+		result = MMC_disk_write(buff, sector, count);
+
+		// translate the reslut code here
+
+		return res;
+
+	case DEV_USB :
+		// translate the arguments here
+
+		result = USB_disk_write(buff, sector, count);
+
+		// translate the reslut code here
+
+		return res;
+	}
+
+	return RES_PARERR;
+}
+
+#endif
+
+
+/*-----------------------------------------------------------------------*/
+/* Miscellaneous Functions                                               */
+/*-----------------------------------------------------------------------*/
+
+DRESULT disk_ioctl (
+	BYTE pdrv,		/* Physical drive nmuber (0..) */
+	BYTE cmd,		/* Control code */
+	void *buff		/* Buffer to send/receive control data */
+)
+{
+	DRESULT res;
+	int result;
+
+	switch (pdrv) {
+	case DEV_RAM :
+
+		// Process of the command for the RAM drive
+
+		return res;
+
+	case DEV_MMC :
+
+		// Process of the command for the MMC/SD card
+
+		return res;
+
+	case DEV_USB :
+
+		// Process of the command the USB drive
+
+		return res;
+	}
+
+	return RES_PARERR;
+}

code/components/esp-fatfs/src/diskio_mh.h

@@ -0,0 +1,79 @@
+/*-----------------------------------------------------------------------/
+/  Low level disk interface modlue include file   (C)ChaN, 2019          /
+/-----------------------------------------------------------------------*/
+
+#ifndef _DISKIO_DEFINED
+#define _DISKIO_DEFINED
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "ff_mh.h"
+
+/* Status of Disk Functions */
+typedef BYTE	DSTATUS;
+
+/* Results of Disk Functions */
+typedef enum {
+	RES_OK = 0,		/* 0: Successful */
+	RES_ERROR,		/* 1: R/W Error */
+	RES_WRPRT,		/* 2: Write Protected */
+	RES_NOTRDY,		/* 3: Not Ready */
+	RES_PARERR		/* 4: Invalid Parameter */
+} DRESULT;
+
+
+/*---------------------------------------*/
+/* Prototypes for disk control functions */
+
+
+DSTATUS disk_initialize (BYTE pdrv);
+DSTATUS disk_status (BYTE pdrv);
+DRESULT disk_read (BYTE pdrv, BYTE* buff, LBA_t sector, UINT count);
+DRESULT disk_write (BYTE pdrv, const BYTE* buff, LBA_t sector, UINT count);
+DRESULT disk_ioctl (BYTE pdrv, BYTE cmd, void* buff);
+
+
+/* Disk Status Bits (DSTATUS) */
+
+#define STA_NOINIT		0x01	/* Drive not initialized */
+#define STA_NODISK		0x02	/* No medium in the drive */
+#define STA_PROTECT		0x04	/* Write protected */
+
+
+/* Command code for disk_ioctrl fucntion */
+
+/* Generic command (Used by FatFs) */
+#define CTRL_SYNC			0	/* Complete pending write process (needed at FF_FS_READONLY == 0) */
+#define GET_SECTOR_COUNT	1	/* Get media size (needed at FF_USE_MKFS == 1) */
+#define GET_SECTOR_SIZE		2	/* Get sector size (needed at FF_MAX_SS != FF_MIN_SS) */
+#define GET_BLOCK_SIZE		3	/* Get erase block size (needed at FF_USE_MKFS == 1) */
+#define CTRL_TRIM			4	/* Inform device that the data on the block of sectors is no longer used (needed at FF_USE_TRIM == 1) */
+
+/* Generic command (Not used by FatFs) */
+#define CTRL_POWER			5	/* Get/Set power status */
+#define CTRL_LOCK			6	/* Lock/Unlock media removal */
+#define CTRL_EJECT			7	/* Eject media */
+#define CTRL_FORMAT			8	/* Create physical format on the media */
+
+/* MMC/SDC specific ioctl command */
+#define MMC_GET_TYPE		10	/* Get card type */
+#define MMC_GET_CSD			11	/* Get CSD */
+#define MMC_GET_CID			12	/* Get CID */
+#define MMC_GET_OCR			13	/* Get OCR */
+#define MMC_GET_SDSTAT		14	/* Get SD status */
+#define ISDIO_READ			55	/* Read data form SD iSDIO register */
+#define ISDIO_WRITE			56	/* Write data to SD iSDIO register */
+#define ISDIO_MRITE			57	/* Masked write data to SD iSDIO register */
+
+/* ATA/CF specific ioctl command */
+#define ATA_GET_REV			20	/* Get F/W revision */
+#define ATA_GET_MODEL		21	/* Get model name */
+#define ATA_GET_SN			22	/* Get serial number */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

code/components/esp-fatfs/src/ff_mh.h

@@ -0,0 +1,422 @@
+/*----------------------------------------------------------------------------/
+/  FatFs - Generic FAT Filesystem module  R0.14b                              /
+/-----------------------------------------------------------------------------/
+/
+/ Copyright (C) 2021, ChaN, all right reserved.
+/
+/ FatFs module is an open source software. Redistribution and use of FatFs in
+/ source and binary forms, with or without modification, are permitted provided
+/ that the following condition is met:
+
+/ 1. Redistributions of source code must retain the above copyright notice,
+/    this condition and the following disclaimer.
+/
+/ This software is provided by the copyright holder and contributors "AS IS"
+/ and any warranties related to this software are DISCLAIMED.
+/ The copyright owner or contributors be NOT LIABLE for any damages caused
+/ by use of this software.
+/
+/----------------------------------------------------------------------------*/
+
+
+#ifndef FF_DEFINED
+#define FF_DEFINED	86631	/* Revision ID */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "ffconf_mh.h"		/* FatFs configuration options */
+
+#if FF_DEFINED != FFCONF_DEF
+#error Wrong configuration file (ffconf.h).
+#endif
+
+
+/* Integer types used for FatFs API */
+
+#if defined(_WIN32)		/* Windows VC++ (for development only) */
+#define FF_INTDEF 2
+#include <windows.h>
+typedef unsigned __int64 QWORD;
+#include <float.h>
+#define isnan(v) _isnan(v)
+#define isinf(v) (!_finite(v))
+
+#elif (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || defined(__cplusplus)	/* C99 or later */
+#define FF_INTDEF 2
+#include <stdint.h>
+typedef unsigned int	UINT;	/* int must be 16-bit or 32-bit */
+typedef unsigned char	BYTE;	/* char must be 8-bit */
+typedef uint16_t		WORD;	/* 16-bit unsigned integer */
+typedef uint32_t		DWORD;	/* 32-bit unsigned integer */
+typedef uint64_t		QWORD;	/* 64-bit unsigned integer */
+typedef WORD			WCHAR;	/* UTF-16 character type */
+
+#else  	/* Earlier than C99 */
+#define FF_INTDEF 1
+typedef unsigned int	UINT;	/* int must be 16-bit or 32-bit */
+typedef unsigned char	BYTE;	/* char must be 8-bit */
+typedef unsigned short	WORD;	/* 16-bit unsigned integer */
+typedef unsigned long	DWORD;	/* 32-bit unsigned integer */
+typedef WORD			WCHAR;	/* UTF-16 character type */
+#endif
+
+
+/* Type of file size and LBA variables */
+
+#if FF_FS_EXFAT
+#if FF_INTDEF != 2
+#error exFAT feature wants C99 or later
+#endif
+typedef QWORD FSIZE_t;
+#if FF_LBA64
+typedef QWORD LBA_t;
+#else
+typedef DWORD LBA_t;
+#endif
+#else
+#if FF_LBA64
+#error exFAT needs to be enabled when enable 64-bit LBA
+#endif
+typedef DWORD FSIZE_t;
+typedef DWORD LBA_t;
+#endif
+
+
+
+/* Type of path name strings on FatFs API (TCHAR) */
+
+#if FF_USE_LFN && FF_LFN_UNICODE == 1 	/* Unicode in UTF-16 encoding */
+typedef WCHAR TCHAR;
+#define _T(x) L ## x
+#define _TEXT(x) L ## x
+#elif FF_USE_LFN && FF_LFN_UNICODE == 2	/* Unicode in UTF-8 encoding */
+typedef char TCHAR;
+#define _T(x) u8 ## x
+#define _TEXT(x) u8 ## x
+#elif FF_USE_LFN && FF_LFN_UNICODE == 3	/* Unicode in UTF-32 encoding */
+typedef DWORD TCHAR;
+#define _T(x) U ## x
+#define _TEXT(x) U ## x
+#elif FF_USE_LFN && (FF_LFN_UNICODE < 0 || FF_LFN_UNICODE > 3)
+#error Wrong FF_LFN_UNICODE setting
+#else									/* ANSI/OEM code in SBCS/DBCS */
+typedef char TCHAR;
+#define _T(x) x
+#define _TEXT(x) x
+#endif
+
+
+
+/* Definitions of volume management */
+
+#if FF_MULTI_PARTITION		/* Multiple partition configuration */
+typedef struct {
+	BYTE pd;	/* Physical drive number */
+	BYTE pt;	/* Partition: 0:Auto detect, 1-4:Forced partition) */
+} PARTITION;
+extern const PARTITION VolToPart[];	/* Volume - Partition mapping table */
+#endif
+
+#if FF_STR_VOLUME_ID
+#ifndef FF_VOLUME_STRS
+extern const char* VolumeStr[FF_VOLUMES];	/* User defied volume ID */
+#endif
+#endif
+
+
+
+/* Filesystem object structure (FATFS) */
+
+typedef struct {
+	BYTE	fs_type;		/* Filesystem type (0:not mounted) */
+	BYTE	pdrv;			/* Associated physical drive */
+	BYTE	n_fats;			/* Number of FATs (1 or 2) */
+	BYTE	wflag;			/* win[] flag (b0:dirty) */
+	BYTE	fsi_flag;		/* FSINFO flags (b7:disabled, b0:dirty) */
+	WORD	id;				/* Volume mount ID */
+	WORD	n_rootdir;		/* Number of root directory entries (FAT12/16) */
+	WORD	csize;			/* Cluster size [sectors] */
+#if FF_MAX_SS != FF_MIN_SS
+	WORD	ssize;			/* Sector size (512, 1024, 2048 or 4096) */
+#endif
+#if FF_USE_LFN
+	WCHAR*	lfnbuf;			/* LFN working buffer */
+#endif
+#if FF_FS_EXFAT
+	BYTE*	dirbuf;			/* Directory entry block scratchpad buffer for exFAT */
+#endif
+#if FF_FS_REENTRANT
+	FF_SYNC_t	sobj;		/* Identifier of sync object */
+#endif
+#if !FF_FS_READONLY
+	DWORD	last_clst;		/* Last allocated cluster */
+	DWORD	free_clst;		/* Number of free clusters */
+#endif
+#if FF_FS_RPATH
+	DWORD	cdir;			/* Current directory start cluster (0:root) */
+#if FF_FS_EXFAT
+	DWORD	cdc_scl;		/* Containing directory start cluster (invalid when cdir is 0) */
+	DWORD	cdc_size;		/* b31-b8:Size of containing directory, b7-b0: Chain status */
+	DWORD	cdc_ofs;		/* Offset in the containing directory (invalid when cdir is 0) */
+#endif
+#endif
+	DWORD	n_fatent;		/* Number of FAT entries (number of clusters + 2) */
+	DWORD	fsize;			/* Size of an FAT [sectors] */
+	LBA_t	volbase;		/* Volume base sector */
+	LBA_t	fatbase;		/* FAT base sector */
+	LBA_t	dirbase;		/* Root directory base sector/cluster */
+	LBA_t	database;		/* Data base sector */
+#if FF_FS_EXFAT
+	LBA_t	bitbase;		/* Allocation bitmap base sector */
+#endif
+	LBA_t	winsect;		/* Current sector appearing in the win[] */
+	BYTE	win[FF_MAX_SS];	/* Disk access window for Directory, FAT (and file data at tiny cfg) */
+} FATFS;
+
+
+
+/* Object ID and allocation information (FFOBJID) */
+
+typedef struct {
+	FATFS*	fs;				/* Pointer to the hosting volume of this object */
+	WORD	id;				/* Hosting volume mount ID */
+	BYTE	attr;			/* Object attribute */
+	BYTE	stat;			/* Object chain status (b1-0: =0:not contiguous, =2:contiguous, =3:fragmented in this session, b2:sub-directory stretched) */
+	DWORD	sclust;			/* Object data start cluster (0:no cluster or root directory) */
+	FSIZE_t	objsize;		/* Object size (valid when sclust != 0) */
+#if FF_FS_EXFAT
+	DWORD	n_cont;			/* Size of first fragment - 1 (valid when stat == 3) */
+	DWORD	n_frag;			/* Size of last fragment needs to be written to FAT (valid when not zero) */
+	DWORD	c_scl;			/* Containing directory start cluster (valid when sclust != 0) */
+	DWORD	c_size;			/* b31-b8:Size of containing directory, b7-b0: Chain status (valid when c_scl != 0) */
+	DWORD	c_ofs;			/* Offset in the containing directory (valid when file object and sclust != 0) */
+#endif
+#if FF_FS_LOCK
+	UINT	lockid;			/* File lock ID origin from 1 (index of file semaphore table Files[]) */
+#endif
+} FFOBJID;
+
+
+
+/* File object structure (FIL) */
+
+typedef struct {
+	FFOBJID	obj;			/* Object identifier (must be the 1st member to detect invalid object pointer) */
+	BYTE	flag;			/* File status flags */
+	BYTE	err;			/* Abort flag (error code) */
+	FSIZE_t	fptr;			/* File read/write pointer (Zeroed on file open) */
+	DWORD	clust;			/* Current cluster of fpter (invalid when fptr is 0) */
+	LBA_t	sect;			/* Sector number appearing in buf[] (0:invalid) */
+#if !FF_FS_READONLY
+	LBA_t	dir_sect;		/* Sector number containing the directory entry (not used at exFAT) */
+	BYTE*	dir_ptr;		/* Pointer to the directory entry in the win[] (not used at exFAT) */
+#endif
+#if FF_USE_FASTSEEK
+	DWORD*	cltbl;			/* Pointer to the cluster link map table (nulled on open, set by application) */
+#endif
+#if !FF_FS_TINY
+	BYTE	buf[FF_MAX_SS];	/* File private data read/write window */
+#endif
+} FIL;
+
+
+
+/* Directory object structure (FF_DIR) */
+
+typedef struct {
+	FFOBJID	obj;			/* Object identifier */
+	DWORD	dptr;			/* Current read/write offset */
+	DWORD	clust;			/* Current cluster */
+	LBA_t	sect;			/* Current sector (0:Read operation has terminated) */
+	BYTE*	dir;			/* Pointer to the directory item in the win[] */
+	BYTE	fn[12];			/* SFN (in/out) {body[8],ext[3],status[1]} */
+#if FF_USE_LFN
+	DWORD	blk_ofs;		/* Offset of current entry block being processed (0xFFFFFFFF:Invalid) */
+#endif
+#if FF_USE_FIND
+	const TCHAR* pat;		/* Pointer to the name matching pattern */
+#endif
+} FF_DIR;
+
+
+
+/* File information structure (FILINFO) */
+
+typedef struct {
+	FSIZE_t	fsize;			/* File size */
+	WORD	fdate;			/* Modified date */
+	WORD	ftime;			/* Modified time */
+	BYTE	fattrib;		/* File attribute */
+#if FF_USE_LFN
+	TCHAR	altname[FF_SFN_BUF + 1];/* Altenative file name */
+	TCHAR	fname[FF_LFN_BUF + 1];	/* Primary file name */
+#else
+	TCHAR	fname[12 + 1];	/* File name */
+#endif
+} FILINFO;
+
+
+
+/* Format parameter structure (MKFS_PARM) */
+
+typedef struct {
+	BYTE fmt;			/* Format option (FM_FAT, FM_FAT32, FM_EXFAT and FM_SFD) */
+	BYTE n_fat;			/* Number of FATs */
+	UINT align;			/* Data area alignment (sector) */
+	UINT n_root;		/* Number of root directory entries */
+	DWORD au_size;		/* Cluster size (byte) */
+} MKFS_PARM;
+
+
+
+/* File function return code (FRESULT) */
+
+typedef enum {
+	FR_OK = 0,				/* (0) Succeeded */
+	FR_DISK_ERR,			/* (1) A hard error occurred in the low level disk I/O layer */
+	FR_INT_ERR,				/* (2) Assertion failed */
+	FR_NOT_READY,			/* (3) The physical drive cannot work */
+	FR_NO_FILE,				/* (4) Could not find the file */
+	FR_NO_PATH,				/* (5) Could not find the path */
+	FR_INVALID_NAME,		/* (6) The path name format is invalid */
+	FR_DENIED,				/* (7) Access denied due to prohibited access or directory full */
+	FR_EXIST,				/* (8) Access denied due to prohibited access */
+	FR_INVALID_OBJECT,		/* (9) The file/directory object is invalid */
+	FR_WRITE_PROTECTED,		/* (10) The physical drive is write protected */
+	FR_INVALID_DRIVE,		/* (11) The logical drive number is invalid */
+	FR_NOT_ENABLED,			/* (12) The volume has no work area */
+	FR_NO_FILESYSTEM,		/* (13) There is no valid FAT volume */
+	FR_MKFS_ABORTED,		/* (14) The f_mkfs() aborted due to any problem */
+	FR_TIMEOUT,				/* (15) Could not get a grant to access the volume within defined period */
+	FR_LOCKED,				/* (16) The operation is rejected according to the file sharing policy */
+	FR_NOT_ENOUGH_CORE,		/* (17) LFN working buffer could not be allocated */
+	FR_TOO_MANY_OPEN_FILES,	/* (18) Number of open files > FF_FS_LOCK */
+	FR_INVALID_PARAMETER	/* (19) Given parameter is invalid */
+} FRESULT;
+
+
+
+/*--------------------------------------------------------------*/
+/* FatFs module application interface                           */
+
+FRESULT f_open (FIL* fp, const TCHAR* path, BYTE mode);				/* Open or create a file */
+FRESULT f_close (FIL* fp);											/* Close an open file object */
+FRESULT f_read (FIL* fp, void* buff, UINT btr, UINT* br);			/* Read data from the file */
+FRESULT f_write (FIL* fp, const void* buff, UINT btw, UINT* bw);	/* Write data to the file */
+FRESULT f_lseek (FIL* fp, FSIZE_t ofs);								/* Move file pointer of the file object */
+FRESULT f_truncate (FIL* fp);										/* Truncate the file */
+FRESULT f_sync (FIL* fp);											/* Flush cached data of the writing file */
+FRESULT f_opendir (FF_DIR* dp, const TCHAR* path);						/* Open a directory */
+FRESULT f_closedir (FF_DIR* dp);										/* Close an open directory */
+FRESULT f_readdir (FF_DIR* dp, FILINFO* fno);							/* Read a directory item */
+FRESULT f_findfirst (FF_DIR* dp, FILINFO* fno, const TCHAR* path, const TCHAR* pattern);	/* Find first file */
+FRESULT f_findnext (FF_DIR* dp, FILINFO* fno);							/* Find next file */
+FRESULT f_mkdir (const TCHAR* path);								/* Create a sub directory */
+FRESULT f_unlink (const TCHAR* path);								/* Delete an existing file or directory */
+FRESULT f_rename (const TCHAR* path_old, const TCHAR* path_new);	/* Rename/Move a file or directory */
+FRESULT f_stat (const TCHAR* path, FILINFO* fno);					/* Get file status */
+FRESULT f_chmod (const TCHAR* path, BYTE attr, BYTE mask);			/* Change attribute of a file/dir */
+FRESULT f_utime (const TCHAR* path, const FILINFO* fno);			/* Change timestamp of a file/dir */
+FRESULT f_chdir (const TCHAR* path);								/* Change current directory */
+FRESULT f_chdrive (const TCHAR* path);								/* Change current drive */
+FRESULT f_getcwd (TCHAR* buff, UINT len);							/* Get current directory */
+FRESULT f_getfree (const TCHAR* path, DWORD* nclst, FATFS** fatfs);	/* Get number of free clusters on the drive */
+FRESULT f_getlabel (const TCHAR* path, TCHAR* label, DWORD* vsn);	/* Get volume label */
+FRESULT f_setlabel (const TCHAR* label);							/* Set volume label */
+FRESULT f_forward (FIL* fp, UINT(*func)(const BYTE*,UINT), UINT btf, UINT* bf);	/* Forward data to the stream */
+FRESULT f_expand (FIL* fp, FSIZE_t fsz, BYTE opt);					/* Allocate a contiguous block to the file */
+FRESULT f_mount (FATFS* fs, const TCHAR* path, BYTE opt);			/* Mount/Unmount a logical drive */
+FRESULT f_mkfs (const TCHAR* path, const MKFS_PARM* opt, void* work, UINT len);	/* Create a FAT volume */
+FRESULT f_fdisk (BYTE pdrv, const LBA_t ptbl[], void* work);		/* Divide a physical drive into some partitions */
+FRESULT f_setcp (WORD cp);											/* Set current code page */
+int f_putc (TCHAR c, FIL* fp);										/* Put a character to the file */
+int f_puts (const TCHAR* str, FIL* cp);								/* Put a string to the file */
+int f_printf (FIL* fp, const TCHAR* str, ...);						/* Put a formatted string to the file */
+TCHAR* f_gets (TCHAR* buff, int len, FIL* fp);						/* Get a string from the file */
+
+#define f_eof(fp) ((int)((fp)->fptr == (fp)->obj.objsize))
+#define f_error(fp) ((fp)->err)
+#define f_tell(fp) ((fp)->fptr)
+#define f_size(fp) ((fp)->obj.objsize)
+#define f_rewind(fp) f_lseek((fp), 0)
+#define f_rewinddir(dp) f_readdir((dp), 0)
+#define f_rmdir(path) f_unlink(path)
+#define f_unmount(path) f_mount(0, path, 0)
+
+
+
+
+/*--------------------------------------------------------------*/
+/* Additional user defined functions                            */
+
+/* RTC function */
+#if !FF_FS_READONLY && !FF_FS_NORTC
+DWORD get_fattime (void);
+#endif
+
+/* LFN support functions */
+#if FF_USE_LFN >= 1						/* Code conversion (defined in unicode.c) */
+WCHAR ff_oem2uni (WCHAR oem, WORD cp);	/* OEM code to Unicode conversion */
+WCHAR ff_uni2oem (DWORD uni, WORD cp);	/* Unicode to OEM code conversion */
+DWORD ff_wtoupper (DWORD uni);			/* Unicode upper-case conversion */
+#endif
+#if FF_USE_LFN == 3						/* Dynamic memory allocation */
+void* ff_memalloc (UINT msize);			/* Allocate memory block */
+void ff_memfree (void* mblock);			/* Free memory block */
+#endif
+
+/* Sync functions */
+#if FF_FS_REENTRANT
+int ff_cre_syncobj (BYTE vol, FF_SYNC_t* sobj);	/* Create a sync object */
+int ff_req_grant (FF_SYNC_t sobj);		/* Lock sync object */
+void ff_rel_grant (FF_SYNC_t sobj);		/* Unlock sync object */
+int ff_del_syncobj (FF_SYNC_t sobj);	/* Delete a sync object */
+#endif
+
+
+
+
+/*--------------------------------------------------------------*/
+/* Flags and offset address                                     */
+
+
+/* File access mode and open method flags (3rd argument of f_open) */
+#define	FA_READ				0x01
+#define	FA_WRITE			0x02
+#define	FA_OPEN_EXISTING	0x00
+#define	FA_CREATE_NEW		0x04
+#define	FA_CREATE_ALWAYS	0x08
+#define	FA_OPEN_ALWAYS		0x10
+#define	FA_OPEN_APPEND		0x30
+
+/* Fast seek controls (2nd argument of f_lseek) */
+#define CREATE_LINKMAP	((FSIZE_t)0 - 1)
+
+/* Format options (2nd argument of f_mkfs) */
+#define FM_FAT		0x01
+#define FM_FAT32	0x02
+#define FM_EXFAT	0x04
+#define FM_ANY		0x07
+#define FM_SFD		0x08
+
+/* Filesystem type (FATFS.fs_type) */
+#define FS_FAT12	1
+#define FS_FAT16	2
+#define FS_FAT32	3
+#define FS_EXFAT	4
+
+/* File attribute bits for directory entry (FILINFO.fattrib) */
+#define	AM_RDO	0x01	/* Read only */
+#define	AM_HID	0x02	/* Hidden */
+#define	AM_SYS	0x04	/* System */
+#define AM_DIR	0x10	/* Directory */
+#define AM_ARC	0x20	/* Archive */
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* FF_DEFINED */

code/components/esp-fatfs/src/ffconf_mh.h

@@ -0,0 +1,309 @@
+#include "sdkconfig.h"
+
+/*---------------------------------------------------------------------------/
+/  FatFs Functional Configurations
+/---------------------------------------------------------------------------*/
+
+#define FFCONF_DEF 86631 /* Revision ID */
+
+/*---------------------------------------------------------------------------/
+/ Function Configurations
+/---------------------------------------------------------------------------*/
+
+#define FF_FS_READONLY 0
+/* This option switches read-only configuration. (0:Read/Write or 1:Read-only)
+/  Read-only configuration removes writing API functions, f_write(), f_sync(),
+/  f_unlink(), f_mkdir(), f_chmod(), f_rename(), f_truncate(), f_getfree()
+/  and optional writing functions as well. */
+
+#define FF_FS_MINIMIZE 0
+/* This option defines minimization level to remove some basic API functions.
+/
+/   0: Basic functions are fully enabled.
+/   1: f_stat(), f_getfree(), f_unlink(), f_mkdir(), f_truncate() and f_rename()
+/      are removed.
+/   2: f_opendir(), f_readdir() and f_closedir() are removed in addition to 1.
+/   3: f_lseek() function is removed in addition to 2. */
+
+#define FF_USE_FIND 0
+/* This option switches filtered directory read functions, f_findfirst() and
+/  f_findnext(). (0:Disable, 1:Enable 2:Enable with matching altname[] too) */
+
+#define FF_USE_MKFS 1
+/* This option switches f_mkfs() function. (0:Disable or 1:Enable) */
+
+#define FF_USE_FASTSEEK CONFIG_FATFS_USE_FASTSEEK
+/* This option switches fast seek function. (0:Disable or 1:Enable) */
+
+#define FF_USE_EXPAND 0
+/* This option switches f_expand function. (0:Disable or 1:Enable) */
+
+#define FF_USE_CHMOD 1
+/* This option switches attribute manipulation functions, f_chmod() and f_utime().
+/  (0:Disable or 1:Enable) Also FF_FS_READONLY needs to be 0 to enable this option. */
+
+#define FF_USE_LABEL 0
+/* This option switches volume label functions, f_getlabel() and f_setlabel().
+/  (0:Disable or 1:Enable) */
+
+#define FF_USE_FORWARD 0
+/* This option switches f_forward() function. (0:Disable or 1:Enable) */
+
+#define FF_USE_STRFUNC 0
+#define FF_PRINT_LLI 0
+#define FF_PRINT_FLOAT 0
+#define FF_STRF_ENCODE 3
+/* FF_USE_STRFUNC switches string functions, f_gets(), f_putc(), f_puts() and
+/  f_printf().
+/
+/   0: Disable. FF_PRINT_LLI, FF_PRINT_FLOAT and FF_STRF_ENCODE have no effect.
+/   1: Enable without LF-CRLF conversion.
+/   2: Enable with LF-CRLF conversion.
+/
+/  FF_PRINT_LLI = 1 makes f_printf() support long long argument and FF_PRINT_FLOAT = 1/2
+   makes f_printf() support floating point argument. These features want C99 or later.
+/  When FF_LFN_UNICODE >= 1 with LFN enabled, string functions convert the character
+/  encoding in it. FF_STRF_ENCODE selects assumption of character encoding ON THE FILE
+/  to be read/written via those functions.
+/
+/   0: ANSI/OEM in current CP
+/   1: Unicode in UTF-16LE
+/   2: Unicode in UTF-16BE
+/   3: Unicode in UTF-8
+*/
+
+/*---------------------------------------------------------------------------/
+/ Locale and Namespace Configurations
+/---------------------------------------------------------------------------*/
+
+#define FF_CODE_PAGE CONFIG_FATFS_CODEPAGE
+/* This option specifies the OEM code page to be used on the target system.
+/  Incorrect code page setting can cause a file open failure.
+/
+/   437 - U.S.
+/   720 - Arabic
+/   737 - Greek
+/   771 - KBL
+/   775 - Baltic
+/   850 - Latin 1
+/   852 - Latin 2
+/   855 - Cyrillic
+/   857 - Turkish
+/   860 - Portuguese
+/   861 - Icelandic
+/   862 - Hebrew
+/   863 - Canadian French
+/   864 - Arabic
+/   865 - Nordic
+/   866 - Russian
+/   869 - Greek 2
+/   932 - Japanese (DBCS)
+/   936 - Simplified Chinese (DBCS)
+/   949 - Korean (DBCS)
+/   950 - Traditional Chinese (DBCS)
+/     0 - Include all code pages above and configured by f_setcp()
+*/
+
+#if defined(CONFIG_FATFS_LFN_STACK)
+#define FF_USE_LFN 2
+#elif defined(CONFIG_FATFS_LFN_HEAP)
+#define FF_USE_LFN 3
+#else /* CONFIG_FATFS_LFN_NONE */
+#define FF_USE_LFN 0
+#endif
+
+#ifdef CONFIG_FATFS_MAX_LFN
+#define FF_MAX_LFN CONFIG_FATFS_MAX_LFN
+#endif
+
+/* The FF_USE_LFN switches the support for LFN (long file name).
+/
+/   0: Disable LFN. FF_MAX_LFN has no effect.
+/   1: Enable LFN with static  working buffer on the BSS. Always NOT thread-safe.
+/   2: Enable LFN with dynamic working buffer on the STACK.
+/   3: Enable LFN with dynamic working buffer on the HEAP.
+/
+/  To enable the LFN, ffunicode.c needs to be added to the project. The LFN function
+/  requiers certain internal working buffer occupies (FF_MAX_LFN + 1) * 2 bytes and
+/  additional (FF_MAX_LFN + 44) / 15 * 32 bytes when exFAT is enabled.
+/  The FF_MAX_LFN defines size of the working buffer in UTF-16 code unit and it can
+/  be in range of 12 to 255. It is recommended to be set it 255 to fully support LFN
+/  specification.
+/  When use stack for the working buffer, take care on stack overflow. When use heap
+/  memory for the working buffer, memory management functions, ff_memalloc() and
+/  ff_memfree() exemplified in ffsystem.c, need to be added to the project. */
+
+#ifdef CONFIG_FATFS_API_ENCODING_UTF_8
+#define FF_LFN_UNICODE 2
+#else /* CONFIG_FATFS_API_ENCODING_ANSI_OEM */
+#define FF_LFN_UNICODE 0
+#endif
+/* This option switches the character encoding on the API when LFN is enabled.
+/
+/   0: ANSI/OEM in current CP (TCHAR = char)
+/   1: Unicode in UTF-16 (TCHAR = WCHAR)
+/   2: Unicode in UTF-8 (TCHAR = char)
+/   3: Unicode in UTF-32 (TCHAR = DWORD)
+/
+/  Also behavior of string I/O functions will be affected by this option.
+/  When LFN is not enabled, this option has no effect. */
+
+#define FF_LFN_BUF 255
+#define FF_SFN_BUF 12
+/* This set of options defines size of file name members in the FILINFO structure
+/  which is used to read out directory items. These values should be suffcient for
+/  the file names to read. The maximum possible length of the read file name depends
+/  on character encoding. When LFN is not enabled, these options have no effect. */
+
+#define FF_FS_RPATH 0
+/* This option configures support for relative path.
+/
+/   0: Disable relative path and remove related functions.
+/   1: Enable relative path. f_chdir() and f_chdrive() are available.
+/   2: f_getcwd() function is available in addition to 1.
+*/
+
+/*---------------------------------------------------------------------------/
+/ Drive/Volume Configurations
+/---------------------------------------------------------------------------*/
+
+#define FF_VOLUMES CONFIG_FATFS_VOLUME_COUNT
+/* Number of volumes (logical drives) to be used. (1-10) */
+
+#define FF_STR_VOLUME_ID 0
+#define FF_VOLUME_STRS "RAM", "NAND", "CF", "SD", "SD2", "USB", "USB2", "USB3"
+/* FF_STR_VOLUME_ID switches support for volume ID in arbitrary strings.
+/  When FF_STR_VOLUME_ID is set to 1 or 2, arbitrary strings can be used as drive
+/  number in the path name. FF_VOLUME_STRS defines the volume ID strings for each
+/  logical drives. Number of items must not be less than FF_VOLUMES. Valid
+/  characters for the volume ID strings are A-Z, a-z and 0-9, however, they are
+/  compared in case-insensitive. If FF_STR_VOLUME_ID >= 1 and FF_VOLUME_STRS is
+/  not defined, a user defined volume string table needs to be defined as:
+/
+/  const char* VolumeStr[FF_VOLUMES] = {"ram","flash","sd","usb",...
+*/
+
+#define FF_MULTI_PARTITION 1
+/* This option switches support for multiple volumes on the physical drive.
+/  By default (0), each logical drive number is bound to the same physical drive
+/  number and only an FAT volume found on the physical drive will be mounted.
+/  When this function is enabled (1), each logical drive number can be bound to
+/  arbitrary physical drive and partition listed in the VolToPart[]. Also f_fdisk()
+/  funciton will be available. */
+
+/* SD card sector size */
+#define FF_SS_SDCARD 512
+/* wear_levelling library sector size */
+#define FF_SS_WL CONFIG_WL_SECTOR_SIZE
+
+#define FF_MIN_SS MIN(FF_SS_SDCARD, FF_SS_WL)
+#define FF_MAX_SS MAX(FF_SS_SDCARD, FF_SS_WL)
+/* This set of options configures the range of sector size to be supported. (512,
+/  1024, 2048 or 4096) Always set both 512 for most systems, generic memory card and
+/  harddisk, but a larger value may be required for on-board flash memory and some
+/  type of optical media. When FF_MAX_SS is larger than FF_MIN_SS, FatFs is configured
+/  for variable sector size mode and disk_ioctl() function needs to implement
+/  GET_SECTOR_SIZE command. */
+
+#define FF_LBA64 0
+/* This option switches support for 64-bit LBA. (0:Disable or 1:Enable)
+/  To enable the 64-bit LBA, also exFAT needs to be enabled. (FF_FS_EXFAT == 1) */
+
+#define FF_MIN_GPT 0x10000000
+/* Minimum number of sectors to switch GPT as partitioning format in f_mkfs and
+/  f_fdisk function. 0x100000000 max. This option has no effect when FF_LBA64 == 0. */
+
+#define FF_USE_TRIM 1
+/* This option switches support for ATA-TRIM. (0:Disable or 1:Enable)
+/  To enable Trim function, also CTRL_TRIM command should be implemented to the
+/  disk_ioctl() function. */
+
+/*---------------------------------------------------------------------------/
+/ System Configurations
+/---------------------------------------------------------------------------*/
+
+#define FF_FS_TINY (!CONFIG_FATFS_PER_FILE_CACHE)
+/* This option switches tiny buffer configuration. (0:Normal or 1:Tiny)
+/  At the tiny configuration, size of file object (FIL) is shrinked FF_MAX_SS bytes.
+/  Instead of private sector buffer eliminated from the file object, common sector
+/  buffer in the filesystem object (FATFS) is used for the file data transfer. */
+
+#define FF_FS_EXFAT 0
+/* This option switches support for exFAT filesystem. (0:Disable or 1:Enable)
+/  To enable exFAT, also LFN needs to be enabled. (FF_USE_LFN >= 1)
+/  Note that enabling exFAT discards ANSI C (C89) compatibility. */
+
+#define FF_FS_NORTC 0
+#define FF_NORTC_MON 1
+#define FF_NORTC_MDAY 1
+#define FF_NORTC_YEAR 2020
+/* The option FF_FS_NORTC switches timestamp functiton. If the system does not have
+/  any RTC function or valid timestamp is not needed, set FF_FS_NORTC = 1 to disable
+/  the timestamp function. Every object modified by FatFs will have a fixed timestamp
+/  defined by FF_NORTC_MON, FF_NORTC_MDAY and FF_NORTC_YEAR in local time.
+/  To enable timestamp function (FF_FS_NORTC = 0), get_fattime() function need to be
+/  added to the project to read current time form real-time clock. FF_NORTC_MON,
+/  FF_NORTC_MDAY and FF_NORTC_YEAR have no effect.
+/  These options have no effect in read-only configuration (FF_FS_READONLY = 1). */
+
+#define FF_FS_NOFSINFO 0
+/* If you need to know correct free space on the FAT32 volume, set bit 0 of this
+/  option, and f_getfree() function at first time after volume mount will force
+/  a full FAT scan. Bit 1 controls the use of last allocated cluster number.
+/
+/  bit0=0: Use free cluster count in the FSINFO if available.
+/  bit0=1: Do not trust free cluster count in the FSINFO.
+/  bit1=0: Use last allocated cluster number in the FSINFO if available.
+/  bit1=1: Do not trust last allocated cluster number in the FSINFO.
+*/
+
+#define FF_FS_LOCK CONFIG_FATFS_FS_LOCK
+/* The option FF_FS_LOCK switches file lock function to control duplicated file open
+/  and illegal operation to open objects. This option must be 0 when FF_FS_READONLY
+/  is 1.
+/
+/  0:  Disable file lock function. To avoid volume corruption, application program
+/      should avoid illegal open, remove and rename to the open objects.
+/  >0: Enable file lock function. The value defines how many files/sub-directories
+/      can be opened simultaneously under file lock control. Note that the file
+/      lock control is independent of re-entrancy. */
+
+/* #include <somertos.h>	// O/S definitions */
+#define FF_FS_REENTRANT 1
+#define FF_FS_TIMEOUT (CONFIG_FATFS_TIMEOUT_MS / portTICK_PERIOD_MS)
+#define FF_SYNC_t SemaphoreHandle_t
+/* The option FF_FS_REENTRANT switches the re-entrancy (thread safe) of the FatFs
+/  module itself. Note that regardless of this option, file access to different
+/  volume is always re-entrant and volume control functions, f_mount(), f_mkfs()
+/  and f_fdisk() function, are always not re-entrant. Only file/directory access
+/  to the same volume is under control of this function.
+/
+/   0: Disable re-entrancy. FF_FS_TIMEOUT and FF_SYNC_t have no effect.
+/   1: Enable re-entrancy. Also user provided synchronization handlers,
+/      ff_req_grant(), ff_rel_grant(), ff_del_syncobj() and ff_cre_syncobj()
+/      function, must be added to the project. Samples are available in
+/      option/syscall.c.
+/
+/  The FF_FS_TIMEOUT defines timeout period in unit of time tick.
+/  The FF_SYNC_t defines O/S dependent sync object type. e.g. HANDLE, ID, OS_EVENT*,
+/  SemaphoreHandle_t and etc. A header file for O/S definitions needs to be
+/  included somewhere in the scope of ff.h. */
+
+#include <sys/param.h>
+#include "freertos/FreeRTOS.h"
+#include "freertos/semphr.h"
+
+/* Some memory allocation functions are declared here in addition to ff.h, so that
+   they can be used also by external code when LFN feature is disabled.
+ */
+void *ff_memalloc(unsigned msize);
+void ff_memfree(void *);
+
+/*--- End of configuration options ---*/
+
+/* Redefine names of disk IO functions to prevent name collisions */
+#define disk_initialize ff_disk_initialize
+#define disk_status ff_disk_status
+#define disk_read ff_disk_read
+#define disk_write ff_disk_write
+#define disk_ioctl ff_disk_ioctl

code/components/esp-fatfs/src/ffsystem_mh.c

@@ -0,0 +1,169 @@
+/*------------------------------------------------------------------------*/
+/* Sample Code of OS Dependent Functions for FatFs                        */
+/* (C)ChaN, 2018                                                          */
+/*------------------------------------------------------------------------*/
+
+
+#include "ff_mh.h"
+
+
+#if FF_USE_LFN == 3	/* Dynamic memory allocation */
+
+/*------------------------------------------------------------------------*/
+/* Allocate a memory block                                                */
+/*------------------------------------------------------------------------*/
+
+void* ff_memalloc (	/* Returns pointer to the allocated memory block (null if not enough core) */
+	UINT msize		/* Number of bytes to allocate */
+)
+{
+	return malloc(msize);	/* Allocate a new memory block with POSIX API */
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Free a memory block                                                    */
+/*------------------------------------------------------------------------*/
+
+void ff_memfree (
+	void* mblock	/* Pointer to the memory block to free (nothing to do if null) */
+)
+{
+	free(mblock);	/* Free the memory block with POSIX API */
+}
+
+#endif
+
+
+
+#if FF_FS_REENTRANT	/* Mutal exclusion */
+
+/*------------------------------------------------------------------------*/
+/* Create a Synchronization Object                                        */
+/*------------------------------------------------------------------------*/
+/* This function is called in f_mount() function to create a new
+/  synchronization object for the volume, such as semaphore and mutex.
+/  When a 0 is returned, the f_mount() function fails with FR_INT_ERR.
+*/
+
+//const osMutexDef_t Mutex[FF_VOLUMES];	/* Table of CMSIS-RTOS mutex */
+
+
+int ff_cre_syncobj (	/* 1:Function succeeded, 0:Could not create the sync object */
+	BYTE vol,			/* Corresponding volume (logical drive number) */
+	FF_SYNC_t* sobj		/* Pointer to return the created sync object */
+)
+{
+	/* Win32 */
+	*sobj = CreateMutex(NULL, FALSE, NULL);
+	return (int)(*sobj != INVALID_HANDLE_VALUE);
+
+	/* uITRON */
+//	T_CSEM csem = {TA_TPRI,1,1};
+//	*sobj = acre_sem(&csem);
+//	return (int)(*sobj > 0);
+
+	/* uC/OS-II */
+//	OS_ERR err;
+//	*sobj = OSMutexCreate(0, &err);
+//	return (int)(err == OS_NO_ERR);
+
+	/* FreeRTOS */
+//	*sobj = xSemaphoreCreateMutex();
+//	return (int)(*sobj != NULL);
+
+	/* CMSIS-RTOS */
+//	*sobj = osMutexCreate(&Mutex[vol]);
+//	return (int)(*sobj != NULL);
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Delete a Synchronization Object                                        */
+/*------------------------------------------------------------------------*/
+/* This function is called in f_mount() function to delete a synchronization
+/  object that created with ff_cre_syncobj() function. When a 0 is returned,
+/  the f_mount() function fails with FR_INT_ERR.
+*/
+
+int ff_del_syncobj (	/* 1:Function succeeded, 0:Could not delete due to an error */
+	FF_SYNC_t sobj		/* Sync object tied to the logical drive to be deleted */
+)
+{
+	/* Win32 */
+	return (int)CloseHandle(sobj);
+
+	/* uITRON */
+//	return (int)(del_sem(sobj) == E_OK);
+
+	/* uC/OS-II */
+//	OS_ERR err;
+//	OSMutexDel(sobj, OS_DEL_ALWAYS, &err);
+//	return (int)(err == OS_NO_ERR);
+
+	/* FreeRTOS */
+//  vSemaphoreDelete(sobj);
+//	return 1;
+
+	/* CMSIS-RTOS */
+//	return (int)(osMutexDelete(sobj) == osOK);
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Request Grant to Access the Volume                                     */
+/*------------------------------------------------------------------------*/
+/* This function is called on entering file functions to lock the volume.
+/  When a 0 is returned, the file function fails with FR_TIMEOUT.
+*/
+
+int ff_req_grant (	/* 1:Got a grant to access the volume, 0:Could not get a grant */
+	FF_SYNC_t sobj	/* Sync object to wait */
+)
+{
+	/* Win32 */
+	return (int)(WaitForSingleObject(sobj, FF_FS_TIMEOUT) == WAIT_OBJECT_0);
+
+	/* uITRON */
+//	return (int)(wai_sem(sobj) == E_OK);
+
+	/* uC/OS-II */
+//	OS_ERR err;
+//	OSMutexPend(sobj, FF_FS_TIMEOUT, &err));
+//	return (int)(err == OS_NO_ERR);
+
+	/* FreeRTOS */
+//	return (int)(xSemaphoreTake(sobj, FF_FS_TIMEOUT) == pdTRUE);
+
+	/* CMSIS-RTOS */
+//	return (int)(osMutexWait(sobj, FF_FS_TIMEOUT) == osOK);
+}
+
+
+/*------------------------------------------------------------------------*/
+/* Release Grant to Access the Volume                                     */
+/*------------------------------------------------------------------------*/
+/* This function is called on leaving file functions to unlock the volume.
+*/
+
+void ff_rel_grant (
+	FF_SYNC_t sobj	/* Sync object to be signaled */
+)
+{
+	/* Win32 */
+	ReleaseMutex(sobj);
+
+	/* uITRON */
+//	sig_sem(sobj);
+
+	/* uC/OS-II */
+//	OSMutexPost(sobj);
+
+	/* FreeRTOS */
+//	xSemaphoreGive(sobj);
+
+	/* CMSIS-RTOS */
+//	osMutexRelease(sobj);
+}
+
+#endif

code/components/esp-fatfs/vfs/esp_vfs_fat_mh.h

@@ -0,0 +1,321 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2022 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#pragma once
+#include <stddef.h>
+#include "esp_err.h"
+#include "driver/gpio.h"
+#include "driver/sdmmc_types.h"
+#include "driver/sdspi_host.h"
+#include "ff_mh.h"
+#include "wear_levelling.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @brief Register FATFS with VFS component
+ *
+ * This function registers given FAT drive in VFS, at the specified base path.
+ * If only one drive is used, fat_drive argument can be an empty string.
+ * Refer to FATFS library documentation on how to specify FAT drive.
+ * This function also allocates FATFS structure which should be used for f_mount
+ * call.
+ *
+ * @note This function doesn't mount the drive into FATFS, it just connects
+ *       POSIX and C standard library IO function with FATFS. You need to mount
+ *       desired drive into FATFS separately.
+ *
+ * @param base_path  path prefix where FATFS should be registered
+ * @param fat_drive  FATFS drive specification; if only one drive is used, can be an empty string
+ * @param max_files  maximum number of files which can be open at the same time
+ * @param[out] out_fs  pointer to FATFS structure which can be used for FATFS f_mount call is returned via this argument.
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_register was already called
+ *      - ESP_ERR_NO_MEM if not enough memory or too many VFSes already registered
+ */
+esp_err_t esp_vfs_fat_register(const char* base_path, const char* fat_drive,
+        size_t max_files, FATFS** out_fs);
+
+/**
+ * @brief Un-register FATFS from VFS
+ *
+ * @note FATFS structure returned by esp_vfs_fat_register is destroyed after
+ *       this call. Make sure to call f_mount function to unmount it before
+ *       calling esp_vfs_fat_unregister_ctx.
+ *       Difference between this function and the one above is that this one
+ *       will release the correct drive, while the one above will release
+ *       the last registered one
+ *
+ * @param base_path     path prefix where FATFS is registered. This is the same
+ *                      used when esp_vfs_fat_register was called
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if FATFS is not registered in VFS
+ */
+esp_err_t esp_vfs_fat_unregister_path(const char* base_path);
+
+
+/**
+ * @brief Configuration arguments for esp_vfs_fat_sdmmc_mount and esp_vfs_fat_spiflash_mount_rw_wl functions
+ */
+typedef struct {
+    /**
+     * If FAT partition can not be mounted, and this parameter is true,
+     * create partition table and format the filesystem.
+     */
+    bool format_if_mount_failed;
+    int max_files;                  ///< Max number of open files
+    /**
+     * If format_if_mount_failed is set, and mount fails, format the card
+     * with given allocation unit size. Must be a power of 2, between sector
+     * size and 128 * sector size.
+     * For SD cards, sector size is always 512 bytes. For wear_levelling,
+     * sector size is determined by CONFIG_WL_SECTOR_SIZE option.
+     *
+     * Using larger allocation unit size will result in higher read/write
+     * performance and higher overhead when storing small files.
+     *
+     * Setting this field to 0 will result in allocation unit set to the
+     * sector size.
+     */
+    size_t allocation_unit_size;
+    /**
+     * Enables real ff_disk_status function implementation for SD cards
+     * (ff_sdmmc_status). Possibly slows down IO performance.
+     *
+     * Try to enable if you need to handle situations when SD cards
+     * are not unmounted properly before physical removal
+     * or you are experiencing issues with SD cards.
+     *
+     * Doesn't do anything for other memory storage media.
+     */
+    bool disk_status_check_enable;
+} esp_vfs_fat_mount_config_t;
+
+// Compatibility definition
+typedef esp_vfs_fat_mount_config_t esp_vfs_fat_sdmmc_mount_config_t;
+
+/**
+ * @brief Convenience function to get FAT filesystem on SD card registered in VFS
+ *
+ * This is an all-in-one function which does the following:
+ * - initializes SDMMC driver or SPI driver with configuration in host_config
+ * - initializes SD card with configuration in slot_config
+ * - mounts FAT partition on SD card using FATFS library, with configuration in mount_config
+ * - registers FATFS library with VFS, with prefix given by base_prefix variable
+ *
+ * This function is intended to make example code more compact.
+ * For real world applications, developers should implement the logic of
+ * probing SD card, locating and mounting partition, and registering FATFS in VFS,
+ * with proper error checking and handling of exceptional conditions.
+ *
+ * @note Use this API to mount a card through SDSPI is deprecated. Please call
+ *       `esp_vfs_fat_sdspi_mount()` instead for that case.
+ *
+ * @param base_path     path where partition should be registered (e.g. "/sdcard")
+ * @param host_config   Pointer to structure describing SDMMC host. When using
+ *                      SDMMC peripheral, this structure can be initialized using
+ *                      SDMMC_HOST_DEFAULT() macro. When using SPI peripheral,
+ *                      this structure can be initialized using SDSPI_HOST_DEFAULT()
+ *                      macro.
+ * @param slot_config   Pointer to structure with slot configuration.
+ *                      For SDMMC peripheral, pass a pointer to sdmmc_slot_config_t
+ *                      structure initialized using SDMMC_SLOT_CONFIG_DEFAULT.
+ * @param mount_config  pointer to structure with extra parameters for mounting FATFS
+ * @param[out] out_card  if not NULL, pointer to the card information structure will be returned via this argument
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_sdmmc_mount was already called
+ *      - ESP_ERR_NO_MEM if memory can not be allocated
+ *      - ESP_FAIL if partition can not be mounted
+ *      - other error codes from SDMMC or SPI drivers, SDMMC protocol, or FATFS drivers
+ */
+esp_err_t esp_vfs_fat_sdmmc_mount(const char* base_path,
+    const sdmmc_host_t* host_config,
+    const void* slot_config,
+    const esp_vfs_fat_mount_config_t* mount_config,
+    sdmmc_card_t** out_card);
+
+/**
+ * @brief Convenience function to get FAT filesystem on SD card registered in VFS
+ *
+ * This is an all-in-one function which does the following:
+ * - initializes an SPI Master device based on the SPI Master driver with configuration in
+ *   slot_config, and attach it to an initialized SPI bus.
+ * - initializes SD card with configuration in host_config_input
+ * - mounts FAT partition on SD card using FATFS library, with configuration in mount_config
+ * - registers FATFS library with VFS, with prefix given by base_prefix variable
+ *
+ * This function is intended to make example code more compact.
+ * For real world applications, developers should implement the logic of
+ * probing SD card, locating and mounting partition, and registering FATFS in VFS,
+ * with proper error checking and handling of exceptional conditions.
+ *
+ * @note This function try to attach the new SD SPI device to the bus specified in host_config.
+ *       Make sure the SPI bus specified in `host_config->slot` have been initialized by
+ *       `spi_bus_initialize()` before.
+ *
+ * @param base_path     path where partition should be registered (e.g. "/sdcard")
+ * @param host_config_input Pointer to structure describing SDMMC host. This structure can be
+ *                          initialized using SDSPI_HOST_DEFAULT() macro.
+ * @param slot_config   Pointer to structure with slot configuration.
+ *                      For SPI peripheral, pass a pointer to sdspi_device_config_t
+ *                      structure initialized using SDSPI_DEVICE_CONFIG_DEFAULT().
+ * @param mount_config  pointer to structure with extra parameters for mounting FATFS
+ * @param[out] out_card If not NULL, pointer to the card information structure will be returned via
+ *                      this argument. It is suggested to hold this handle and use it to unmount the card later if
+ *                      needed. Otherwise it's not suggested to use more than one card at the same time and unmount one
+ *                      of them in your application.
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_sdmmc_mount was already called
+ *      - ESP_ERR_NO_MEM if memory can not be allocated
+ *      - ESP_FAIL if partition can not be mounted
+ *      - other error codes from SDMMC or SPI drivers, SDMMC protocol, or FATFS drivers
+ */
+esp_err_t esp_vfs_fat_sdspi_mount(const char* base_path,
+                                  const sdmmc_host_t* host_config_input,
+                                  const sdspi_device_config_t* slot_config,
+                                  const esp_vfs_fat_mount_config_t* mount_config,
+                                  sdmmc_card_t** out_card);
+
+/**
+ * @brief Unmount FAT filesystem and release resources acquired using esp_vfs_fat_sdmmc_mount
+ *
+ * @deprecated Use `esp_vfs_fat_sdcard_unmount()` instead.
+ *
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_sdmmc_mount hasn't been called
+ */
+esp_err_t esp_vfs_fat_sdmmc_unmount(void);
+
+/**
+ * @brief Unmount an SD card from the FAT filesystem and release resources acquired using
+ *        `esp_vfs_fat_sdmmc_mount()` or `esp_vfs_fat_sdspi_mount()`
+ *
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_ARG if the card argument is unregistered
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_sdmmc_mount hasn't been called
+ */
+esp_err_t esp_vfs_fat_sdcard_unmount(const char* base_path, sdmmc_card_t *card);
+
+/**
+ * @brief Convenience function to initialize FAT filesystem in SPI flash and register it in VFS
+ *
+ * This is an all-in-one function which does the following:
+ *
+ * - finds the partition with defined partition_label. Partition label should be
+ *   configured in the partition table.
+ * - initializes flash wear levelling library on top of the given partition
+ * - mounts FAT partition using FATFS library on top of flash wear levelling
+ *   library
+ * - registers FATFS library with VFS, with prefix given by base_prefix variable
+ *
+ * This function is intended to make example code more compact.
+ *
+ * @param base_path        path where FATFS partition should be mounted (e.g. "/spiflash")
+ * @param partition_label  label of the partition which should be used
+ * @param mount_config     pointer to structure with extra parameters for mounting FATFS
+ * @param[out] wl_handle   wear levelling driver handle
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_NOT_FOUND if the partition table does not contain FATFS partition with given label
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_spiflash_mount_rw_wl was already called
+ *      - ESP_ERR_NO_MEM if memory can not be allocated
+ *      - ESP_FAIL if partition can not be mounted
+ *      - other error codes from wear levelling library, SPI flash driver, or FATFS drivers
+ */
+esp_err_t esp_vfs_fat_spiflash_mount_rw_wl(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config,
+    wl_handle_t* wl_handle);
+
+/**
+ * @brief Unmount FAT filesystem and release resources acquired using esp_vfs_fat_spiflash_mount_rw_wl
+ *
+ * @param base_path  path where partition should be registered (e.g. "/spiflash")
+ * @param wl_handle  wear levelling driver handle returned by esp_vfs_fat_spiflash_mount_rw_wl
+ *
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_spiflash_mount_rw_wl hasn't been called
+ */
+esp_err_t esp_vfs_fat_spiflash_unmount_rw_wl(const char* base_path, wl_handle_t wl_handle);
+
+/**
+ * @brief Convenience function to initialize read-only FAT filesystem and register it in VFS
+ *
+ * This is an all-in-one function which does the following:
+ *
+ * - finds the partition with defined partition_label. Partition label should be
+ *   configured in the partition table.
+ * - mounts FAT partition using FATFS library
+ * - registers FATFS library with VFS, with prefix given by base_prefix variable
+ *
+ * @note Wear levelling is not used when FAT is mounted in read-only mode using this function.
+ *
+ * @param base_path        path where FATFS partition should be mounted (e.g. "/spiflash")
+ * @param partition_label  label of the partition which should be used
+ * @param mount_config     pointer to structure with extra parameters for mounting FATFS
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_NOT_FOUND if the partition table does not contain FATFS partition with given label
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_spiflash_mount_ro was already called for the same partition
+ *      - ESP_ERR_NO_MEM if memory can not be allocated
+ *      - ESP_FAIL if partition can not be mounted
+ *      - other error codes from SPI flash driver, or FATFS drivers
+ */
+esp_err_t esp_vfs_fat_spiflash_mount_ro(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config);
+
+/**
+ * @brief Unmount FAT filesystem and release resources acquired using esp_vfs_fat_spiflash_mount_ro
+ *
+ * @param base_path  path where partition should be registered (e.g. "/spiflash")
+ * @param partition_label label of partition to be unmounted
+ *
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if esp_vfs_fat_spiflash_mount_rw_wl hasn't been called
+ */
+esp_err_t esp_vfs_fat_spiflash_unmount_ro(const char* base_path, const char* partition_label);
+
+esp_err_t esp_vfs_fat_spiflash_mount(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config,
+    wl_handle_t* wl_handle)
+    __attribute__((deprecated("esp_vfs_fat_spiflash_mount is deprecated, please use esp_vfs_fat_spiflash_mount_rw_wl instead")));
+esp_err_t esp_vfs_fat_spiflash_unmount(const char* base_path, wl_handle_t wl_handle)
+    __attribute__((deprecated("esp_vfs_fat_spiflash_unmount is deprecated, please use esp_vfs_fat_spiflash_unmount_rw_wl instead")));
+esp_err_t esp_vfs_fat_rawflash_mount(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config)
+    __attribute__((deprecated("esp_vfs_fat_rawflash_mount is deprecated, please use esp_vfs_fat_spiflash_mount_ro instead")));
+esp_err_t esp_vfs_fat_rawflash_unmount(const char* base_path, const char* partition_label)
+    __attribute__((deprecated("esp_vfs_fat_rawflash_unmount is deprecated, please use esp_vfs_fat_spiflash_unmount_ro instead")));
+
+/**
+ * @brief  Get information for FATFS partition
+ *
+ * @param base_path  Path where partition should be registered (e.g. "/spiflash")
+ * @param[out] out_total_bytes  Size of the file system
+ * @param[out] out_free_bytes   Current used bytes in the file system
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_STATE if partition not found
+ *      - ESP_FAIL if another FRESULT error (saved in errno)
+ */
+esp_err_t esp_vfs_fat_info(const char* base_path, uint64_t* out_total_bytes, uint64_t* out_free_bytes);
+
+#ifdef __cplusplus
+}
+#endif

code/components/esp-fatfs/vfs/vfs_fat_internal_mh.h

@@ -1,4 +1,4 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
+// Copyright 2018 Espressif Systems (Shanghai) PTE LTD
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -12,19 +12,19 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#include <stdint.h>
-#include <stdlib.h>
+#pragma once
 
-#include <common_functions.h>
+#include "esp_vfs_fat_mh.h"
+#include <sys/param.h>
+#include <stddef.h>
 
-void esp_nn_relu6_s8_ansi(int8_t *data, uint16_t size)
+static inline size_t esp_vfs_fat_get_allocation_unit_size(
+        size_t sector_size, size_t requested_size)
 {
-    int32_t i;
-
-    for (i = 0; i < size; i++) {
-        int32_t ip = data[i];
-
-        ip = max(ip, 0);
-        data[i] = min(ip, 6);
-    }
+    size_t alloc_unit_size = requested_size;
+    const size_t max_sectors_per_cylinder = 128;
+    const size_t max_size = sector_size * max_sectors_per_cylinder;
+    alloc_unit_size = MAX(alloc_unit_size, sector_size);
+    alloc_unit_size = MIN(alloc_unit_size, max_size);
+    return alloc_unit_size;
 }

code/components/esp-fatfs/vfs/vfs_fat_sdmmc_mh.c

@@ -0,0 +1,375 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2022 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "esp_log.h"
+#include "esp_vfs.h"
+#include "esp_vfs_fat_mh.h"
+#include "vfs_fat_internal_mh.h"
+#include "driver/sdspi_host.h"
+#include "sdmmc_cmd_mh.h"
+#include "diskio_impl_mh.h"
+#include "diskio_sdmmc_mh.h"
+#include "soc/soc_caps.h"
+#include "driver/sdmmc_defs.h"
+
+#if SOC_SDMMC_HOST_SUPPORTED
+#include "driver/sdmmc_host.h"
+#endif
+
+static const char* TAG = "vfs_fat_sdmmc";
+static sdmmc_card_t* s_card = NULL;
+static uint8_t s_pdrv = FF_DRV_NOT_USED;
+static char * s_base_path = NULL;
+
+#define CHECK_EXECUTE_RESULT(err, str) do { \
+    if ((err) !=ESP_OK) { \
+        ESP_LOGE(TAG, str" (0x%x).", err); \
+        goto cleanup; \
+    } \
+    } while(0)
+
+static void call_host_deinit(const sdmmc_host_t *host_config);
+static esp_err_t partition_card(const esp_vfs_fat_mount_config_t *mount_config,
+                                const char *drv, sdmmc_card_t *card, BYTE pdrv);
+
+static esp_err_t mount_prepare_mem(const char *base_path,
+        BYTE *out_pdrv,
+        char **out_dup_path,
+        sdmmc_card_t** out_card)
+{
+    esp_err_t err = ESP_OK;
+    char* dup_path = NULL;
+    sdmmc_card_t* card = NULL;
+
+    // connect SDMMC driver to FATFS
+    BYTE pdrv = FF_DRV_NOT_USED;
+    if (ff_diskio_get_drive(&pdrv) != ESP_OK || pdrv == FF_DRV_NOT_USED) {
+        ESP_LOGD(TAG, "the maximum count of volumes is already mounted");
+        return ESP_ERR_NO_MEM;
+
+    }
+
+    // not using ff_memalloc here, as allocation in internal RAM is preferred
+    card = (sdmmc_card_t*)malloc(sizeof(sdmmc_card_t));
+    if (card == NULL) {
+        ESP_LOGD(TAG, "could not locate new sdmmc_card_t");
+        err = ESP_ERR_NO_MEM;
+        goto cleanup;
+    }
+
+    dup_path = strdup(base_path);
+    if(!dup_path){
+        ESP_LOGD(TAG, "could not copy base_path");
+        err = ESP_ERR_NO_MEM;
+        goto cleanup;
+    }
+
+    *out_card = card;
+    *out_pdrv = pdrv;
+    *out_dup_path = dup_path;
+    return ESP_OK;
+cleanup:
+    free(card);
+    free(dup_path);
+    return err;
+}
+
+static esp_err_t mount_to_vfs_fat(const esp_vfs_fat_mount_config_t *mount_config, sdmmc_card_t *card, uint8_t pdrv,
+                                  const char *base_path)
+{
+    FATFS* fs = NULL;
+    esp_err_t err;
+    ff_diskio_register_sdmmc(pdrv, card);
+    ff_sdmmc_set_disk_status_check(pdrv, mount_config->disk_status_check_enable);
+    ESP_LOGD(TAG, "using pdrv=%i", pdrv);
+    char drv[3] = {(char)('0' + pdrv), ':', 0};
+
+    // connect FATFS to VFS
+    err = esp_vfs_fat_register(base_path, drv, mount_config->max_files, &fs);
+    if (err == ESP_ERR_INVALID_STATE) {
+        // it's okay, already registered with VFS
+    } else if (err != ESP_OK) {
+        ESP_LOGD(TAG, "esp_vfs_fat_register failed 0x(%x)", err);
+        goto fail;
+    }
+
+    // Try to mount partition
+    FRESULT res = f_mount(fs, drv, 1);
+    if (res != FR_OK) {
+        err = ESP_FAIL;
+        ESP_LOGW(TAG, "failed to mount card (%d)", res);
+        if (!((res == FR_NO_FILESYSTEM || res == FR_INT_ERR)
+              && mount_config->format_if_mount_failed)) {
+            goto fail;
+        }
+
+        err = partition_card(mount_config, drv, card, pdrv);
+        if (err != ESP_OK) {
+            goto fail;
+        }
+
+        ESP_LOGW(TAG, "mounting again");
+        res = f_mount(fs, drv, 0);
+        if (res != FR_OK) {
+            err = ESP_FAIL;
+            ESP_LOGD(TAG, "f_mount failed after formatting (%d)", res);
+            goto fail;
+        }
+    }
+    return ESP_OK;
+
+fail:
+    if (fs) {
+        f_mount(NULL, drv, 0);
+    }
+    esp_vfs_fat_unregister_path(base_path);
+    ff_diskio_unregister(pdrv);
+    return err;
+}
+
+static esp_err_t partition_card(const esp_vfs_fat_mount_config_t *mount_config,
+                                const char *drv, sdmmc_card_t *card, BYTE pdrv)
+{
+    FRESULT res = FR_OK;
+    esp_err_t err;
+    const size_t workbuf_size = 4096;
+    void* workbuf = NULL;
+    ESP_LOGW(TAG, "partitioning card");
+
+    workbuf = ff_memalloc(workbuf_size);
+    if (workbuf == NULL) {
+        return ESP_ERR_NO_MEM;
+    }
+
+    LBA_t plist[] = {100, 0, 0, 0};
+    res = f_fdisk(pdrv, plist, workbuf);
+    if (res != FR_OK) {
+        err = ESP_FAIL;
+        ESP_LOGD(TAG, "f_fdisk failed (%d)", res);
+        goto fail;
+    }
+    size_t alloc_unit_size = esp_vfs_fat_get_allocation_unit_size(
+                card->csd.sector_size,
+                mount_config->allocation_unit_size);
+    ESP_LOGW(TAG, "formatting card, allocation unit size=%d", alloc_unit_size);
+    const MKFS_PARM opt = {(BYTE)FM_ANY, 0, 0, 0, alloc_unit_size};
+    res = f_mkfs(drv, &opt, workbuf, workbuf_size);
+    if (res != FR_OK) {
+        err = ESP_FAIL;
+        ESP_LOGD(TAG, "f_mkfs failed (%d)", res);
+        goto fail;
+    }
+
+    free(workbuf);
+    return ESP_OK;
+fail:
+    free(workbuf);
+    return err;
+}
+
+#if SOC_SDMMC_HOST_SUPPORTED
+static esp_err_t init_sdmmc_host(int slot, const void *slot_config, int *out_slot)
+{
+    *out_slot = slot;
+    return sdmmc_host_init_slot(slot, (const sdmmc_slot_config_t*) slot_config);
+}
+
+
+esp_err_t esp_vfs_fat_sdmmc_mount(const char* base_path,
+                                  const sdmmc_host_t* host_config,
+                                  const void* slot_config,
+                                  const esp_vfs_fat_mount_config_t* mount_config,
+                                  sdmmc_card_t** out_card)
+{
+    esp_err_t err;
+    int card_handle = -1;   //uninitialized
+    sdmmc_card_t* card = NULL;
+    BYTE pdrv = FF_DRV_NOT_USED;
+    char* dup_path = NULL;
+    bool host_inited = false;
+
+    err = mount_prepare_mem(base_path, &pdrv, &dup_path, &card);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "mount_prepare failed");
+        return err;
+    }
+
+    err = (*host_config->init)();
+    CHECK_EXECUTE_RESULT(err, "host init failed");
+    //deinit() needs to be called to revert the init
+    host_inited = true;
+    //If this failed (indicated by card_handle != -1), slot deinit needs to called()
+    //leave card_handle as is to indicate that (though slot deinit not implemented yet.
+    err = init_sdmmc_host(host_config->slot, slot_config, &card_handle);
+    CHECK_EXECUTE_RESULT(err, "slot init failed");
+
+    // probe and initialize card
+    err = sdmmc_card_init(host_config, card);
+    CHECK_EXECUTE_RESULT(err, "sdmmc_card_init failed");
+
+    err = mount_to_vfs_fat(mount_config, card, pdrv, dup_path);
+    CHECK_EXECUTE_RESULT(err, "mount_to_vfs failed");
+
+    if (out_card != NULL) {
+        *out_card = card;
+    }
+    if (s_card == NULL) {
+        //store the ctx locally to be back-compatible
+        s_card = card;
+        s_pdrv = pdrv;
+        s_base_path = dup_path;
+    } else {
+        free(dup_path);
+    }
+    return ESP_OK;
+cleanup:
+    if (host_inited) {
+        call_host_deinit(host_config);
+    }
+    free(card);
+    free(dup_path);
+    return err;
+}
+#endif
+
+static esp_err_t init_sdspi_host(int slot, const void *slot_config, int *out_slot)
+{
+    esp_err_t err = sdspi_host_init_device((const sdspi_device_config_t*)slot_config, out_slot);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG,
+"Failed to attach sdspi device onto an SPI bus (rc=0x%x), please initialize the \
+bus first and check the device parameters."
+            , err);
+    }
+    return err;
+}
+
+esp_err_t esp_vfs_fat_sdspi_mount(const char* base_path,
+                                  const sdmmc_host_t* host_config_input,
+                                  const sdspi_device_config_t* slot_config,
+                                  const esp_vfs_fat_mount_config_t* mount_config,
+                                  sdmmc_card_t** out_card)
+{
+    const sdmmc_host_t* host_config = host_config_input;
+    esp_err_t err;
+    int card_handle = -1;   //uninitialized
+    bool host_inited = false;
+    BYTE pdrv = FF_DRV_NOT_USED;
+    sdmmc_card_t* card = NULL;
+    char* dup_path = NULL;
+
+    err = mount_prepare_mem(base_path, &pdrv, &dup_path, &card);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "mount_prepare failed");
+        return err;
+    }
+
+    //the init() function is usually empty, doesn't require any deinit to revert it
+    err = (*host_config->init)();
+    CHECK_EXECUTE_RESULT(err, "host init failed");
+
+    err = init_sdspi_host(host_config->slot, slot_config, &card_handle);
+    CHECK_EXECUTE_RESULT(err, "slot init failed");
+    //Set `host_inited` to true to indicate that host_config->deinit() needs
+    //to be called to revert `init_sdspi_host`
+    host_inited = true;
+
+    /*
+     * The `slot` argument inside host_config should be replaced by the SD SPI handled returned
+     * above. But the input pointer is const, so create a new variable.
+     */
+    sdmmc_host_t new_config;
+    if (card_handle != host_config->slot) {
+        new_config = *host_config_input;
+        host_config = &new_config;
+        new_config.slot = card_handle;
+    }
+
+    // probe and initialize card
+    err = sdmmc_card_init(host_config, card);
+    CHECK_EXECUTE_RESULT(err, "sdmmc_card_init failed");
+
+    err = mount_to_vfs_fat(mount_config, card, pdrv, dup_path);
+    CHECK_EXECUTE_RESULT(err, "mount_to_vfs failed");
+
+    if (out_card != NULL) {
+        *out_card = card;
+    }
+    if (s_card == NULL) {
+        //store the ctx locally to be back-compatible
+        s_card = card;
+        s_pdrv = pdrv;
+        s_base_path = dup_path;
+    } else {
+        free(dup_path);
+    }
+    return ESP_OK;
+
+cleanup:
+    if (host_inited) {
+        call_host_deinit(host_config);
+    }
+    free(card);
+    free(dup_path);
+    return err;
+
+}
+
+static void local_card_remove(void)
+{
+    s_card = NULL;
+    free(s_base_path);
+    s_base_path = NULL;
+    s_pdrv = FF_DRV_NOT_USED;
+}
+
+static void call_host_deinit(const sdmmc_host_t *host_config)
+{
+    if (host_config->flags & SDMMC_HOST_FLAG_DEINIT_ARG) {
+        host_config->deinit_p(host_config->slot);
+    } else {
+        host_config->deinit();
+    }
+}
+
+static esp_err_t unmount_card_core(const char *base_path, sdmmc_card_t *card)
+{
+    BYTE pdrv = ff_diskio_get_pdrv_card(card);
+    if (pdrv == 0xff) {
+        return ESP_ERR_INVALID_ARG;
+    }
+
+    // unmount
+    char drv[3] = {(char)('0' + pdrv), ':', 0};
+    f_mount(0, drv, 0);
+    // release SD driver
+    ff_diskio_unregister(pdrv);
+
+    call_host_deinit(&card->host);
+    free(card);
+
+    esp_err_t err = esp_vfs_fat_unregister_path(base_path);
+    return err;
+}
+
+esp_err_t esp_vfs_fat_sdmmc_unmount(void)
+{
+    sdmmc_card_t* card = s_card;
+    esp_err_t err = unmount_card_core(s_base_path, card);
+    local_card_remove();
+    return err;
+}
+
+esp_err_t esp_vfs_fat_sdcard_unmount(const char *base_path, sdmmc_card_t *card)
+{
+    esp_err_t err = unmount_card_core(base_path, card);
+    if (s_card == card) {
+        local_card_remove();
+    }
+    return err;
+}

code/components/esp-fatfs/vfs/vfs_fat_spiflash_mh.c

@@ -0,0 +1,218 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2022 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "esp_log.h"
+#include "esp_vfs.h"
+#include "esp_vfs_fat_mh.h"
+#include "vfs_fat_internal_mh.h"
+#include "diskio_impl_mh.h"
+
+#include "diskio_rawflash_mh.h"
+
+#include "wear_levelling.h"
+#include "diskio_wl_mh.h"
+
+static const char* TAG = "vfs_fat_spiflash";
+
+esp_err_t esp_vfs_fat_spiflash_mount_rw_wl(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config,
+    wl_handle_t* wl_handle)
+{
+    esp_err_t result = ESP_OK;
+    const size_t workbuf_size = 4096;
+    void *workbuf = NULL;
+
+    esp_partition_subtype_t subtype = partition_label ?
+            ESP_PARTITION_SUBTYPE_ANY : ESP_PARTITION_SUBTYPE_DATA_FAT;
+    const esp_partition_t *data_partition = esp_partition_find_first(ESP_PARTITION_TYPE_DATA,
+                                                subtype, partition_label);
+    if (data_partition == NULL) {
+        ESP_LOGE(TAG, "Failed to find FATFS partition (type='data', subtype='fat', partition_label='%s'). Check the partition table.", partition_label);
+        return ESP_ERR_NOT_FOUND;
+    }
+
+    result = wl_mount(data_partition, wl_handle);
+    if (result != ESP_OK) {
+        ESP_LOGE(TAG, "failed to mount wear levelling layer. result = %i", result);
+        return result;
+    }
+    // connect driver to FATFS
+    BYTE pdrv = 0xFF;
+    if (ff_diskio_get_drive(&pdrv) != ESP_OK) {
+        ESP_LOGD(TAG, "the maximum count of volumes is already mounted");
+        return ESP_ERR_NO_MEM;
+    }
+    ESP_LOGD(TAG, "using pdrv=%i", pdrv);
+    char drv[3] = {(char)('0' + pdrv), ':', 0};
+
+    result = ff_diskio_register_wl_partition(pdrv, *wl_handle);
+    if (result != ESP_OK) {
+        ESP_LOGE(TAG, "ff_diskio_register_wl_partition failed pdrv=%i, error - 0x(%x)", pdrv, result);
+        goto fail;
+    }
+    FATFS *fs;
+    result = esp_vfs_fat_register(base_path, drv, mount_config->max_files, &fs);
+    if (result == ESP_ERR_INVALID_STATE) {
+        // it's okay, already registered with VFS
+    } else if (result != ESP_OK) {
+        ESP_LOGD(TAG, "esp_vfs_fat_register failed 0x(%x)", result);
+        goto fail;
+    }
+
+    // Try to mount partition
+    FRESULT fresult = f_mount(fs, drv, 1);
+    if (fresult != FR_OK) {
+        ESP_LOGW(TAG, "f_mount failed (%d)", fresult);
+        if (!((fresult == FR_NO_FILESYSTEM || fresult == FR_INT_ERR)
+              && mount_config->format_if_mount_failed)) {
+            result = ESP_FAIL;
+            goto fail;
+        }
+        workbuf = ff_memalloc(workbuf_size);
+        if (workbuf == NULL) {
+            result = ESP_ERR_NO_MEM;
+            goto fail;
+        }
+        size_t alloc_unit_size = esp_vfs_fat_get_allocation_unit_size(
+                CONFIG_WL_SECTOR_SIZE,
+                mount_config->allocation_unit_size);
+        ESP_LOGI(TAG, "Formatting FATFS partition, allocation unit size=%d", alloc_unit_size);
+        const MKFS_PARM opt = {(BYTE)(FM_ANY | FM_SFD), 0, 0, 0, alloc_unit_size};
+        fresult = f_mkfs(drv, &opt, workbuf, workbuf_size);
+        if (fresult != FR_OK) {
+            result = ESP_FAIL;
+            ESP_LOGE(TAG, "f_mkfs failed (%d)", fresult);
+            goto fail;
+        }
+        free(workbuf);
+        workbuf = NULL;
+        ESP_LOGI(TAG, "Mounting again");
+        fresult = f_mount(fs, drv, 0);
+        if (fresult != FR_OK) {
+            result = ESP_FAIL;
+            ESP_LOGE(TAG, "f_mount failed after formatting (%d)", fresult);
+            goto fail;
+        }
+    }
+    return ESP_OK;
+
+fail:
+    free(workbuf);
+    esp_vfs_fat_unregister_path(base_path);
+    ff_diskio_unregister(pdrv);
+    return result;
+}
+
+esp_err_t esp_vfs_fat_spiflash_unmount_rw_wl(const char* base_path, wl_handle_t wl_handle)
+{
+    BYTE pdrv = ff_diskio_get_pdrv_wl(wl_handle);
+    if (pdrv == 0xff) {
+        return ESP_ERR_INVALID_STATE;
+    }
+    char drv[3] = {(char)('0' + pdrv), ':', 0};
+
+    f_mount(0, drv, 0);
+    ff_diskio_unregister(pdrv);
+    ff_diskio_clear_pdrv_wl(wl_handle);
+    // release partition driver
+    esp_err_t err_drv = wl_unmount(wl_handle);
+    esp_err_t err = esp_vfs_fat_unregister_path(base_path);
+    if (err == ESP_OK) err = err_drv;
+    return err;
+}
+
+
+esp_err_t esp_vfs_fat_spiflash_mount_ro(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config)
+{
+    esp_err_t result = ESP_OK;
+
+    const esp_partition_t *data_partition = esp_partition_find_first(ESP_PARTITION_TYPE_DATA,
+            ESP_PARTITION_SUBTYPE_DATA_FAT, partition_label);
+    if (data_partition == NULL) {
+        ESP_LOGE(TAG, "Failed to find FATFS partition (type='data', subtype='fat', partition_label='%s'). Check the partition table.", partition_label);
+        return ESP_ERR_NOT_FOUND;
+    }
+
+    // connect driver to FATFS
+    BYTE pdrv = 0xFF;
+    if (ff_diskio_get_drive(&pdrv) != ESP_OK) {
+        ESP_LOGD(TAG, "the maximum count of volumes is already mounted");
+        return ESP_ERR_NO_MEM;
+    }
+    ESP_LOGD(TAG, "using pdrv=%i", pdrv);
+    char drv[3] = {(char)('0' + pdrv), ':', 0};
+
+    result = ff_diskio_register_raw_partition(pdrv, data_partition);
+    if (result != ESP_OK) {
+        ESP_LOGE(TAG, "ff_diskio_register_raw_partition failed pdrv=%i, error - 0x(%x)", pdrv, result);
+        goto fail;
+    }
+
+    FATFS *fs;
+    result = esp_vfs_fat_register(base_path, drv, mount_config->max_files, &fs);
+    if (result == ESP_ERR_INVALID_STATE) {
+        // it's okay, already registered with VFS
+    } else if (result != ESP_OK) {
+        ESP_LOGD(TAG, "esp_vfs_fat_register failed 0x(%x)", result);
+        goto fail;
+    }
+
+    // Try to mount partition
+    FRESULT fresult = f_mount(fs, drv, 1);
+    if (fresult != FR_OK) {
+        ESP_LOGW(TAG, "f_mount failed (%d)", fresult);
+        result = ESP_FAIL;
+        goto fail;
+    }
+    return ESP_OK;
+
+fail:
+    esp_vfs_fat_unregister_path(base_path);
+    ff_diskio_unregister(pdrv);
+    return result;
+}
+
+esp_err_t esp_vfs_fat_spiflash_unmount_ro(const char* base_path, const char* partition_label)
+{
+    const esp_partition_t *data_partition = esp_partition_find_first(ESP_PARTITION_TYPE_DATA,
+            ESP_PARTITION_SUBTYPE_DATA_FAT, partition_label);
+
+    if (data_partition == NULL) {
+        ESP_LOGE(TAG, "Failed to find FATFS partition (type='data', subtype='fat', partition_label='%s'). Check the partition table.", partition_label);
+        return ESP_ERR_NOT_FOUND;
+    }
+
+    BYTE pdrv = ff_diskio_get_pdrv_raw(data_partition);
+    if (pdrv == 0xff) {
+        return ESP_ERR_INVALID_STATE;
+    }
+    char drv[3] = {(char)('0' + pdrv), ':', 0};
+
+    f_mount(0, drv, 0);
+    ff_diskio_unregister(pdrv);
+    esp_err_t err = esp_vfs_fat_unregister_path(base_path);
+    return err;
+}
+
+
+esp_err_t esp_vfs_fat_spiflash_mount(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config,
+    wl_handle_t* wl_handle)
+    __attribute__((alias("esp_vfs_fat_spiflash_mount_rw_wl")));
+esp_err_t esp_vfs_fat_spiflash_unmount(const char* base_path, wl_handle_t wl_handle)
+    __attribute__((alias("esp_vfs_fat_spiflash_unmount_rw_wl")));
+esp_err_t esp_vfs_fat_rawflash_mount(const char* base_path,
+    const char* partition_label,
+    const esp_vfs_fat_mount_config_t* mount_config)
+    __attribute__((alias("esp_vfs_fat_spiflash_mount_ro")));
+esp_err_t esp_vfs_fat_rawflash_unmount(const char* base_path, const char* partition_label)
+    __attribute__((alias("esp_vfs_fat_spiflash_unmount_ro")));

code/components/esp-nn/.gitignore

@@ -1,57 +0,0 @@
-.config
-*.o
-*.i
-*.s
-*.orig
-*.pyc
-
-# gtags
-GTAGS
-GRTAGS
-GPATH
-
-# emacs
-.dir-locals.el
-
-# emacs temp file suffixes
-*~
-.#*
-\#*#
-
-# eclipse setting
-.settings
-
-# MacOS directory files
-.DS_Store
-
-# Example project files
-examples/**/sdkconfig
-examples/**/sdkconfig.old
-examples/**/build
-
-# Test app files
-test_app/build
-test_app/sdkconfig
-test_app/sdkconfig.old
-
-# Doc build artifacts
-docs/_build/
-docs/doxygen-warning-log.txt
-docs/sphinx-warning-log.txt
-docs/sphinx-warning-log-sanitized.txt
-docs/xml/
-docs/xml_in/
-docs/man/
-docs/doxygen_sqlite3.db
-
-TEST_LOGS
-
-
-# gcov coverage reports
-*.gcda
-*.gcno
-coverage.info
-coverage_report/
-
-# VS Code Settings
-.vscode/

code/components/esp-nn/.gitlab-ci.yml

@@ -1,55 +0,0 @@
-stages:
-  - build
-
-variables:
-  BATCH_BUILD: "1"
-  V: "0"
-  MAKEFLAGS: "-j8 --no-keep-going"
-  IDF_PATH: "$CI_PROJECT_DIR/esp-idf"
-  LOG_PATH: "$CI_PROJECT_DIR"
-
-.set_git_config: &set_git_config
-  # Set git config
-  - git config user.email "test@espressif.com"
-  - git config user.name "Espressif"
-
-.add_ssh_key: &add_ssh_key
-  # Add gitlab ssh key
-  - mkdir -p ~/.ssh
-  - chmod 700 ~/.ssh
-  - echo -n $GITLAB_KEY > ~/.ssh/id_rsa_base64
-  - base64 --decode --ignore-garbage ~/.ssh/id_rsa_base64 > ~/.ssh/id_rsa
-  - chmod 600 ~/.ssh/id_rsa
-  - echo -e "Host gitlab.espressif.cn\n\tStrictHostKeyChecking no\n" >> ~/.ssh/config
-
-before_script:
-  # Add gitlab ssh key
-  - *add_ssh_key
-  # Set git config
-  - *set_git_config
-
-.build_esp32s3: &build_esp32s3
-  - idf.py set-target esp32s3 build
-
-.build_esp32: &build_esp32
-  - idf.py set-target esp32 build
-
-build_demo:
-  stage: build
-  image: $CI_DOCKER_REGISTRY/esp32-ci-env:esp-nn
-  tags:
-    - build
-  script:
-    # Clone IDF
-    - git clone --recursive --single-branch -b release/v4.4 --reference-if-able /local_references/gitlab/ https://gitlab-ci-token:${BOT_TOKEN}@gitlab.espressif.cn:6688/espressif/esp-idf.git
-    - cd esp-idf
-    - ./install.sh
-    - . ./export.sh
-    - cd ..
-    # Build examples now
-    - cd test_app
-    # Build esp32s3
-    - *build_esp32s3
-    # Build esp32
-    - *build_esp32
-    - cd -

code/components/esp-nn/CMakeLists.txt

@@ -1,48 +0,0 @@
-idf_build_get_property(idf_target IDF_TARGET)
-
-set(c_srcs
-    "src/activation_functions/esp_nn_relu_ansi.c"
-    "src/basic_math/esp_nn_add_ansi.c"
-    "src/basic_math/esp_nn_mul_ansi.c"
-    "src/convolution/esp_nn_conv_ansi.c"
-    "src/convolution/esp_nn_depthwise_conv_ansi.c"
-    "src/fully_connected/esp_nn_fully_connected_ansi.c"
-    "src/softmax/esp_nn_softmax_ansi.c"
-    "src/softmax/esp_nn_softmax_opt.c"
-    "src/pooling/esp_nn_avg_pool_ansi.c"
-    "src/pooling/esp_nn_max_pool_ansi.c")
-
-if(CONFIG_IDF_TARGET_ESP32S3)
-    set(s3_srcs
-        "src/common/esp_nn_common_functions_esp32s3.S"
-        "src/common/esp_nn_multiply_by_quantized_mult_esp32s3.S"
-        "src/common/esp_nn_multiply_by_quantized_mult_ver1_esp32s3.S"
-        "src/activation_functions/esp_nn_relu_s8_esp32s3.S"
-        "src/basic_math/esp_nn_add_s8_esp32s3.S"
-        "src/basic_math/esp_nn_mul_s8_esp32s3.S"
-        "src/convolution/esp_nn_conv_esp32s3.c"
-        "src/convolution/esp_nn_depthwise_conv_s8_esp32s3.c"
-        "src/convolution/esp_nn_conv_s16_mult8_esp32s3.S"
-        "src/convolution/esp_nn_conv_s16_mult8_1x1_esp32s3.S"
-        "src/convolution/esp_nn_conv_s16_mult4_1x1_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s8_mult1_3x3_padded_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s16_mult1_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s16_mult1_3x3_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s16_mult1_3x3_no_pad_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s16_mult8_3x3_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s16_mult4_esp32s3.S"
-        "src/convolution/esp_nn_depthwise_conv_s16_mult8_esp32s3.S"
-        "src/fully_connected/esp_nn_fully_connected_s8_esp32s3.S"
-        "src/pooling/esp_nn_max_pool_s8_esp32s3.S"
-        "src/pooling/esp_nn_avg_pool_s8_esp32s3.S")
-endif()
-
-idf_component_register(SRCS "${c_srcs}"
-                            "${s3_srcs}"
-                       INCLUDE_DIRS "include" "src/common")
-
-if(CONFIG_IDF_TARGET_ESP32S3)
-    target_compile_options(${COMPONENT_LIB} PRIVATE -mlongcalls -fno-unroll-loops -O2 -Wno-unused-function)
-else()
-    target_compile_options(${COMPONENT_LIB} PRIVATE -Wno-unused-function)
-endif()

code/components/esp-nn/Kconfig.projbuild

@@ -1,29 +0,0 @@
-menu "ESP-NN"
-
-choice NN_OPTIMIZATIONS
-   bool "Optimization for nn functions"
-   default NN_OPTIMIZED
-   help
-      Use ANSI-C versions for verification and debug purpose.
-      Optimisations are automatically picked up for a chipset.
-      For ESP32-S3, assembly Optimisations are selected.
-      For ESP32, just the ANSI C versions are selected for now.
-
-config NN_ANSI_C
-   bool "ANSI C"
-   help
-      ANSI C versions for verification and debug purposes.
-config NN_OPTIMIZED
-   bool "Optimized versions"
-   help
-      Optimisations are automatically picked up for a chipset.
-      For ESP32-S3, assembly Optimisations are selected.
-      For ESP32, just the ANSI C versions are selected for now.
-endchoice
-
-config NN_OPTIMIZATIONS
-   int
-   default 0 if NN_ANSI_C
-   default 1 if NN_OPTIMIZED
-
-endmenu

code/components/esp-nn/LICENSE

@@ -1,202 +0,0 @@
-
-                                 Apache License
-                           Version 2.0, January 2004
-                        http://www.apache.org/licenses/
-
-   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
-
-   1. Definitions.
-
-      "License" shall mean the terms and conditions for use, reproduction,
-      and distribution as defined by Sections 1 through 9 of this document.
-
-      "Licensor" shall mean the copyright owner or entity authorized by
-      the copyright owner that is granting the License.
-
-      "Legal Entity" shall mean the union of the acting entity and all
-      other entities that control, are controlled by, or are under common
-      control with that entity. For the purposes of this definition,
-      "control" means (i) the power, direct or indirect, to cause the
-      direction or management of such entity, whether by contract or
-      otherwise, or (ii) ownership of fifty percent (50%) or more of the
-      outstanding shares, or (iii) beneficial ownership of such entity.
-
-      "You" (or "Your") shall mean an individual or Legal Entity
-      exercising permissions granted by this License.
-
-      "Source" form shall mean the preferred form for making modifications,
-      including but not limited to software source code, documentation
-      source, and configuration files.
-
-      "Object" form shall mean any form resulting from mechanical
-      transformation or translation of a Source form, including but
-      not limited to compiled object code, generated documentation,
-      and conversions to other media types.
-
-      "Work" shall mean the work of authorship, whether in Source or
-      Object form, made available under the License, as indicated by a
-      copyright notice that is included in or attached to the work
-      (an example is provided in the Appendix below).
-
-      "Derivative Works" shall mean any work, whether in Source or Object
-      form, that is based on (or derived from) the Work and for which the
-      editorial revisions, annotations, elaborations, or other modifications
-      represent, as a whole, an original work of authorship. For the purposes
-      of this License, Derivative Works shall not include works that remain
-      separable from, or merely link (or bind by name) to the interfaces of,
-      the Work and Derivative Works thereof.
-
-      "Contribution" shall mean any work of authorship, including
-      the original version of the Work and any modifications or additions
-      to that Work or Derivative Works thereof, that is intentionally
-      submitted to Licensor for inclusion in the Work by the copyright owner
-      or by an individual or Legal Entity authorized to submit on behalf of
-      the copyright owner. For the purposes of this definition, "submitted"
-      means any form of electronic, verbal, or written communication sent
-      to the Licensor or its representatives, including but not limited to
-      communication on electronic mailing lists, source code control systems,
-      and issue tracking systems that are managed by, or on behalf of, the
-      Licensor for the purpose of discussing and improving the Work, but
-      excluding communication that is conspicuously marked or otherwise
-      designated in writing by the copyright owner as "Not a Contribution."
-
-      "Contributor" shall mean Licensor and any individual or Legal Entity
-      on behalf of whom a Contribution has been received by Licensor and
-      subsequently incorporated within the Work.
-
-   2. Grant of Copyright License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      copyright license to reproduce, prepare Derivative Works of,
-      publicly display, publicly perform, sublicense, and distribute the
-      Work and such Derivative Works in Source or Object form.
-
-   3. Grant of Patent License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      (except as stated in this section) patent license to make, have made,
-      use, offer to sell, sell, import, and otherwise transfer the Work,
-      where such license applies only to those patent claims licensable
-      by such Contributor that are necessarily infringed by their
-      Contribution(s) alone or by combination of their Contribution(s)
-      with the Work to which such Contribution(s) was submitted. If You
-      institute patent litigation against any entity (including a
-      cross-claim or counterclaim in a lawsuit) alleging that the Work
-      or a Contribution incorporated within the Work constitutes direct
-      or contributory patent infringement, then any patent licenses
-      granted to You under this License for that Work shall terminate
-      as of the date such litigation is filed.
-
-   4. Redistribution. You may reproduce and distribute copies of the
-      Work or Derivative Works thereof in any medium, with or without
-      modifications, and in Source or Object form, provided that You
-      meet the following conditions:
-
-      (a) You must give any other recipients of the Work or
-          Derivative Works a copy of this License; and
-
-      (b) You must cause any modified files to carry prominent notices
-          stating that You changed the files; and
-
-      (c) You must retain, in the Source form of any Derivative Works
-          that You distribute, all copyright, patent, trademark, and
-          attribution notices from the Source form of the Work,
-          excluding those notices that do not pertain to any part of
-          the Derivative Works; and
-
-      (d) If the Work includes a "NOTICE" text file as part of its
-          distribution, then any Derivative Works that You distribute must
-          include a readable copy of the attribution notices contained
-          within such NOTICE file, excluding those notices that do not
-          pertain to any part of the Derivative Works, in at least one
-          of the following places: within a NOTICE text file distributed
-          as part of the Derivative Works; within the Source form or
-          documentation, if provided along with the Derivative Works; or,
-          within a display generated by the Derivative Works, if and
-          wherever such third-party notices normally appear. The contents
-          of the NOTICE file are for informational purposes only and
-          do not modify the License. You may add Your own attribution
-          notices within Derivative Works that You distribute, alongside
-          or as an addendum to the NOTICE text from the Work, provided
-          that such additional attribution notices cannot be construed
-          as modifying the License.
-
-      You may add Your own copyright statement to Your modifications and
-      may provide additional or different license terms and conditions
-      for use, reproduction, or distribution of Your modifications, or
-      for any such Derivative Works as a whole, provided Your use,
-      reproduction, and distribution of the Work otherwise complies with
-      the conditions stated in this License.
-
-   5. Submission of Contributions. Unless You explicitly state otherwise,
-      any Contribution intentionally submitted for inclusion in the Work
-      by You to the Licensor shall be under the terms and conditions of
-      this License, without any additional terms or conditions.
-      Notwithstanding the above, nothing herein shall supersede or modify
-      the terms of any separate license agreement you may have executed
-      with Licensor regarding such Contributions.
-
-   6. Trademarks. This License does not grant permission to use the trade
-      names, trademarks, service marks, or product names of the Licensor,
-      except as required for reasonable and customary use in describing the
-      origin of the Work and reproducing the content of the NOTICE file.
-
-   7. Disclaimer of Warranty. Unless required by applicable law or
-      agreed to in writing, Licensor provides the Work (and each
-      Contributor provides its Contributions) on an "AS IS" BASIS,
-      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
-      implied, including, without limitation, any warranties or conditions
-      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
-      PARTICULAR PURPOSE. You are solely responsible for determining the
-      appropriateness of using or redistributing the Work and assume any
-      risks associated with Your exercise of permissions under this License.
-
-   8. Limitation of Liability. In no event and under no legal theory,
-      whether in tort (including negligence), contract, or otherwise,
-      unless required by applicable law (such as deliberate and grossly
-      negligent acts) or agreed to in writing, shall any Contributor be
-      liable to You for damages, including any direct, indirect, special,
-      incidental, or consequential damages of any character arising as a
-      result of this License or out of the use or inability to use the
-      Work (including but not limited to damages for loss of goodwill,
-      work stoppage, computer failure or malfunction, or any and all
-      other commercial damages or losses), even if such Contributor
-      has been advised of the possibility of such damages.
-
-   9. Accepting Warranty or Additional Liability. While redistributing
-      the Work or Derivative Works thereof, You may choose to offer,
-      and charge a fee for, acceptance of support, warranty, indemnity,
-      or other liability obligations and/or rights consistent with this
-      License. However, in accepting such obligations, You may act only
-      on Your own behalf and on Your sole responsibility, not on behalf
-      of any other Contributor, and only if You agree to indemnify,
-      defend, and hold each Contributor harmless for any liability
-      incurred by, or claims asserted against, such Contributor by reason
-      of your accepting any such warranty or additional liability.
-
-   END OF TERMS AND CONDITIONS
-
-   APPENDIX: How to apply the Apache License to your work.
-
-      To apply the Apache License to your work, attach the following
-      boilerplate notice, with the fields enclosed by brackets "[]"
-      replaced with your own identifying information. (Don't include
-      the brackets!)  The text should be enclosed in the appropriate
-      comment syntax for the file format. We also recommend that a
-      file or class name and description of purpose be included on the
-      same "printed page" as the copyright notice for easier
-      identification within third-party archives.
-
-   Copyright [yyyy] [name of copyright owner]
-
-   Licensed under the Apache License, Version 2.0 (the "License");
-   you may not use this file except in compliance with the License.
-   You may obtain a copy of the License at
-
-       http://www.apache.org/licenses/LICENSE-2.0
-
-   Unless required by applicable law or agreed to in writing, software
-   distributed under the License is distributed on an "AS IS" BASIS,
-   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-   See the License for the specific language governing permissions and
-   limitations under the License.

code/components/esp-nn/README.md

@@ -1,54 +0,0 @@
-# ESP-NN
-
-The library contains optimised NN (Neural Network) functions for various Espressif chipsets.
-
-* Supported platforms:
-   * TensorFlow Lite Micro (TFLite Micro). Repo can be found [here](https://github.com/espressif/tflite-micro-esp-examples)
-
-* Supported ESP chipsets include:
-   * ESP32-S3 (Assembly versions optimised to benefit from vector instructions of ESP32-S3)
-   * ESP32 (ANSI C versions)
-
-## Performance
-
-### Kernelwise performance for s8 versions:
-
-  * Kernelwise performance on ESP32-S3 chip
-    * Numbers are ticks taken for kernel to execute
-    * Chip config: 240MHz, SPI: QPI 80MHz, Data cache: 64KB
-
-    | Function        | ANSI C  | ESP32-S3 Opt  | Opt Ratio | Data info   | Memory    |
-    | ----------------| --------|---------|---------|-------------|-----------|
-    | elementwise_add | 320397  | 87119   | 3.68    | size = 1615 | External  |
-    | elementwise_mul | 125958  | 44239   | 2.85    | size = 1615 | External  |
-    | convolution     | 4663012 | 428675  | 10.88   | input(10,10), filter(64x1x1x64) | External |
-    | convolution     | 301014  | 32433   | 9.28    | input(8,8), filter(16x1x1x16) | External |
-    | convolution     | 2115418 | 1020923 | 2.07    | input(10,10), filter(64x3x3x3) | External |
-    | depthwise conv  | 1190062 | 203278  | 5.85    | input (18, 18), pad(0,0), stride(1,1) filter: 1x3x3x16 | External |
-    | depthwise conv  | 837072  | 182335  | 4.59    | input (12, 12), pad(1,1), stride(1,1)  filter: 8x5x5x4 | External |
-    | max pool        | 485714  | 76747   | 6.33    | input(16,16), filter (1x3x3x16) | Internal |
-    | avg pool        | 541462  | 160580  | 3.37    | input(16,16), filter (1x3x3x16) | Internal |
-    | fully connected | 15853   | 9547    | 1.66    | len: 265, ch = 3 | Internal |
-    | prelu (relu6)   | 19472   | 2734    | 7.12    | size, 1615  | Internal  |
-
-
-## Configuration
-
-  * To configure, please use `idf.py menuconfig` and under `ESP-NN` select `NN_OPTIMIZATIONS`
-  * There are two options presented:
-     * Optimized versions
-     * ANSI C
-
-  * Default selection is for `Optimized versions`. For ESP32-S3, assembly versions are automatically selected, whereas for ESP32,  ANSI-C versions are selected by default.
-  * For debugging purposes, you may want to select `ANSI C`
-
-
-## Contributing
-
-If you encounter an issue with ESP-NN, or wish to submit a feature request, please use the Issues section on the Github.
-
-For general questions related to this library, please use the esp32.com forum.
-
-## Copyrights and License
-
-All original source code in this repository is Copyright (C) 2020-2021 Espressif Systems. This source code is licensed under the Apache License 2.0 as described in the file LICENSE.

code/components/esp-nn/include/esp_nn_ansi_c.h

@@ -1,46 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-/**
- * @file        Header definitions to include for ANSI C versions.
- *              These are just typedefs to pick up ANSI versions.
- */
-
-#pragma once
-
-#include "esp_nn_ansi_headers.h"
-
-#define esp_nn_add_elementwise_s8 esp_nn_add_elementwise_s8_ansi
-#define esp_nn_mul_elementwise_s8 esp_nn_mul_elementwise_s8_ansi
-
-#define esp_nn_depthwise_conv_s8 esp_nn_depthwise_conv_s8_ansi
-
-#define esp_nn_conv_s8 esp_nn_conv_s8_ansi
-
-#define esp_nn_get_conv_scratch_size esp_nn_get_conv_scratch_size_ansi
-#define esp_nn_set_conv_scratch_buf esp_nn_set_conv_scratch_buf_ansi
-
-#define esp_nn_get_depthwise_conv_scratch_size esp_nn_get_depthwise_conv_scratch_size_ansi
-#define esp_nn_set_depthwise_conv_scratch_buf esp_nn_set_depthwise_conv_scratch_buf_ansi
-
-#define esp_nn_relu6_s8 esp_nn_relu6_s8_ansi
-
-#define esp_nn_avg_pool_s8 esp_nn_avg_pool_s8_ansi
-#define esp_nn_max_pool_s8 esp_nn_max_pool_s8_ansi
-
-#define esp_nn_fully_connected_s8 esp_nn_fully_connected_s8_ansi
-
-#define esp_nn_get_softmax_scratch_size esp_nn_get_softmax_scratch_size_ansi
-#define esp_nn_set_softmax_scratch_buf esp_nn_set_softmax_scratch_buf_ansi
-#define esp_nn_softmax_s8 esp_nn_softmax_s8_ansi

code/components/esp-nn/include/esp_nn_ansi_headers.h

@@ -1,283 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#pragma once
-
-/**
- * @file        Header definitions to include for esp_nn reference functions
- */
-
-#include <stdint.h>
-
-/************************** Basic math functions ****************************/
-
-/**
- * @brief       elementwise addition
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- *
- *              shift values are expected to be <= 0
- */
-void esp_nn_add_elementwise_s8_ansi(const int8_t *input1_data,
-                                    const int8_t *input2_data,
-                                    const int32_t input1_offset,
-                                    const int32_t input2_offset,
-                                    const int32_t input1_mult,
-                                    const int32_t input2_mult,
-                                    const int32_t input1_shift,
-                                    const int32_t input2_shift,
-                                    const int32_t left_shift,
-                                    int8_t *output,
-                                    const int32_t out_offset,
-                                    const int32_t out_mult,
-                                    const int32_t out_shift,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max,
-                                    const int32_t size);
-/**
- * @brief       elementwise multiplication
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- *
- *              output shift is expected to be <= 0
- */
-void esp_nn_mul_elementwise_s8_ansi(const int8_t *input1_data,
-                                    const int8_t *input2_data,
-                                    const int32_t input1_offset,
-                                    const int32_t input2_offset,
-                                    int8_t *output,
-                                    const int32_t out_offset,
-                                    const int32_t out_mult,
-                                    const int32_t out_shift,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max,
-                                    const int32_t size);
-
-
-/************************** Convolution functions *****************************/
-
-/**
- * @brief       depthwise convolution per channel
- *
- * @note        inputs type: int8_t, output: int8_t
- *              Version used in tflite is per channel.
- *              This version follows the same footsprints.
- *              Meaning, it has per out_channel shift and multiplier for
- *              requantization
- *
- *              optimization notes: Though input_offset is int32 type,
- *              offset values are contained in 8 bits [-128, 127]
- */
-void esp_nn_depthwise_conv_s8_ansi(const int8_t *input_data,
-                                   const uint16_t input_wd,
-                                   const uint16_t input_ht,
-                                   const uint16_t channels,
-                                   const int32_t input_offset,
-                                   const uint16_t pad_wd,
-                                   const uint16_t pad_ht,
-                                   const uint16_t stride_wd,
-                                   const uint16_t stride_ht,
-                                   const uint16_t ch_mult,
-                                   const int8_t *filter_data,
-                                   const uint16_t filter_wd,
-                                   const uint16_t filter_ht,
-                                   const int32_t *bias,
-                                   int8_t *out_data,
-                                   const uint16_t out_wd,
-                                   const uint16_t out_ht,
-                                   const int32_t out_offset,
-                                   const int32_t *out_shift,
-                                   const int32_t *out_mult,
-                                   const int32_t activation_min,
-                                   const int32_t activation_max);
-
-/**
- * @brief       2d-convolution channelwise
- *
- * @note        operation: result += (input + offset) * filter
- *
- *              inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_conv_s8_ansi(const int8_t *input_data,
-                         const uint16_t input_wd,
-                         const uint16_t input_ht,
-                         const uint16_t in_channels,
-                         const int32_t input_offset,
-                         const uint16_t pad_wd,
-                         const uint16_t pad_ht,
-                         const uint16_t stride_wd,
-                         const uint16_t stride_ht,
-                         const int8_t *filter_data,
-                         const uint16_t filter_wd,
-                         const uint16_t filter_ht,
-                         const int32_t *bias,
-                         int8_t *out_data,
-                         const uint16_t out_wd,
-                         const uint16_t out_ht,
-                         const uint16_t out_channels,
-                         const int32_t out_offset,
-                         const int32_t *out_shift,
-                         const int32_t *out_mult,
-                         const int32_t activation_min,
-                         const int32_t activation_max);
-
-int esp_nn_get_conv_scratch_size_ansi(const uint16_t input_wd,
-                                      const uint16_t input_ht,
-                                      const uint16_t in_ch,
-                                      const uint16_t out_ch,
-                                      const uint16_t filter_wd,
-                                      const uint16_t filter_ht);
-void esp_nn_set_conv_scratch_buf_ansi(const void *buf);
-
-int esp_nn_get_depthwise_conv_scratch_size_ansi(const uint16_t input_wd,
-                                                const uint16_t input_ht,
-                                                const uint16_t channels,
-                                                const uint16_t ch_mult,
-                                                const uint16_t filter_wd,
-                                                const uint16_t filter_ht);
-void esp_nn_set_depthwise_conv_scratch_buf_ansi(const void *buf);
-
-/************************** Activation functions *****************************/
-
-/**
- * @brief       relu6
- *
- * @note        inout: int8_t
- */
-void esp_nn_relu6_s8_ansi(int8_t *data, uint16_t size);
-
-/************************** Pooling functions *****************************/
-
-
-/**
- * @brief       max_pool
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_max_pool_s8_ansi(const int8_t *input,
-                             const uint16_t input_wd,
-                             const uint16_t input_ht,
-                             int8_t *output,
-                             const uint16_t output_wd,
-                             const uint16_t output_ht,
-                             const uint16_t stride_wd,
-                             const uint16_t stride_ht,
-                             const uint16_t filter_wd,
-                             const uint16_t filter_ht,
-                             const uint16_t pad_wd,
-                             const uint16_t pad_ht,
-                             const int32_t activation_min,
-                             const int32_t activation_max,
-                             const uint16_t channels);
-
-/**
- * @brief       avg_pool
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_avg_pool_s8_ansi(const int8_t *input,
-                             const uint16_t input_wd,
-                             const uint16_t input_ht,
-                             int8_t *output,
-                             const uint16_t output_wd,
-                             const uint16_t output_ht,
-                             const uint16_t stride_wd,
-                             const uint16_t stride_ht,
-                             const uint16_t filter_wd,
-                             const uint16_t filter_ht,
-                             const uint16_t pad_wd,
-                             const uint16_t pad_ht,
-                             const int32_t activation_min,
-                             const int32_t activation_max,
-                             const uint16_t channels);
-
-
-/************************** Fully connected functions ***********************/
-
-/**
- * @brief       fully connected
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_fully_connected_s8_ansi(const int8_t *input_data,
-                                    const int32_t input_offset,
-                                    const uint16_t row_len,
-                                    const int8_t *filter_data,
-                                    const int32_t filter_offset,
-                                    const int32_t *bias,
-                                    int8_t *out_data,
-                                    const uint16_t out_channels,
-                                    const int32_t out_offset,
-                                    const int32_t out_shift,
-                                    const int32_t out_mult,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max);
-
-/**
- * @brief   Get scratch buffer size needed by softmax function
- *
- * @param   width
- * @param   height
- * @return  size in bytes
- *
- * @note    buffer must be 4 byte aligned
- */
-int32_t esp_nn_get_softmax_scratch_size_ansi(const int32_t width, const int32_t height);
-
-/* ANSI C function to be hooked up when optimised version needed */
-int32_t esp_nn_get_softmax_scratch_size_opt(const int32_t width, const int32_t height);
-
-/**
- * @brief   Set scratch buffer to be used by softmax function
- *
- * @param   buffer  this can be NULL if one needs to unset it
- *                  must be aligned to 4 bytes
- */
-void esp_nn_set_softmax_scratch_buf_ansi(void *buffer);
-
-/* ANSI C function to be hooked up when optimised version needed */
-void esp_nn_set_softmax_scratch_buf_opt(void *buffer);
-
-/**
- * @brief       reference softmax function
- *
- * @note        inputs type: int8_t, output: int8_t
- */
-void esp_nn_softmax_s8_ansi(const int8_t *input_data,
-                            const int32_t height,
-                            const int32_t width,
-                            const int32_t mult,
-                            const int32_t shift,
-                            const int32_t diff_min,
-                            int8_t *output_data);
-
-/**
- * @brief       optimised version of softmax function
- *
- * @note        the function uses extra buffer (4 * width bytes)
- *              hence, scratch buffers must be set before calling this.
- */
-void esp_nn_softmax_s8_opt(const int8_t *input_data,
-                           const int32_t height,
-                           const int32_t width,
-                           const int32_t mult,
-                           const int32_t shift,
-                           const int32_t diff_min,
-                           int8_t *output_data);

code/components/esp-nn/include/esp_nn_esp32.h

@@ -1,48 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-/**
- * @file        Header definitions to include for esp_nn optimized functions for
- *              the ESP32 platform.
- *              We are hooking up just the C versions for now.
- *              The file hence is exactly same as `esp_nn_ansi_c.h`
- */
-
-#pragma once
-
-#include "esp_nn_ansi_headers.h"
-
-#define esp_nn_add_elementwise_s8 esp_nn_add_elementwise_s8_ansi
-#define esp_nn_mul_elementwise_s8 esp_nn_mul_elementwise_s8_ansi
-
-#define esp_nn_depthwise_conv_s8 esp_nn_depthwise_conv_s8_ansi
-
-#define esp_nn_conv_s8 esp_nn_conv_s8_ansi
-
-#define esp_nn_get_conv_scratch_size esp_nn_get_conv_scratch_size_ansi
-#define esp_nn_set_conv_scratch_buf esp_nn_set_conv_scratch_buf_ansi
-
-#define esp_nn_get_depthwise_conv_scratch_size esp_nn_get_depthwise_conv_scratch_size_ansi
-#define esp_nn_set_depthwise_conv_scratch_buf esp_nn_set_depthwise_conv_scratch_buf_ansi
-
-#define esp_nn_relu6_s8 esp_nn_relu6_s8_ansi
-
-#define esp_nn_avg_pool_s8 esp_nn_avg_pool_s8_ansi
-#define esp_nn_max_pool_s8 esp_nn_max_pool_s8_ansi
-
-#define esp_nn_fully_connected_s8 esp_nn_fully_connected_s8_ansi
-
-#define esp_nn_get_softmax_scratch_size esp_nn_get_softmax_scratch_size_opt
-#define esp_nn_set_softmax_scratch_buf esp_nn_set_softmax_scratch_buf_opt
-#define esp_nn_softmax_s8 esp_nn_softmax_s8_opt

code/components/esp-nn/include/esp_nn_esp32s3.h

@@ -1,261 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-/**
- * @file        Header definitions to include for esp_nn optimized functions for
- *              the ESP32-S3 platform
- */
-
-#pragma once
-
-#include <stdint.h>
-#include "esp_nn_ansi_headers.h"
-
-/************************** Basic math functions *****************************/
-
-
-/**
- * @brief       elementwise addition
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- *
- *              shift values are expected to be <= 0
- */
-void esp_nn_add_elementwise_s8_esp32s3(const int8_t *input1_data,
-                                       const int8_t *input2_data,
-                                       const int32_t input1_offset,
-                                       const int32_t input2_offset,
-                                       const int32_t input1_mult,
-                                       const int32_t input2_mult,
-                                       const int32_t input1_shift,
-                                       const int32_t input2_shift,
-                                       const int32_t left_shift,
-                                       int8_t *output,
-                                       const int32_t out_offset,
-                                       const int32_t out_mult,
-                                       const int32_t out_shift,
-                                       const int32_t activation_min,
-                                       const int32_t activation_max,
-                                       const int32_t size);
-
-/**
- * @brief       elementwise multiplication
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- *
- *              output shift is expected to be <= 0
- */
-void esp_nn_mul_elementwise_s8_esp32s3(const int8_t *input1_data,
-                                       const int8_t *input2_data,
-                                       const int32_t input1_offset,
-                                       const int32_t input2_offset,
-                                       int8_t *output,
-                                       const int32_t out_offset,
-                                       const int32_t out_mult,
-                                       const int32_t out_shift,
-                                       const int32_t activation_min,
-                                       const int32_t activation_max,
-                                       const int32_t size);
-
-
-/************************** Convolution functions *****************************/
-
-/**
- * @brief       depthwise convolution per channel
- *
- * @note        inputs type: int8_t, output: int8_t
- *              Version used in tflite is per channel.
- *              This version follows the same footsprints.
- *              Meaning, it has per out_channel shift and multiplier for
- *              requantization
- *
- *              optimization notes: Though input_offset is int32 type,
- *              offset values are contained in 8 bits [-128, 127]
- */
-void esp_nn_depthwise_conv_s8_esp32s3(const int8_t *input_data,
-                                      const uint16_t input_wd,
-                                      const uint16_t input_ht,
-                                      const uint16_t channels,
-                                      const int32_t input_offset,
-                                      const uint16_t pad_wd,
-                                      const uint16_t pad_ht,
-                                      const uint16_t stride_wd,
-                                      const uint16_t stride_ht,
-                                      const uint16_t ch_mult,
-                                      const int8_t *filter_data,
-                                      const uint16_t filter_wd,
-                                      const uint16_t filter_ht,
-                                      const int32_t *bias,
-                                      int8_t *out_data,
-                                      const uint16_t out_wd,
-                                      const uint16_t out_ht,
-                                      const int32_t out_offset,
-                                      const int32_t *out_shift,
-                                      const int32_t *out_mult,
-                                      const int32_t activation_min,
-                                      const int32_t activation_max);
-
-/**
- * @brief       2d - convolution channelwise
- *
- * @note        operation: result += (input + offset) * filter
- *
- *              inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_conv_s8_esp32s3(const int8_t *input_data,
-                            const uint16_t input_wd,
-                            const uint16_t input_ht,
-                            const uint16_t in_channels,
-                            const int32_t input_offset,
-                            const uint16_t pad_wd,
-                            const uint16_t pad_ht,
-                            const uint16_t stride_wd,
-                            const uint16_t stride_ht,
-                            const int8_t *filter_data,
-                            const uint16_t filter_wd,
-                            const uint16_t filter_ht,
-                            const int32_t *bias,
-                            int8_t *out_data,
-                            const uint16_t out_wd,
-                            const uint16_t out_ht,
-                            const uint16_t out_channels,
-                            const int32_t out_offset,
-                            const int32_t *out_shift,
-                            const int32_t *out_mult,
-                            const int32_t activation_min,
-                            const int32_t activation_max);
-
-int esp_nn_get_conv_scratch_size_esp32s3(const uint16_t input_wd,
-                                         const uint16_t input_ht,
-                                         const uint16_t in_ch,
-                                         const uint16_t out_ch,
-                                         const uint16_t filter_wd,
-                                         const uint16_t filter_ht);
-void esp_nn_set_conv_scratch_buf_esp32s3(const void *buf);
-
-int esp_nn_get_depthwise_conv_scratch_size_esp32s3(const uint16_t input_wd,
-                                                   const uint16_t input_ht,
-                                                   const uint16_t channels,
-                                                   const uint16_t ch_mult,
-                                                   const uint16_t filter_wd,
-                                                   const uint16_t filter_ht);
-void esp_nn_set_depthwise_conv_scratch_buf_esp32s3(const void *buf);
-
-/************************** Pooling functions *****************************/
-
-/**
- * @brief       max_pool
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_max_pool_s8_esp32s3(const int8_t *input,
-                                const uint16_t input_wd,
-                                const uint16_t input_ht,
-                                int8_t *output,
-                                const uint16_t output_wd,
-                                const uint16_t output_ht,
-                                const uint16_t stride_wd,
-                                const uint16_t stride_ht,
-                                const uint16_t filter_wd,
-                                const uint16_t filter_ht,
-                                const uint16_t pad_wd,
-                                const uint16_t pad_ht,
-                                const int32_t activation_min,
-                                const int32_t activation_max,
-                                const uint16_t channels);
-
-/**
- * @brief       avg_pool
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- */
-void esp_nn_avg_pool_s8_esp32s3(const int8_t *input,
-                                const uint16_t input_wd,
-                                const uint16_t input_ht,
-                                int8_t *output,
-                                const uint16_t output_wd,
-                                const uint16_t output_ht,
-                                const uint16_t stride_wd,
-                                const uint16_t stride_ht,
-                                const uint16_t filter_wd,
-                                const uint16_t filter_ht,
-                                const uint16_t pad_wd,
-                                const uint16_t pad_ht,
-                                const int32_t activation_min,
-                                const int32_t activation_max,
-                                const uint16_t channels);
-
-
-/************************** Fully connected functions *****************************/
-
-/**
- * @brief       fully connected
- *
- * @note        inputs type: int8_t, output: int8_t
- *              input offsets: although int32_t, they are contained in 8 bits [-128, 127]
- *
- *              Current version works only on aligned input.
- *              row_len and channels should both be multiple of 8.
- */
-void esp_nn_fully_connected_s8_esp32s3(const int8_t *input_data,
-                                       const int32_t input_offset,
-                                       const uint16_t row_len,
-                                       const int8_t *filter_data,
-                                       const int32_t filter_offset,
-                                       const int32_t *bias,
-                                       int8_t *out_data,
-                                       const uint16_t out_channels,
-                                       const int32_t out_offset,
-                                       const int32_t out_shift,
-                                       const int32_t out_mult,
-                                       const int32_t activation_min,
-                                       const int32_t activation_max);
-
-/**
- * @brief       relu6
- *
- * @note        inout: int8_t
- */
-void esp_nn_relu6_s8_esp32s3(int8_t *data, uint16_t size);
-
-/********************** function defines ***************************/
-
-#define esp_nn_add_elementwise_s8 esp_nn_add_elementwise_s8_esp32s3
-#define esp_nn_mul_elementwise_s8 esp_nn_mul_elementwise_s8_esp32s3
-
-#define esp_nn_depthwise_conv_s8 esp_nn_depthwise_conv_s8_esp32s3
-
-#define esp_nn_get_conv_scratch_size esp_nn_get_conv_scratch_size_esp32s3
-#define esp_nn_set_conv_scratch_buf esp_nn_set_conv_scratch_buf_esp32s3
-
-#define esp_nn_get_depthwise_conv_scratch_size esp_nn_get_depthwise_conv_scratch_size_esp32s3
-#define esp_nn_set_depthwise_conv_scratch_buf esp_nn_set_depthwise_conv_scratch_buf_esp32s3
-
-#define esp_nn_conv_s8 esp_nn_conv_s8_esp32s3
-
-#define esp_nn_relu6_s8 esp_nn_relu6_s8_esp32s3
-
-#define esp_nn_avg_pool_s8 esp_nn_avg_pool_s8_esp32s3
-#define esp_nn_max_pool_s8 esp_nn_max_pool_s8_esp32s3
-
-#define esp_nn_fully_connected_s8 esp_nn_fully_connected_s8_esp32s3
-
-#define esp_nn_get_softmax_scratch_size esp_nn_get_softmax_scratch_size_opt
-#define esp_nn_set_softmax_scratch_buf esp_nn_set_softmax_scratch_buf_opt
-#define esp_nn_softmax_s8 esp_nn_softmax_s8_opt

code/components/esp-nn/src/basic_math/esp_nn_add_ansi.c

@@ -1,97 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-void esp_nn_add_elementwise_u8_ansi(const uint8_t *input1_data,
-                                    const uint8_t *input2_data,
-                                    const int32_t input1_offset,
-                                    const int32_t input2_offset,
-                                    const int32_t input1_mult,
-                                    const int32_t input2_mult,
-                                    const int32_t input1_shift,
-                                    const int32_t input2_shift,
-                                    const int32_t left_shift,
-                                    uint8_t *output,
-                                    const int32_t out_offset,
-                                    const int32_t out_mult,
-                                    const int32_t out_shift,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max,
-                                    const int32_t size)
-{
-    for (int i = 0; i < size; i++) {
-        int32_t tmp1 = input1_data[i] + input1_offset;
-        int32_t tmp2 = input2_data[i] + input2_offset;
-
-        tmp1 <<= left_shift;
-        tmp2 <<= left_shift;
-
-        tmp1 = esp_nn_sat_round_doubling_high_mul(tmp1, input1_mult);
-        tmp2 = esp_nn_sat_round_doubling_high_mul(tmp2, input2_mult);
-
-        tmp1 = esp_nn_div_by_power_of_two(tmp1, -input1_shift);
-        tmp2 = esp_nn_div_by_power_of_two(tmp2, -input2_shift);
-
-        int32_t out = tmp1 + tmp2;
-        out = esp_nn_sat_round_doubling_high_mul(out, out_mult);
-        out = esp_nn_div_by_power_of_two(out, -out_shift);
-        out = out + out_offset;
-
-        out = max(activation_min, min(out, activation_max));
-        output[i] = (uint8_t) out;
-    }
-}
-
-void esp_nn_add_elementwise_s8_ansi(const int8_t *input1_data,
-                                    const int8_t *input2_data,
-                                    const int32_t input1_offset,
-                                    const int32_t input2_offset,
-                                    const int32_t input1_mult,
-                                    const int32_t input2_mult,
-                                    const int32_t input1_shift,
-                                    const int32_t input2_shift,
-                                    const int32_t left_shift,
-                                    int8_t *output,
-                                    const int32_t out_offset,
-                                    const int32_t out_mult,
-                                    const int32_t out_shift,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max,
-                                    const int32_t size)
-{
-    for (int i = 0; i < size; i++) {
-        int32_t tmp1 = input1_data[i] + input1_offset;
-        int32_t tmp2 = input2_data[i] + input2_offset;
-
-        tmp1 <<= left_shift;
-        tmp2 <<= left_shift;
-
-        tmp1 = esp_nn_sat_round_doubling_high_mul(tmp1, input1_mult);
-        tmp2 = esp_nn_sat_round_doubling_high_mul(tmp2, input2_mult);
-
-        tmp1 = esp_nn_div_by_power_of_two(tmp1, -input1_shift);
-        tmp2 = esp_nn_div_by_power_of_two(tmp2, -input2_shift);
-
-        int32_t out = tmp1 + tmp2;
-        out = esp_nn_sat_round_doubling_high_mul(out, out_mult);
-        out = esp_nn_div_by_power_of_two(out, -out_shift);
-        out = out + out_offset;
-
-        out = max(activation_min, min(out, activation_max));
-        output[i] = (int8_t) out;
-    }
-}

code/components/esp-nn/src/basic_math/esp_nn_mul_ansi.c

@@ -1,42 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-void esp_nn_mul_elementwise_s8_ansi(const int8_t *input1_data,
-                                    const int8_t *input2_data,
-                                    const int32_t input1_offset,
-                                    const int32_t input2_offset,
-                                    int8_t *output,
-                                    const int32_t out_offset,
-                                    const int32_t out_mult,
-                                    const int32_t out_shift,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max,
-                                    const int32_t size)
-{
-    for (int i = 0; i < size; i++) {
-        int32_t tmp1 = input1_data[i] + input1_offset;
-        int32_t tmp2 = input2_data[i] + input2_offset;
-
-        int32_t out = tmp1 * tmp2;
-        out = esp_nn_multiply_by_quantized_mult(out, out_mult, out_shift);
-        out = out + out_offset;
-
-        out = max(activation_min, min(out, activation_max));
-        output[i] = (int8_t) out;
-    }
-}

code/components/esp-nn/src/common/common_functions.h

@@ -1,218 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#pragma once
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <string.h>
-
-/**
- * c99 standard still doesn't strictly inline functions
- * We need to use attribute as well to do this.
- */
-#define __NN_FORCE_INLINE__ __attribute((always_inline)) static inline
-
-/* min/max macros */
-#ifndef max
-#define max(a, b) ({            \
-    __typeof__ (a) _a = (a);    \
-    __typeof__ (b) _b = (b);    \
-    _a > _b ? _a : _b;          \
-})
-
-#define min(a, b) ({            \
-    __typeof__ (a) _a = (a);    \
-    __typeof__ (b) _b = (b);    \
-    _a < _b ? _a : _b;          \
-})
-#endif
-
-__NN_FORCE_INLINE__ int32_t esp_nn_clz32(uint32_t in)
-{
-    __asm__ volatile("nsau %0, %0" : "+r" (in));
-    return in;
-}
-
-__NN_FORCE_INLINE__ int32_t esp_nn_pick_sat_high32_of64(int64_t val64)
-{
-    int32_t sign = (int32_t) (val64 >> 63);
-    int32_t to_add = sign & ((1ul << 31) - 1);
-    return (int32_t) ((int64_t) (val64 + to_add) >> 31);
-}
-
-/**
- * Signed saturate a 32 bit value to 8 bits keeping output in 32 bit variable.
- */
-__NN_FORCE_INLINE__ int32_t esp_nn_saturate8(int32_t in)
-{
-    __asm__ volatile("clamps %0, %0, 7" : "+a"(in));
-    return in;
-}
-
-__NN_FORCE_INLINE__ int32_t esp_nn_sat_round_doubling_high_mul(int32_t in0, int32_t in1)
-{
-    int32_t result;
-    int64_t in0_64 = (int64_t) in0;
-    bool overflow = (in0 == in1) && (in0 == (int32_t) INT32_MIN);
-
-    /* Nudge value */
-    int64_t nudge_val = 1 << 30;
-    if ((in0 < 0) ^ (in1 < 0)) {
-        nudge_val = 1 - nudge_val;
-    }
-
-    /* Multiply and add nudge */
-    int64_t mult = in0_64 * in1 + nudge_val;
-
-    /* Round and pickup 32 bits */
-    result = esp_nn_pick_sat_high32_of64(mult);
-
-    return overflow ? INT32_MAX : result;
-}
-
-/**
- * fast version
- * this will fail for values closer to INT32_MAX and INT32_MIN by `1 << (exponent - 1)`.
- * We can afford to do this because we are at the very last stage of filter.
- * Also it is pretty rare condition as our output is going to be 8 bit.
- */
-__NN_FORCE_INLINE__ int32_t esp_nn_div_by_power_of_two_fast(int32_t val, int32_t exponent)
-{
-    int32_t to_add = (1 << (exponent - 1)) - (val < 0);
-    return (int32_t) ((val + to_add) >> exponent);
-}
-
-__NN_FORCE_INLINE__ int32_t esp_nn_div_by_power_of_two(int32_t val, int32_t exponent)
-{
-    int32_t result;
-
-    const int32_t mask = (1 << exponent) - 1;
-    const int32_t remainder = val & mask;
-
-    result = val >> exponent;
-    int32_t threshold = (mask >> 1) + (result < 0);
-
-    if (remainder > threshold) {
-        result += 1;
-    }
-    return result;
-}
-
-__NN_FORCE_INLINE__ int32_t esp_nn_multiply_by_quantized_mult(int32_t x, int32_t mult, int32_t shift)
-{
-    int32_t left_shift = shift > 0 ? shift : 0;
-    int32_t right_shift = shift > 0 ? 0 : -shift;
-    int32_t result = esp_nn_sat_round_doubling_high_mul(x * (1 << left_shift), mult);
-    return esp_nn_div_by_power_of_two(result, right_shift);
-}
-
-__NN_FORCE_INLINE__ int32_t esp_nn_multiply_by_quantized_mult_fast(int32_t x, int32_t mult, int32_t shift)
-{
-    int32_t left_shift = max(shift, 0);
-    int32_t right_shift = left_shift - shift;
-
-    int64_t nudge_val = 1 << 30;
-    int64_t in0_64 = (int64_t) (x << left_shift);
-
-    /* Multiply and add nudge */
-    int64_t mult_64 = in0_64 * mult + nudge_val;
-    int32_t result = (int32_t) (mult_64 >> 31);
-    if (right_shift) {
-        result = esp_nn_div_by_power_of_two_fast(result, right_shift);
-    }
-    return result;
-}
-
-static void esp_nn_aligned_s8_pad_with_value(const int8_t *src, int8_t *dst,
-                                             const uint16_t input_wd,
-                                             const uint16_t input_ht,
-                                             const uint16_t channels,
-                                             const int32_t pad_val,
-                                             const uint16_t pad_wd,
-                                             const uint16_t pad_ht)
-{
-    /* memset with pad_val */
-    memset(dst, pad_val, ((input_wd + 2 * pad_wd) * (input_ht + 2 * pad_ht)) * channels * 2);
-    dst += (pad_wd + input_wd + pad_wd) * channels;
-
-    for (int i = 0; i < input_ht; i++) {
-        dst += pad_wd * channels;
-        for (int j = 0; j < input_wd * channels; j++) {
-            *dst++ = *src++;
-        }
-        dst += pad_wd * channels;
-    }
-}
-
-#if 0
-static void esp_nn_aligned_s8_pad_end_with_value(const int8_t *src, int8_t *dst,
-                                                 const uint16_t input_wd,
-                                                 const uint16_t input_ht,
-                                                 const uint16_t channels,
-                                                 const int32_t pad_val,
-                                                 const uint16_t pad_wd,
-                                                 const uint16_t pad_ht)
-{
-    for (int i = 0; i < input_ht; i++) {
-        for (int j = 0; j < input_wd * channels; j++) {
-            *dst++ = *src++;
-        }
-        memset(dst, pad_val, pad_wd * channels);
-        dst += pad_wd * channels;
-    }
-    /* pad end `pad_ht` lines at end */
-    memset(dst, pad_val, (input_wd + pad_wd) * pad_ht * channels);
-}
-#endif
-
-/**
- * @brief       convert 8 bit input data to 16 bit
- *
- * @param       src int8_t source data
- * @param       dst int16_t dst data
- * @param       size length of data
- * @param       offset  offset to be added to src data. Range: [-128, 127]
- */
-__NN_FORCE_INLINE__ void esp_nn_s8_to_s16_with_offset(const int8_t *src, int16_t *dst,
-                                                      const int size, const int32_t offset)
-{
-    int i = 0;
-    for (; i < size; i += 2) {
-        dst[i + 0] = src[i + 0] + offset;
-        dst[i + 1] = src[i + 1] + offset;
-    }
-    if(i < size) {
-        dst[i] = src[i] + offset;
-    }
-}
-
-/**
- * @brief       convert 8 bit input data to 16 bit
- *
- * @param       src int8_t source data
- * @param       dst int16_t dst data
- * @param       size length of data
- */
-__NN_FORCE_INLINE__ void esp_nn_s8_to_s16(const int8_t *src, int16_t *dst, const int size)
-{
-    int i = 0;
-    for (; i < size; i += 2) {
-        dst[i + 0] = src[i + 0];
-        dst[i + 1] = src[i + 1];
-    }
-    if(i < size) {
-        dst[i] = src[i];
-    }
-}

code/components/esp-nn/src/convolution/esp_nn_conv_ansi.c

@@ -1,175 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-int esp_nn_get_conv_scratch_size_ansi(const uint16_t input_wd,
-                                      const uint16_t input_ht,
-                                      const uint16_t in_ch,
-                                      const uint16_t out_ch,
-                                      const uint16_t filter_wd,
-                                      const uint16_t filter_ht)
-{
-    return 0;
-}
-
-void esp_nn_set_conv_scratch_buf_ansi(const void *buf)
-{
-
-}
-
-/**
- * Assumption 1: i/p channels == o/p channels
- * Assumption 2: Pointers are valid
- * Assumption 3: dialation width = 1
- */
-void esp_nn_conv_u8_ansi(const uint8_t *input_data,
-                         const uint16_t input_wd,
-                         const uint16_t input_ht,
-                         const uint16_t in_channels,
-                         const int32_t input_offset,
-                         const uint16_t pad_wd,
-                         const uint16_t pad_ht,
-                         const uint16_t stride_wd,
-                         const uint16_t stride_ht,
-                         const uint8_t *filter_data,
-                         const uint16_t filter_wd,
-                         const uint16_t filter_ht,
-                         const int32_t filter_offset,
-                         const int32_t *bias,
-                         uint8_t *out_data,
-                         const uint16_t out_wd,
-                         const uint16_t out_ht,
-                         const uint16_t out_channels,
-                         const int32_t out_offset,
-                         const int32_t out_shift,
-                         const int32_t out_mult,
-                         const int32_t activation_min,
-                         const int32_t activation_max)
-{
-    for (int out_y = 0; out_y < out_ht; out_y++) { //height loop
-        const int16_t base_y = (out_y * stride_ht) - pad_ht;
-        for (int out_x = 0; out_x < out_wd; out_x++) { //width_loop
-            const int16_t base_x = (out_x * stride_wd) - pad_wd;
-            for (int out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {//channel_loop
-                int32_t result = 0;
-
-                /* Select filter so as the point doesn't lie outside block */
-                int filter_y_start = max(0, -base_y);
-                int filter_x_start = max(0, -base_x);
-                int filter_y_end = min(filter_ht, input_ht - base_y);
-                int filter_x_end = min(filter_wd, input_wd - base_x);
-
-                for (int filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                    const int32_t idx_y = base_y + filter_y_idx;
-                    for (int filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                        const int32_t idx_x = base_x + filter_x_idx;
-                        for (int in_ch_idx = 0; in_ch_idx < in_channels; in_ch_idx++) {
-                            int32_t input_index = (idx_y * input_wd + idx_x) * in_channels + in_ch_idx;
-                            int32_t filter_index = ((out_ch_idx * filter_ht + filter_y_idx)
-                                                    * filter_wd + filter_x_idx) * in_channels
-                                                   + in_ch_idx;
-                            int32_t input_val = input_data[input_index] + input_offset;
-                            int32_t filter_val = filter_data[filter_index] + filter_offset;
-                            result += input_val * filter_val;
-                        }
-                    }
-                }
-                if (bias) {
-                    result += bias[out_ch_idx];
-                }
-                result = esp_nn_multiply_by_quantized_mult(result, out_mult, out_shift);
-                result += out_offset;
-                result = max(result, activation_min);
-                result = min(result, activation_max);
-
-                int out_index = (out_y * out_wd + out_x) * out_channels + out_ch_idx;
-                out_data[out_index] = (uint8_t) result;
-            }
-        }
-    }
-}
-
-/**
- * Assumption 1: i/p channels == o/p channels
- * Assumption 2: Pointers are valid
- * Assumption 3: dialation width = 1
- */
-void esp_nn_conv_s8_ansi(const int8_t *input_data,
-                         const uint16_t input_wd,
-                         const uint16_t input_ht,
-                         const uint16_t in_channels,
-                         const int32_t input_offset,
-                         const uint16_t pad_wd,
-                         const uint16_t pad_ht,
-                         const uint16_t stride_wd,
-                         const uint16_t stride_ht,
-                         const int8_t *filter_data,
-                         const uint16_t filter_wd,
-                         const uint16_t filter_ht,
-                         const int32_t *bias,
-                         int8_t *out_data,
-                         const uint16_t out_wd,
-                         const uint16_t out_ht,
-                         const uint16_t out_channels,
-                         const int32_t out_offset,
-                         const int32_t *out_shift,
-                         const int32_t *out_mult,
-                         const int32_t activation_min,
-                         const int32_t activation_max)
-{
-    int32_t out_ch_idx, out_y, out_x, in_ch_idx, filter_y_idx, filter_x_idx;
-
-    for (out_y = 0; out_y < out_ht; out_y++) {
-        for (out_x = 0; out_x < out_wd; out_x++) {
-            for (out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {
-                int32_t conv_out = 0;
-
-                const int32_t base_y = stride_ht * out_y - pad_ht;
-                const int32_t base_x = stride_wd * out_x - pad_wd;
-
-                const int32_t filter_y_start = max(0, -base_y);
-                const int32_t filter_x_start = max(0, -base_x);
-
-                const int32_t filter_y_end = min(filter_ht, input_ht - base_y);
-                const int32_t filter_x_end = min(filter_wd, input_wd - base_x);
-
-                for (filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                    for (filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                        const int32_t in_row = base_y + filter_y_idx;
-                        const int32_t in_col = base_x + filter_x_idx;
-                        int32_t input_base_offset = (in_row * input_wd + in_col) * in_channels;
-                        int32_t filter_base_offset = out_ch_idx * in_channels * filter_ht * filter_wd +
-                                                       (filter_y_idx * filter_wd + filter_x_idx) * in_channels;
-                        for (in_ch_idx = 0; in_ch_idx < in_channels; in_ch_idx++) {
-                            conv_out +=
-                                (input_data[input_base_offset + in_ch_idx] + input_offset) *
-                                filter_data[filter_base_offset + in_ch_idx];
-                        }
-                    }
-                }
-                if (bias) {
-                    conv_out += bias[out_ch_idx];
-                }
-                conv_out = esp_nn_multiply_by_quantized_mult(conv_out, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                conv_out += out_offset;
-                conv_out = max(conv_out, activation_min);
-                conv_out = min(conv_out, activation_max);
-                *out_data++ = (int8_t) conv_out;
-            }
-        }
-    }
-}

code/components/esp-nn/src/convolution/esp_nn_conv_esp32s3.c

@@ -1,436 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdio.h>
-
-#include <common_functions.h>
-
-static int16_t *scratch_buffer = NULL;
-
-extern void esp_nn_conv_s16_mult8_1x1_esp32s3(const int8_t *input_data,
-                                              const uint16_t input_wd,
-                                              const uint16_t input_ht,
-                                              const uint16_t in_channels,
-                                              const int32_t input_offset,
-                                              const int16_t *filter_data,
-                                              const int32_t *bias,
-                                              int8_t *out_data,
-                                              const uint16_t out_wd,
-                                              const uint16_t out_ht,
-                                              const uint16_t out_channels,
-                                              const int32_t out_offset,
-                                              const int32_t *out_shift,
-                                              const int32_t *out_mult,
-                                              const int32_t activation_min,
-                                              const int32_t activation_max,
-                                              void *buffer /* scratch buffer */);
-
-extern void esp_nn_conv_s16_mult4_1x1_esp32s3(const int16_t *input_data,
-                                              const uint16_t input_wd,
-                                              const uint16_t input_ht,
-                                              const uint16_t in_channels,
-                                              const int16_t *filter_data,
-                                              const int32_t *bias,
-                                              int8_t *out_data,
-                                              const uint16_t out_wd,
-                                              const uint16_t out_ht,
-                                              const uint16_t out_channels,
-                                              const int32_t out_offset,
-                                              const int32_t *out_shift,
-                                              const int32_t *out_mult,
-                                              const int32_t activation_min,
-                                              const int32_t activation_max,
-                                              void *buffer /* scratch buffer */);
-
-extern void esp_nn_conv_s16_mult8_esp32s3(const int16_t *input_data,
-                                          const uint16_t input_wd,
-                                          const uint16_t input_ht,
-                                          const uint16_t in_channels,
-                                          const uint16_t pad_wd,
-                                          const uint16_t pad_ht,
-                                          const uint16_t stride_wd,
-                                          const uint16_t stride_ht,
-                                          const int16_t *filter_data,
-                                          const uint16_t filter_wd,
-                                          const uint16_t filter_ht,
-                                          const int32_t *bias,
-                                          int8_t *out_data,
-                                          const uint16_t out_wd,
-                                          const uint16_t out_ht,
-                                          const uint16_t out_channels,
-                                          const int32_t out_offset,
-                                          const int32_t *out_shift,
-                                          const int32_t *out_mult,
-                                          const int32_t activation_min,
-                                          const int32_t activation_max);
-
-extern void esp_nn_aligned_s8_to_s16_with_offset_esp32s3(const int8_t *src, int16_t *dst,
-                                                         const int size, const int32_t offset);
-
-extern void esp_nn_s8_to_s16_esp32s3(const int8_t *src, int16_t *dst, const int size);
-
-static void esp_nn_conv_s8_unrolled(const int8_t *input_data,
-                                    const uint16_t input_wd,
-                                    const uint16_t input_ht,
-                                    const uint16_t in_channels,
-                                    const int32_t input_offset,
-                                    const uint16_t pad_wd,
-                                    const uint16_t pad_ht,
-                                    const uint16_t stride_wd,
-                                    const uint16_t stride_ht,
-                                    const int8_t *filter_data,
-                                    const uint16_t filter_wd,
-                                    const uint16_t filter_ht,
-                                    const int32_t *bias,
-                                    int8_t *out_data,
-                                    const uint16_t out_wd,
-                                    const uint16_t out_ht,
-                                    const uint16_t out_channels,
-                                    const int32_t out_offset,
-                                    const int32_t *out_shift,
-                                    const int32_t *out_mult,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max)
-{
-    int32_t out_ch_idx, out_y, out_x, in_ch_idx, filter_y_idx, filter_x_idx;
-
-    for (out_y = 0; out_y < out_ht; out_y++) {
-        for (out_x = 0; out_x < out_wd; out_x++) {
-            for (out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {
-                int32_t conv_out = 0;
-
-                const int32_t base_y = stride_ht * out_y - pad_ht;
-                const int32_t base_x = stride_wd * out_x - pad_wd;
-
-                const int32_t filter_y_start = max(0, -base_y);
-                const int32_t filter_x_start = max(0, -base_x);
-
-                const int32_t filter_y_end = min(filter_ht, input_ht - base_y);
-                const int32_t filter_x_end = min(filter_wd, input_wd - base_x);
-
-                for (filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                    for (filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                        const int32_t in_row = base_y + filter_y_idx;
-                        const int32_t in_col = base_x + filter_x_idx;
-                        int32_t input_base_offset = (in_row * input_wd + in_col) * in_channels;
-                        int32_t filter_base_offset = out_ch_idx * in_channels * filter_ht * filter_wd +
-                                                       (filter_y_idx * filter_wd + filter_x_idx) * in_channels;
-                        for (in_ch_idx = 0; in_ch_idx < in_channels; in_ch_idx++) {
-                            conv_out +=
-                                (input_data[input_base_offset + in_ch_idx] + input_offset) *
-                                filter_data[filter_base_offset + in_ch_idx];
-                        }
-                    }
-                }
-                if (bias) {
-                    conv_out += bias[out_ch_idx];
-                }
-                conv_out = esp_nn_multiply_by_quantized_mult_fast(conv_out, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                conv_out += out_offset;
-                conv_out = max(conv_out, activation_min);
-                conv_out = min(conv_out, activation_max);
-                *out_data++ = (int8_t) conv_out;
-            }
-        }
-    }
-}
-
-static void esp_nn_conv_s8_pad_valid(const int8_t *input_data,
-                                     const uint16_t input_wd,
-                                     const uint16_t input_ht,
-                                     const uint16_t in_channels,
-                                     const int32_t input_offset,
-                                     const uint16_t stride_wd,
-                                     const uint16_t stride_ht,
-                                     const int8_t *filter_data,
-                                     const uint16_t filter_wd,
-                                     const uint16_t filter_ht,
-                                     const int32_t *bias,
-                                     int8_t *out_data,
-                                     const uint16_t out_wd,
-                                     const uint16_t out_ht,
-                                     const uint16_t out_channels,
-                                     const int32_t out_offset,
-                                     const int32_t *out_shift,
-                                     const int32_t *out_mult,
-                                     const int32_t activation_min,
-                                     const int32_t activation_max)
-{
-    int32_t out_ch_idx, out_y, out_x, in_ch_idx, filter_y_idx, filter_x_idx;
-
-    for (out_y = 0; out_y < out_ht; out_y++) {
-        for (out_x = 0; out_x < out_wd; out_x++) {
-            for (out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {
-                int32_t conv_out = 0;
-
-                const int32_t base_y = stride_ht * out_y;
-                const int32_t base_x = stride_wd * out_x;
-
-                for (filter_y_idx = 0; filter_y_idx < filter_ht; filter_y_idx++) {
-                    for (filter_x_idx = 0; filter_x_idx < filter_wd; filter_x_idx++) {
-                        const int32_t in_row = base_y + filter_y_idx;
-                        const int32_t in_col = base_x + filter_x_idx;
-                        int32_t input_base_offset = (in_row * input_wd + in_col) * in_channels;
-                        int32_t filter_base_offset = out_ch_idx * in_channels * filter_ht * filter_wd +
-                                                       (filter_y_idx * filter_wd + filter_x_idx) * in_channels;
-                        const int8_t *input_data_ptr = input_data + input_base_offset;
-                        const int8_t *filter_data_ptr = filter_data + filter_base_offset;
-                        for (in_ch_idx = 0; in_ch_idx < in_channels; in_ch_idx++) {
-                            conv_out += (*input_data_ptr++ + input_offset) * *filter_data_ptr++;
-                        }
-                    }
-                }
-                if (bias) {
-                    conv_out += bias[out_ch_idx];
-                }
-                conv_out = esp_nn_multiply_by_quantized_mult_fast(conv_out, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                conv_out += out_offset;
-                conv_out = max(conv_out, activation_min);
-                conv_out = min(conv_out, activation_max);
-                *out_data++ = (int8_t) conv_out;
-            }
-        }
-    }
-}
-
-static void esp_nn_conv_s8_pad_valid_3x3(const int8_t *input_data,
-                                         const uint16_t input_wd,
-                                         const uint16_t input_ht,
-                                         const uint16_t in_channels,
-                                         const int32_t input_offset,
-                                         const uint16_t stride_wd,
-                                         const uint16_t stride_ht,
-                                         const int8_t *filter_data,
-                                         const int32_t *bias,
-                                         int8_t *out_data,
-                                         const uint16_t out_wd,
-                                         const uint16_t out_ht,
-                                         const uint16_t out_channels,
-                                         const int32_t out_offset,
-                                         const int32_t *out_shift,
-                                         const int32_t *out_mult,
-                                         const int32_t activation_min,
-                                         const int32_t activation_max)
-{
-    int32_t out_ch_idx, out_y, out_x, in_ch_idx, filter_y_idx, filter_x_idx;
-
-    for (out_y = 0; out_y < out_ht; out_y++) {
-        for (out_x = 0; out_x < out_wd; out_x++) {
-            const int32_t base_y = stride_ht * out_y;
-            const int32_t base_x = stride_wd * out_x;
-            for (out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {
-                int32_t conv_out = 0;
-                for (filter_y_idx = 0; filter_y_idx < 3; filter_y_idx++) {
-                    for (filter_x_idx = 0; filter_x_idx < 3; filter_x_idx++) {
-                        const int32_t in_row = base_y + filter_y_idx;
-                        const int32_t in_col = base_x + filter_x_idx;
-                        int32_t input_base_offset = (in_row * input_wd + in_col) * in_channels;
-                        int32_t filter_base_offset = out_ch_idx * in_channels * 3 * 3 +
-                                                       (filter_y_idx * 3 + filter_x_idx) * in_channels;
-                        const int8_t *input_data_ptr = input_data + input_base_offset;
-                        const int8_t *filter_data_ptr = filter_data + filter_base_offset;
-                        for (in_ch_idx = 0; in_ch_idx < in_channels; in_ch_idx++) {
-                            conv_out += (*input_data_ptr++ + input_offset) * *filter_data_ptr++;
-                        }
-                    }
-                }
-                if (bias) {
-                    conv_out += bias[out_ch_idx];
-                }
-                conv_out = esp_nn_multiply_by_quantized_mult_fast(conv_out, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                conv_out += out_offset;
-                conv_out = max(conv_out, activation_min);
-                conv_out = min(conv_out, activation_max);
-                *out_data++ = (int8_t) conv_out;
-            }
-        }
-    }
-}
-
-static void esp_nn_conv_s8_pad_valid_ch3_3x3(const int8_t *input_data,
-                                             const uint16_t input_wd,
-                                             const uint16_t input_ht,
-                                             const int32_t input_offset,
-                                             const uint16_t stride_wd,
-                                             const uint16_t stride_ht,
-                                             const int8_t *filter_data,
-                                             const int32_t *bias,
-                                             int8_t *out_data,
-                                             const uint16_t out_wd,
-                                             const uint16_t out_ht,
-                                             const uint16_t out_channels,
-                                             const int32_t out_offset,
-                                             const int32_t *out_shift,
-                                             const int32_t *out_mult,
-                                             const int32_t activation_min,
-                                             const int32_t activation_max)
-{
-    int32_t out_ch_idx, out_y, out_x, filter_y_idx;
-
-    /* use scratch_buffer to pre-compute offset factor */
-    int16_t *filter_sum = (int16_t *) scratch_buffer;
-    const int8_t *filter_ptr = filter_data;
-    for (out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {
-        int16_t sum_val = 0;
-        for (int i = 0; i < 9; i++) {
-            sum_val += *filter_ptr++;
-            sum_val += *filter_ptr++;
-            sum_val += *filter_ptr++;
-        }
-        *filter_sum++ = sum_val;
-    }
-
-    for (out_y = 0; out_y < out_ht; out_y++) {
-        for (out_x = 0; out_x < out_wd; out_x++) {
-            const int8_t *filter_data_ptr = filter_data;
-            const int32_t base_y = stride_ht * out_y;
-            const int32_t base_x = stride_wd * out_x;
-            const int8_t *input_base_ptr = input_data + (base_y * input_wd + base_x) * 3;
-            int16_t *filter_sum = (int16_t *) scratch_buffer;
-            for (out_ch_idx = 0; out_ch_idx < out_channels; out_ch_idx++) {
-                int32_t conv_out = 0;
-
-                for (filter_y_idx = 0; filter_y_idx < 3; filter_y_idx++) {
-                    const int8_t *input_data_ptr = input_base_ptr + (filter_y_idx * input_wd) * 3;
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                    conv_out += (*input_data_ptr++) * (*filter_data_ptr++);
-                }
-
-                conv_out += *filter_sum++ * input_offset;
-
-                if (bias) {
-                    conv_out += bias[out_ch_idx];
-                }
-                conv_out = esp_nn_multiply_by_quantized_mult_fast(conv_out, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                conv_out += out_offset;
-                conv_out = max(conv_out, activation_min);
-                conv_out = min(conv_out, activation_max);
-                *out_data++ = (int8_t) conv_out;
-            }
-        }
-    }
-}
-
-int esp_nn_get_conv_scratch_size_esp32s3(const uint16_t input_wd,
-                                         const uint16_t input_ht,
-                                         const uint16_t in_ch,
-                                         const uint16_t out_ch,
-                                         const uint16_t filter_wd,
-                                         const uint16_t filter_ht)
-{
-    int filter_size = filter_wd * filter_ht * in_ch * out_ch;
-    int input_size = input_wd * input_ht * in_ch;
-    int transpose_buf_size = 8 * in_ch; /* to store intermediate data */
-    int align_buf_size = 32; /* extra buffer for alignment */
-    return 2 * (filter_size + input_size +  transpose_buf_size) + align_buf_size;
-}
-
-void esp_nn_set_conv_scratch_buf_esp32s3(void *buf)
-{
-    scratch_buffer = (int16_t *) buf;
-}
-
-void esp_nn_conv_s8_esp32s3(const int8_t *input,
-                            const uint16_t input_wd,
-                            const uint16_t input_ht,
-                            const uint16_t channels,
-                            const int32_t input_offset,
-                            const uint16_t pad_wd,
-                            const uint16_t pad_ht,
-                            const uint16_t stride_wd,
-                            const uint16_t stride_ht,
-                            const int8_t *filter_data,
-                            const uint16_t filter_wd,
-                            const uint16_t filter_ht,
-                            const int32_t *bias,
-                            int8_t *out_data,
-                            const uint16_t out_wd,
-                            const uint16_t out_ht,
-                            const uint16_t out_channels,
-                            const int32_t out_offset,
-                            const int32_t *out_shift,
-                            const int32_t *out_mult,
-                            const int32_t activation_min,
-                            const int32_t activation_max)
-{
-    int filter_size = filter_wd * filter_ht * channels * out_channels;
-    int input_size = input_wd * input_ht * channels;
-    int align_len = 16 - (filter_size & 15);
-    int16_t *filter_data16 = scratch_buffer;
-    int16_t *input_data16 = scratch_buffer + filter_size + align_len;
-
-    if (scratch_buffer == NULL) {
-        printf("esp_nn_conv error! scratch_buffer not set!\n");
-        return;
-    }
-
-    if (channels % 8 == 0 && filter_wd == 1 && filter_ht == 1 &&
-            pad_wd == 0 && pad_ht == 0 && stride_wd == 1 && stride_ht == 1) {
-        int scratch_offset = (int) (filter_data16 + filter_size);
-        void *scratch_buf = (void *) (scratch_offset + 16 - (scratch_offset & 15));
-        esp_nn_s8_to_s16_esp32s3(filter_data, filter_data16, filter_size);
-        esp_nn_conv_s16_mult8_1x1_esp32s3(
-            input, input_wd, input_ht, channels, input_offset, filter_data16,
-            bias, out_data, out_wd, out_ht, out_channels, out_offset,
-            out_shift, out_mult, activation_min, activation_max, scratch_buf);
-    } else if (channels % 4 == 0 && filter_wd == 1 && filter_ht == 1 &&
-            (input_wd * input_ht) % 16 == 0 && /* TODO: remove this check */
-            pad_wd == 0 && pad_ht == 0 && stride_wd == 1 && stride_ht == 1) {
-        int scratch_offset = (int) (input_data16 + input_size);
-        void *scratch_buf = (void *) (scratch_offset + 16 - (scratch_offset & 15));
-        esp_nn_s8_to_s16_esp32s3(filter_data, filter_data16, filter_size);
-        esp_nn_aligned_s8_to_s16_with_offset_esp32s3(input, input_data16, input_size, input_offset);
-        esp_nn_conv_s16_mult4_1x1_esp32s3(
-            input_data16, input_wd, input_ht, channels, filter_data16,
-            bias, out_data, out_wd, out_ht, out_channels, out_offset,
-            out_shift, out_mult, activation_min, activation_max, scratch_buf);
-    } else if (channels % 8 == 0) {
-        esp_nn_s8_to_s16_esp32s3(filter_data, filter_data16, filter_size);
-        esp_nn_aligned_s8_to_s16_with_offset_esp32s3(input, input_data16, input_size, input_offset);
-        esp_nn_conv_s16_mult8_esp32s3(
-            input_data16, input_wd, input_ht, channels, pad_wd, pad_ht,
-            stride_wd, stride_ht, filter_data16, filter_wd, filter_ht, bias,
-            out_data, out_wd, out_ht, out_channels, out_offset, out_shift,
-            out_mult, activation_min, activation_max);
-    } else if (pad_wd == 0 && pad_ht == 0) {
-        if (filter_wd == 3 && filter_ht == 3 && channels == 3) {
-            esp_nn_conv_s8_pad_valid_ch3_3x3(input, input_wd, input_ht, input_offset,
-                                             stride_wd, stride_ht, filter_data, bias,
-                                             out_data, out_wd, out_ht, out_channels, out_offset,
-                                             out_shift, out_mult, activation_min, activation_max);
-        } else {
-            esp_nn_conv_s8_pad_valid(input, input_wd, input_ht, channels, input_offset,
-                                     stride_wd, stride_ht, filter_data, filter_wd, filter_ht, bias,
-                                     out_data, out_wd, out_ht, out_channels, out_offset, out_shift,
-                                     out_mult, activation_min, activation_max);
-        }
-    } else {
-        /* Basic unrolled version */
-        esp_nn_conv_s8_unrolled(input, input_wd, input_ht, channels, input_offset,
-                                pad_wd, pad_ht, stride_wd, stride_ht,
-                                filter_data, filter_wd, filter_ht, bias,
-                                out_data, out_wd, out_ht, out_channels, out_offset, out_shift,
-                                out_mult, activation_min, activation_max);
-    }
-}

code/components/esp-nn/src/convolution/esp_nn_depthwise_conv_ansi.c

@@ -1,97 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-int esp_nn_get_depthwise_conv_scratch_size_ansi(const uint16_t input_wd,
-                                                const uint16_t input_ht,
-                                                const uint16_t channels,
-                                                const uint16_t ch_mult,
-                                                const uint16_t filter_wd,
-                                                const uint16_t filter_ht)
-{
-    return 0;
-}
-
-void esp_nn_set_depthwise_conv_scratch_buf_ansi(const void *buf)
-{
-
-}
-
-void esp_nn_depthwise_conv_s8_ansi(const int8_t *input_data,
-                                   const uint16_t input_wd,
-                                   const uint16_t input_ht,
-                                   const uint16_t channels,
-                                   const int32_t input_offset,
-                                   const uint16_t pad_wd,
-                                   const uint16_t pad_ht,
-                                   const uint16_t stride_wd,
-                                   const uint16_t stride_ht,
-                                   const uint16_t ch_mult,
-                                   const int8_t *filter_data,
-                                   const uint16_t filter_wd,
-                                   const uint16_t filter_ht,
-                                   const int32_t *bias,
-                                   int8_t *out_data,
-                                   const uint16_t out_wd,
-                                   const uint16_t out_ht,
-                                   const int32_t out_offset,
-                                   const int32_t *out_shift,
-                                   const int32_t *out_mult,
-                                   const int32_t activation_min,
-                                   const int32_t activation_max)
-{
-    int out_idx = 0;
-    for (int out_y = 0; out_y < out_ht; out_y++) { //height loop
-        const int16_t base_y = (out_y * stride_ht) - pad_ht;
-        for (int out_x = 0; out_x < out_wd; out_x++) { //width_loop
-            const int16_t base_x = (out_x * stride_wd) - pad_wd;
-            for (int ch_idx = 0; ch_idx < channels; ch_idx++) {//channel_loop
-                for (int ch_mult_idx = 0; ch_mult_idx < ch_mult; ch_mult_idx++) {
-                    int32_t result = 0;
-                    const int out_ch_idx = ch_mult_idx + ch_idx * ch_mult;
-
-                    /* Select filter so as the point doesn't lie outside block */
-                    int filter_y_start = max(0, -base_y);
-                    int filter_x_start = max(0, -base_x);
-                    int filter_y_end = min(filter_ht, input_ht - base_y);
-                    int filter_x_end = min(filter_wd, input_wd - base_x);
-
-                    for (int filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                        const int32_t idx_y = base_y + filter_y_idx;
-                        for (int filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                            const int32_t idx_x = base_x + filter_x_idx;
-                            int32_t input_index = (idx_y * input_wd + idx_x) * channels + ch_idx;
-                            int32_t filter_index = (filter_y_idx * filter_wd + filter_x_idx) * (channels * ch_mult) + out_ch_idx;
-                            int32_t input_val = input_data[input_index] + input_offset;
-                            int32_t filter_val = filter_data[filter_index];
-                            result += input_val * filter_val;
-                        }
-                    }
-                    if (bias) {
-                        result += bias[out_ch_idx];
-                    }
-                    result = esp_nn_multiply_by_quantized_mult(result, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                    result += out_offset;
-                    result = max(result, activation_min);
-                    result = min(result, activation_max);
-
-                    out_data[out_idx++] = result;
-                }
-            }
-        }
-    }
-}

code/components/esp-nn/src/convolution/esp_nn_depthwise_conv_s8_esp32s3.c

@@ -1,483 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdio.h>
-
-#include <common_functions.h>
-
-static int16_t *scratch_buffer = NULL;
-
-extern void esp_nn_depthwise_conv_s16_mult8_3x3_esp32s3(const int16_t *input_data,
-                                                        const uint16_t input_wd,
-                                                        const uint16_t input_ht,
-                                                        const uint16_t channels,
-                                                        const uint16_t pad_wd,
-                                                        const uint16_t pad_ht,
-                                                        const uint16_t stride_wd,
-                                                        const uint16_t stride_ht,
-                                                        const uint16_t ch_mult,
-                                                        const int16_t *filter_data,
-                                                        const int32_t *bias,
-                                                        int8_t *out_data,
-                                                        const uint16_t out_wd,
-                                                        const uint16_t out_ht,
-                                                        const int32_t out_offset,
-                                                        const int32_t *out_shift,
-                                                        const int32_t *out_mult,
-                                                        const int32_t activation_min,
-                                                        const int32_t activation_max);
-
-extern void esp_nn_depthwise_conv_s8_mult1_3x3_padded_esp32s3(const int8_t *input_data,
-                                                              const uint16_t input_wd,
-                                                              const uint16_t input_ht,
-                                                              const uint16_t channels,
-                                                              const int32_t input_offset,
-                                                              const uint16_t stride_wd,
-                                                              const uint16_t stride_ht,
-                                                              const int8_t *filter_data,
-                                                              const int32_t *bias,
-                                                              int8_t *out_data,
-                                                              const uint16_t out_wd,
-                                                              const uint16_t out_ht,
-                                                              const int32_t out_offset,
-                                                              const int32_t *out_shift,
-                                                              const int32_t *out_mult,
-                                                              const int32_t activation_min,
-                                                              const int32_t activation_max);
-
-extern void esp_nn_depthwise_conv_s16_mult1_3x3_no_pad_esp32s3(const int16_t *input_data,
-                                                               const uint16_t input_wd,
-                                                               const uint16_t input_ht,
-                                                               const uint16_t channels,
-                                                               const uint16_t stride_wd,
-                                                               const uint16_t stride_ht,
-                                                               const int16_t *filter_data,
-                                                               const int32_t *bias,
-                                                               int8_t *out_data,
-                                                               const uint16_t out_wd,
-                                                               const uint16_t out_ht,
-                                                               const int32_t out_offset,
-                                                               const int32_t *out_shift,
-                                                               const int32_t *out_mult,
-                                                               const int32_t activation_min,
-                                                               const int32_t activation_max);
-
-extern void esp_nn_depthwise_conv_s16_mult8_esp32s3(const int16_t *input_data,
-                                                    const uint16_t input_wd,
-                                                    const uint16_t input_ht,
-                                                    const uint16_t channels,
-                                                    const uint16_t pad_wd,
-                                                    const uint16_t pad_ht,
-                                                    const uint16_t stride_wd,
-                                                    const uint16_t stride_ht,
-                                                    const uint16_t ch_mult,
-                                                    const int16_t *filter_data,
-                                                    const uint16_t filter_wd,
-                                                    const uint16_t filter_ht,
-                                                    const int32_t *bias,
-                                                    int8_t *out_data,
-                                                    const uint16_t out_wd,
-                                                    const uint16_t out_ht,
-                                                    const int32_t out_offset,
-                                                    const int32_t *out_shift,
-                                                    const int32_t *out_mult,
-                                                    const int32_t activation_min,
-                                                    const int32_t activation_max);
-
-extern void esp_nn_depthwise_conv_s16_mult4_esp32s3(const int16_t *input_data,
-                                                    const uint16_t input_wd,
-                                                    const uint16_t input_ht,
-                                                    const uint16_t channels,
-                                                    const uint16_t pad_wd,
-                                                    const uint16_t pad_ht,
-                                                    const uint16_t stride_wd,
-                                                    const uint16_t stride_ht,
-                                                    const uint16_t ch_mult,
-                                                    const int16_t *filter_data,
-                                                    const uint16_t filter_wd,
-                                                    const uint16_t filter_ht,
-                                                    const int32_t *bias,
-                                                    int8_t *out_data,
-                                                    const uint16_t out_wd,
-                                                    const uint16_t out_ht,
-                                                    const int32_t out_offset,
-                                                    const int32_t *out_shift,
-                                                    const int32_t *out_mult,
-                                                    const int32_t activation_min,
-                                                    const int32_t activation_max);
-
-extern void esp_nn_depthwise_conv_s16_mult1_3x3_esp32s3(const int16_t *input_data,
-                                                        const uint16_t input_wd,
-                                                        const uint16_t input_ht,
-                                                        const uint16_t channels,
-                                                        const uint16_t pad_wd,
-                                                        const uint16_t pad_ht,
-                                                        const uint16_t stride_wd,
-                                                        const uint16_t stride_ht,
-                                                        const int16_t *filter_data,
-                                                        const int32_t *bias,
-                                                        int8_t *out_data,
-                                                        const uint16_t out_wd,
-                                                        const uint16_t out_ht,
-                                                        const int32_t out_offset,
-                                                        const int32_t *out_shift,
-                                                        const int32_t *out_mult,
-                                                        const int32_t activation_min,
-                                                        const int32_t activation_max);
-
-extern void esp_nn_depthwise_conv_s16_mult1_esp32s3(const int16_t *input_data,
-                                                    const uint16_t input_wd,
-                                                    const uint16_t input_ht,
-                                                    const uint16_t channels,
-                                                    const uint16_t pad_wd,
-                                                    const uint16_t pad_ht,
-                                                    const uint16_t stride_wd,
-                                                    const uint16_t stride_ht,
-                                                    const int16_t *filter_data,
-                                                    const uint16_t filter_wd,
-                                                    const uint16_t filter_ht,
-                                                    const int32_t *bias,
-                                                    int8_t *out_data,
-                                                    const uint16_t out_wd,
-                                                    const uint16_t out_ht,
-                                                    const int32_t out_offset,
-                                                    const int32_t *out_shift,
-                                                    const int32_t *out_mult,
-                                                    const int32_t activation_min,
-                                                    const int32_t activation_max);
-
-extern void esp_nn_s8_to_s16_esp32s3(const int8_t *src, int16_t *dst, const int size);
-
-extern void esp_nn_aligned_s8_to_s16_with_offset_esp32s3(const int8_t *src, int16_t *dst,
-                                                         const int size, const int32_t offset);
-
-static void esp_nn_depthwise_conv_s8_unrolled(const int8_t *input_data,
-                                              const uint16_t input_wd,
-                                              const uint16_t input_ht,
-                                              const uint16_t channels,
-                                              const int32_t input_offset,
-                                              const uint16_t pad_wd,
-                                              const uint16_t pad_ht,
-                                              const uint16_t stride_wd,
-                                              const uint16_t stride_ht,
-                                              const uint16_t ch_mult,
-                                              const int8_t *filter_data,
-                                              const uint16_t filter_wd,
-                                              const uint16_t filter_ht,
-                                              const int32_t *bias,
-                                              int8_t *out_data,
-                                              const uint16_t out_wd,
-                                              const uint16_t out_ht,
-                                              const int32_t out_offset,
-                                              const int32_t *out_shift,
-                                              const int32_t *out_mult,
-                                              const int32_t activation_min,
-                                              const int32_t activation_max)
-{
-    int out_idx = 0;
-    for (int out_y = 0; out_y < out_ht; out_y++) { //height loop
-        const int16_t base_y = (out_y * stride_ht) - pad_ht;
-        for (int out_x = 0; out_x < out_wd; out_x++) { //width_loop
-            const int16_t base_x = (out_x * stride_wd) - pad_wd;
-            for (int ch_idx = 0; ch_idx < channels; ch_idx++) {//channel_loop
-                int ch_mult_idx = 0;
-                for (; ch_mult_idx < ch_mult - 3; ch_mult_idx += 4) {
-                    int32_t result0 = 0, result1 = 0, result2 = 0, result3 = 0;
-                    const int out_ch_idx = ch_mult_idx + ch_idx * ch_mult;
-
-                    /* Select filter so as the point doesn't lie outside block */
-                    int filter_y_start = max(0, -base_y);
-                    int filter_x_start = max(0, -base_x);
-                    int filter_y_end = min(filter_ht, input_ht - base_y);
-                    int filter_x_end = min(filter_wd, input_wd - base_x);
-
-                    for (int filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                        const int32_t idx_y = base_y + filter_y_idx;
-                        for (int filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                            const int32_t idx_x = base_x + filter_x_idx;
-                            int32_t input_index = (idx_y * input_wd + idx_x) * channels + ch_idx;
-                            int32_t filter_index = (filter_y_idx * filter_wd + filter_x_idx) * (channels * ch_mult) + out_ch_idx;
-                            int32_t input_val = input_data[input_index] + input_offset;
-                            int32_t filter_val0 = filter_data[filter_index + 0];
-                            int32_t filter_val1 = filter_data[filter_index + 1];
-                            int32_t filter_val2 = filter_data[filter_index + 2];
-                            int32_t filter_val3 = filter_data[filter_index + 3];
-                            result0 += input_val * filter_val0;
-                            result1 += input_val * filter_val1;
-                            result2 += input_val * filter_val2;
-                            result3 += input_val * filter_val3;
-                        }
-                    }
-                    if (bias) {
-                        result0 += bias[out_ch_idx + 0];
-                        result1 += bias[out_ch_idx + 1];
-                        result2 += bias[out_ch_idx + 2];
-                        result3 += bias[out_ch_idx + 3];
-                    }
-                    result0 = esp_nn_multiply_by_quantized_mult(result0,
-                                out_mult[out_ch_idx + 0], out_shift[out_ch_idx + 0]);
-                    result1 = esp_nn_multiply_by_quantized_mult(result1,
-                                out_mult[out_ch_idx + 1], out_shift[out_ch_idx + 1]);
-                    result2 = esp_nn_multiply_by_quantized_mult(result2,
-                                out_mult[out_ch_idx + 2], out_shift[out_ch_idx + 2]);
-                    result3 = esp_nn_multiply_by_quantized_mult(result3,
-                                out_mult[out_ch_idx + 3], out_shift[out_ch_idx + 3]);
-
-                    result0 += out_offset;
-                    result1 += out_offset;
-                    result2 += out_offset;
-                    result3 += out_offset;
-
-                    result0 = max(result0, activation_min);
-                    result1 = max(result1, activation_min);
-                    result2 = max(result2, activation_min);
-                    result3 = max(result3, activation_min);
-
-                    result0 = min(result0, activation_max);
-                    result1 = min(result1, activation_max);
-                    result2 = min(result2, activation_max);
-                    result3 = min(result3, activation_max);
-
-                    out_data[out_idx++] = result0;
-                    out_data[out_idx++] = result1;
-                    out_data[out_idx++] = result2;
-                    out_data[out_idx++] = result3;
-                }
-
-                /* left-over */
-                for (; ch_mult_idx < ch_mult; ch_mult_idx++) {
-                    int32_t result = 0;
-                    const int out_ch_idx = ch_mult_idx + ch_idx * ch_mult;
-
-                    /* Select filter so as the point doesn't lie outside block */
-                    int filter_y_start = max(0, -base_y);
-                    int filter_x_start = max(0, -base_x);
-                    int filter_y_end = min(filter_ht, input_ht - base_y);
-                    int filter_x_end = min(filter_wd, input_wd - base_x);
-
-                    for (int filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                        const int32_t idx_y = base_y + filter_y_idx;
-                        for (int filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                            const int32_t idx_x = base_x + filter_x_idx;
-                            int32_t input_index = (idx_y * input_wd + idx_x) * channels + ch_idx;
-                            int32_t filter_index = (filter_y_idx * filter_wd + filter_x_idx) * (channels * ch_mult) + out_ch_idx;
-                            int32_t input_val = input_data[input_index] + input_offset;
-                            int32_t filter_val = filter_data[filter_index];
-                            result += input_val * filter_val;
-                        }
-                    }
-                    if (bias) {
-                        result += bias[out_ch_idx];
-                    }
-                    result = esp_nn_multiply_by_quantized_mult(result, out_mult[out_ch_idx], out_shift[out_ch_idx]);
-                    result += out_offset;
-                    result = max(result, activation_min);
-                    result = min(result, activation_max);
-
-                    out_data[out_idx++] = result;
-                }
-            }
-        }
-    }
-}
-
-void esp_nn_depthwise_conv_s8_ch_mult1(const int8_t *input_data,
-                                       const uint16_t input_wd,
-                                       const uint16_t input_ht,
-                                       const uint16_t channels,
-                                       const int32_t input_offset,
-                                       const uint16_t pad_wd,
-                                       const uint16_t pad_ht,
-                                       const uint16_t stride_wd,
-                                       const uint16_t stride_ht,
-                                       const int8_t *filter_data,
-                                       const uint16_t filter_wd,
-                                       const uint16_t filter_ht,
-                                       const int32_t *bias,
-                                       int8_t *out_data,
-                                       const uint16_t out_wd,
-                                       const uint16_t out_ht,
-                                       const int32_t out_offset,
-                                       const int32_t *out_shift,
-                                       const int32_t *out_mult,
-                                       const int32_t activation_min,
-                                       const int32_t activation_max)
-{
-    int out_idx = 0;
-    for (int out_y = 0; out_y < out_ht; out_y++) { //height loop
-        const int16_t base_y = (out_y * stride_ht) - pad_ht;
-        for (int out_x = 0; out_x < out_wd; out_x++) { //width_loop
-            const int16_t base_x = (out_x * stride_wd) - pad_wd;
-            for (int ch_idx = 0; ch_idx < channels; ch_idx++) {//channel_loop
-                int32_t result = 0;
-                /* Select filter so as the point doesn't lie outside block */
-                int filter_y_start = max(0, -base_y);
-                int filter_x_start = max(0, -base_x);
-                int filter_y_end = min(filter_ht, input_ht - base_y);
-                int filter_x_end = min(filter_wd, input_wd - base_x);
-
-                for (int filter_y_idx = filter_y_start; filter_y_idx < filter_y_end; filter_y_idx++) {
-                    const int32_t idx_y = base_y + filter_y_idx;
-                    for (int filter_x_idx = filter_x_start; filter_x_idx < filter_x_end; filter_x_idx++) {
-                        const int32_t idx_x = base_x + filter_x_idx;
-                        int32_t input_index = (idx_y * input_wd + idx_x) * channels + ch_idx;
-                        int32_t filter_index = (filter_y_idx * filter_wd + filter_x_idx) * channels + ch_idx;
-                        int32_t input_val = input_data[input_index] + input_offset;
-                        int32_t filter_val = filter_data[filter_index];
-                        result += input_val * filter_val;
-                    }
-                }
-                if (bias) {
-                    result += bias[ch_idx];
-                }
-                result = esp_nn_multiply_by_quantized_mult(result, out_mult[ch_idx], out_shift[ch_idx]);
-                result += out_offset;
-                result = max(result, activation_min);
-                result = min(result, activation_max);
-
-                out_data[out_idx++] = result;
-            }
-        }
-    }
-}
-
-int esp_nn_get_depthwise_conv_scratch_size_esp32s3(const uint16_t input_wd,
-                                                   const uint16_t input_ht,
-                                                   const uint16_t channels,
-                                                   const uint16_t ch_mult,
-                                                   const uint16_t filter_wd,
-                                                   const uint16_t filter_ht)
-{
-    int filter_size = filter_wd * filter_ht * channels * ch_mult;
-    int padding_used = ((filter_wd == 3) && (filter_ht == 3)) * 2;
-    int input_size = (input_wd + padding_used) * (input_ht + padding_used) * channels;
-    return  2 * (filter_size + input_size) + 16; //16 for alignment
-}
-
-void esp_nn_set_depthwise_conv_scratch_buf_esp32s3(void *buf)
-{
-    scratch_buffer = (int16_t *) buf;
-}
-
-/**
- * Assumption 1: i/p channels == o/p channels
- * Assumption 2: Pointers are valid
- * Assumption 3: dialation width = 1
- */
-void esp_nn_depthwise_conv_s8_esp32s3(const int8_t *input_data,
-                                      const uint16_t input_wd,
-                                      const uint16_t input_ht,
-                                      const uint16_t channels,
-                                      const int32_t input_offset,
-                                      const uint16_t pad_wd,
-                                      const uint16_t pad_ht,
-                                      const uint16_t stride_wd,
-                                      const uint16_t stride_ht,
-                                      const uint16_t ch_mult,
-                                      const int8_t *filter_data,
-                                      const uint16_t filter_wd,
-                                      const uint16_t filter_ht,
-                                      const int32_t *bias,
-                                      int8_t *out_data,
-                                      const uint16_t out_wd,
-                                      const uint16_t out_ht,
-                                      const int32_t out_offset,
-                                      const int32_t *out_shift,
-                                      const int32_t *out_mult,
-                                      const int32_t activation_min,
-                                      const int32_t activation_max)
-{
-    int filter_size = filter_wd * filter_ht * channels * ch_mult;
-    int align_len = 16 - (filter_size & 15);
-    int input_size = input_wd * input_ht * channels;
-    int16_t *filter_data16 = scratch_buffer;
-    int16_t *input_data16 = scratch_buffer + filter_size + align_len;
-    if (scratch_buffer == NULL) {
-        printf("esp_nn_depthwise_conv error! scratch_buffer not set!\n");
-        return;
-    }
-
-    if ((ch_mult == 1) && (channels % 8 == 0)) {
-        if ((filter_wd == 3) && (filter_ht == 3)) {
-            if ((channels % 16 == 0) && (pad_wd == 1) && (pad_ht == 1)) {
-                /* process in 8 bits */
-                int8_t *filter_aligned = (int8_t *) scratch_buffer;
-                int8_t *input_padded = (int8_t *) scratch_buffer + filter_size + align_len;
-                memcpy(filter_aligned, filter_data, filter_size);
-                esp_nn_aligned_s8_pad_with_value(input_data, input_padded, input_wd, input_ht, channels,
-                                                 -input_offset, pad_wd, pad_ht);
-                esp_nn_depthwise_conv_s8_mult1_3x3_padded_esp32s3(input_padded, input_wd + 2 * pad_wd,
-                                                                  input_ht + 2 * pad_ht, channels, input_offset,
-                                                                  stride_wd, stride_ht, filter_aligned, bias,
-                                                                  out_data, out_wd, out_ht, out_offset, out_shift,
-                                                                  out_mult, activation_min, activation_max);
-            } else if ((pad_wd == 0) && (pad_ht == 0) &&
-                    // because this does not handle padding offset cases yet, run just for stride (1, 1).
-                    // end padding of input with `-input_offset` should solve this
-                    (stride_wd == 1) && (stride_ht == 1)) {
-                /* process in 8 bits */
-                int8_t *filter_aligned = (int8_t *) scratch_buffer;
-                memcpy(filter_aligned, filter_data, filter_size);
-                esp_nn_depthwise_conv_s8_mult1_3x3_padded_esp32s3(input_data, input_wd, input_ht, channels, input_offset,
-                                                                  stride_wd, stride_ht, filter_aligned,
-                                                                  bias, out_data, out_wd, out_ht, out_offset, out_shift,
-                                                                  out_mult, activation_min, activation_max);
-            } else { /* (channels % 8) == 0 && pad_wd == 1 && pad_ht == 1 */
-                esp_nn_s8_to_s16_esp32s3(filter_data, filter_data16, filter_size);
-                esp_nn_aligned_s8_to_s16_with_offset_esp32s3(input_data, input_data16, input_size, input_offset);
-                esp_nn_depthwise_conv_s16_mult1_3x3_esp32s3(input_data16, input_wd, input_ht, channels,
-                                                            pad_wd, pad_ht, stride_wd, stride_ht, filter_data16,
-                                                            bias, out_data, out_wd, out_ht, out_offset, out_shift,
-                                                            out_mult, activation_min, activation_max);
-            }
-        } else { // all other ch_mult == 1, `channels % 8 == 0`
-            esp_nn_depthwise_conv_s8_ch_mult1(input_data, input_wd, input_ht, channels, input_offset,
-                                              pad_wd, pad_ht, stride_wd, stride_ht,
-                                              filter_data, filter_wd, filter_ht,
-                                              bias, out_data, out_wd, out_ht, out_offset, out_shift,
-                                              out_mult, activation_min, activation_max);
-        }
-    } else if (ch_mult % 8 == 0) {
-        esp_nn_s8_to_s16_esp32s3(filter_data, filter_data16, filter_size);
-        esp_nn_aligned_s8_to_s16_with_offset_esp32s3(input_data, input_data16, input_size, input_offset);
-        if (filter_wd == 3 && filter_ht == 3) {
-            esp_nn_depthwise_conv_s16_mult8_3x3_esp32s3(input_data16, input_wd, input_ht, channels,
-                                                        pad_wd, pad_ht, stride_wd, stride_ht, ch_mult,
-                                                        filter_data16, bias,
-                                                        out_data, out_wd, out_ht, out_offset, out_shift,
-                                                        out_mult, activation_min, activation_max);
-        } else {
-            esp_nn_depthwise_conv_s16_mult8_esp32s3(input_data16, input_wd, input_ht, channels,
-                                                    pad_wd, pad_ht, stride_wd, stride_ht, ch_mult,
-                                                    filter_data16, filter_wd, filter_ht, bias,
-                                                    out_data, out_wd, out_ht, out_offset, out_shift,
-                                                    out_mult, activation_min, activation_max);
-        }
-    } else if (ch_mult % 4 == 0) {
-        esp_nn_s8_to_s16_esp32s3(filter_data, filter_data16, filter_size);
-        esp_nn_aligned_s8_to_s16_with_offset_esp32s3(input_data, input_data16, input_size, input_offset);
-        esp_nn_depthwise_conv_s16_mult4_esp32s3(input_data16, input_wd, input_ht, channels,
-                                                pad_wd, pad_ht, stride_wd, stride_ht, ch_mult,
-                                                filter_data16, filter_wd, filter_ht, bias,
-                                                out_data, out_wd, out_ht, out_offset, out_shift,
-                                                out_mult, activation_min, activation_max);
-    } else {
-        esp_nn_depthwise_conv_s8_unrolled(input_data, input_wd, input_ht, channels, input_offset,
-                                          pad_wd, pad_ht, stride_wd, stride_ht, ch_mult,
-                                          filter_data, filter_wd, filter_ht,
-                                          bias, out_data, out_wd, out_ht, out_offset, out_shift,
-                                          out_mult, activation_min, activation_max);
-    }
-}

code/components/esp-nn/src/fully_connected/esp_nn_fully_connected_ansi.c

@@ -1,50 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-void esp_nn_fully_connected_s8_ansi(const int8_t *input_data,
-                                    const int32_t input_offset,
-                                    const uint16_t row_len,
-                                    const int8_t *filter_data,
-                                    const int32_t filter_offset,
-                                    const int32_t *bias,
-                                    int8_t *out_data,
-                                    const uint16_t out_channels,
-                                    const int32_t out_offset,
-                                    const int32_t out_shift,
-                                    const int32_t out_mult,
-                                    const int32_t activation_min,
-                                    const int32_t activation_max)
-{
-    for (int32_t out_c = 0; out_c < out_channels; ++out_c) {
-        int32_t result = 0;
-        for (int32_t data_idx = 0; data_idx < row_len; data_idx++) {
-            int32_t filter_index = row_len * out_c + data_idx;
-            int32_t input_val = input_data[data_idx];
-            int32_t filter_val = filter_data[filter_index];
-            result += (filter_val + filter_offset) * (input_val + input_offset);
-        }
-        if (bias) {
-            result += bias[out_c];
-        }
-        result = esp_nn_multiply_by_quantized_mult(result, out_mult, out_shift);
-        result += out_offset;
-        result = max(result, activation_min);
-        result = min(result, activation_max);
-        out_data[out_c] = (int8_t) result;
-    }
-}

code/components/esp-nn/src/pooling/esp_nn_avg_pool_ansi.c

@@ -1,72 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-void esp_nn_avg_pool_s8_ansi(const int8_t *input,
-                             const uint16_t input_wd,
-                             const uint16_t input_ht,
-                             int8_t *output,
-                             const uint16_t output_wd,
-                             const uint16_t output_ht,
-                             const uint16_t stride_wd,
-                             const uint16_t stride_ht,
-                             const uint16_t filter_wd,
-                             const uint16_t filter_ht,
-                             const uint16_t pad_wd,
-                             const uint16_t pad_ht,
-                             const int32_t activation_min,
-                             const int32_t activation_max,
-                             const uint16_t channels)
-{
-    int32_t base_y = -pad_ht;
-    for (int32_t out_y = 0; out_y < output_ht; out_y++, base_y += stride_ht) {
-        int32_t base_x = -pad_wd;
-        for (int32_t out_x = 0; out_x < output_wd; out_x++, base_x += stride_wd) {
-            for (int32_t ch_idx = 0; ch_idx < channels; ch_idx++) {
-                int32_t result = 0;
-                int32_t filter_cnt = 0;
-                /* Make sure filter does not cross the input box */
-                int32_t filter_y_start = max(0, -base_y);
-                int32_t filter_x_start = max(0, -base_x);
-
-                int32_t filter_y_end = min(filter_ht, input_ht - base_y);
-                int32_t filter_x_end = min(filter_wd, input_wd - base_x);
-
-                for (int32_t filter_y = filter_y_start; filter_y < filter_y_end; filter_y++) {
-                    for (int32_t filter_x = filter_x_start; filter_x < filter_x_end; filter_x++) {
-                        int32_t in_x_idx = base_x + filter_x;
-                        int32_t in_y_idx = base_y + filter_y;
-                        int32_t input_index = (in_y_idx * input_wd + in_x_idx) * channels + ch_idx;
-                        result += input[input_index];
-                        filter_cnt++;
-                    }
-                }
-
-                /* Rounded average */
-                result = result > 0 ? (result + filter_cnt / 2) / filter_cnt
-                                    : (result - filter_cnt / 2) / filter_cnt;
-
-                /* Activation function */
-                result = max(result, activation_min);
-                result = min(result, activation_max);
-
-                int32_t output_index = (out_y * output_wd + out_x) * channels + ch_idx;
-                output[output_index] = (int8_t) result;
-            }
-        }
-    }
-}

code/components/esp-nn/src/pooling/esp_nn_max_pool_ansi.c

@@ -1,66 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-
-#include <common_functions.h>
-
-void esp_nn_max_pool_s8_ansi(const int8_t *input,
-                             const uint16_t input_wd,
-                             const uint16_t input_ht,
-                             int8_t *output,
-                             const uint16_t output_wd,
-                             const uint16_t output_ht,
-                             const uint16_t stride_wd,
-                             const uint16_t stride_ht,
-                             const uint16_t filter_wd,
-                             const uint16_t filter_ht,
-                             const uint16_t pad_wd,
-                             const uint16_t pad_ht,
-                             const int32_t activation_min,
-                             const int32_t activation_max,
-                             const uint16_t channels)
-{
-    int32_t base_y = -pad_ht;
-    for (int32_t out_y = 0; out_y < output_ht; out_y++, base_y += stride_ht) {
-        int32_t base_x = -pad_wd;
-        for (int32_t out_x = 0; out_x < output_wd; out_x++, base_x += stride_wd) {
-            /* Make sure filter does not cross the input box */
-            int32_t filter_y_start = max(0, -base_y);
-            int32_t filter_x_start = max(0, -base_x);
-            int32_t filter_y_end = min(filter_ht, input_ht - base_y);
-            int32_t filter_x_end = min(filter_wd, input_wd - base_x);
-
-            for (int32_t ch_idx = 0; ch_idx < channels; ch_idx++) {
-                int8_t result = INT8_MIN;
-
-                for (int32_t filter_y = filter_y_start; filter_y < filter_y_end; filter_y++) {
-                    for (int32_t filter_x = filter_x_start; filter_x < filter_x_end; filter_x++) {
-                        int32_t in_x_idx = base_x + filter_x;
-                        int32_t in_y_idx = base_y + filter_y;
-                        int32_t input_index = (in_y_idx * input_wd + in_x_idx) * channels + ch_idx;
-                        result = max(input[input_index], result);
-                    }
-                }
-
-                /* Activation function */
-                result = max(result, activation_min);
-                result = min(result, activation_max);
-
-                int32_t output_index = (out_y * output_wd + out_x) * channels + ch_idx;
-                output[output_index] = result;
-            }
-        }
-    }
-}

code/components/esp-nn/src/softmax/esp_nn_softmax_ansi.c

@@ -1,88 +0,0 @@
-// Copyright 2022 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include "softmax_common.h"
-
-int32_t esp_nn_get_softmax_scratch_size_ansi(const int32_t width, const int32_t height)
-{
-    (void) width;
-    (void) height;
-    return 0;
-}
-
-void esp_nn_set_softmax_scratch_buf_ansi(void *buffer)
-{
-    (void) buffer;
-    return;
-}
-
-void esp_nn_softmax_s8_ansi(const int8_t *input_data,
-                            const int32_t height,
-                            const int32_t width,
-                            const int32_t mult,
-                            const int32_t shift,
-                            const int32_t diff_min,
-                            int8_t *output_data)
-{
-    // The representation chosen for the input to the exp() function is Q5.26.
-    // We need to leave extra space since values that we skip might be as large as
-    // -32 before multiplying by input mult, and therefore as large as
-    // -16 afterwards.  Note that exp(-8) is definitely not insignificant to
-    // accumulation, but exp(-16) definitely is.
-#define ACCUM_BITS  12
-#define DIFF_BITS   5
-
-    const int32_t mask = (1 << shift);
-    int32_t col = 0;
-    const int8_t *in_ptr = input_data;
-    int8_t *out_ptr = output_data;
-
-    for (int row_idx = 0; row_idx < height; row_idx++) {
-        int8_t max_in_row = in_ptr[0];
-        for (col = 1; col < width; col++) {
-            max_in_row = max(max_in_row, in_ptr[col]);
-        }
-
-        int32_t input_diff = 0;
-        int32_t sum_of_exps = 0;
-
-        for (col = 0; col < width; col++) {
-            input_diff = in_ptr[col] - max_in_row;
-            if (input_diff >= diff_min) {
-                const int32_t input_diff_rescaled = SAT_HIGH_MUL(input_diff * mask, mult);
-                const int32_t exp_raw = esp_nn_exp_on_negative_values(input_diff_rescaled);
-                sum_of_exps += DIV_POW2(exp_raw, ACCUM_BITS);
-            }
-        }
-
-        const int32_t headroom_plus1 = esp_nn_clz32((uint32_t) sum_of_exps);
-        const int32_t shifted_scale = ONE_OVER_ONE_X((sum_of_exps << headroom_plus1) - (1 << 31));
-        const int32_t bits_over_unit = ACCUM_BITS - headroom_plus1 + 31 - sizeof(int8_t) * 8;
-
-        for (col = 0; col < width; col++) {
-            input_diff = in_ptr[col] - max_in_row;
-            if (input_diff >= diff_min) {
-                const int32_t input_diff_rescaled = SAT_HIGH_MUL(input_diff * mask, mult);
-                const int32_t exp_raw = esp_nn_exp_on_negative_values(input_diff_rescaled);
-                const int32_t shifted_output = SAT_HIGH_MUL(shifted_scale, exp_raw);
-                const int32_t result = DIV_POW2(shifted_output, bits_over_unit) - 128;
-                out_ptr[col] = (int8_t) esp_nn_saturate8(result);
-            } else {
-                out_ptr[col] = -128;
-            }
-        }
-        in_ptr  += width;
-        out_ptr += width;
-    }
-}

code/components/esp-nn/src/softmax/esp_nn_softmax_opt.c

@@ -1,108 +0,0 @@
-// Copyright 2022 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include "softmax_common.h"
-#include <stdio.h>
-
-static int32_t *scratch_buf = NULL;
-
-/**
- * @brief   Get scratch buffer size needed by softmax function
- *
- * @param   width
- * @param   height
- * @return  size in bytes
- *
- * @note    buffer must be 4 byte aligned
- */
-int32_t esp_nn_get_softmax_scratch_size_opt(const int32_t width, const int32_t height)
-{
-    (void) height;
-    return width * 4;
-}
-
-/**
- * @brief   Set scratch buffer to be used by softmax function
- *
- * @param   buffer  this can be NULL if one needs to unset it
- *                  must be aligned to 4 bytes
- */
-void esp_nn_set_softmax_scratch_buf_opt(void *buffer)
-{
-    scratch_buf = (int32_t *) buffer;
-}
-
-void esp_nn_softmax_s8_opt(const int8_t *input_data,
-                           const int32_t height,
-                           const int32_t width,
-                           const int32_t mult,
-                           const int32_t shift,
-                           const int32_t diff_min,
-                           int8_t *output_data)
-{
-    if (scratch_buf == NULL) {
-        printf("%s error! scratch buffer not set\n", __FUNCTION__);
-        return;
-    }
-    // The representation chosen for the input to the exp() function is Q5.26.
-    // We need to leave extra space since values that we skip might be as large as
-    // -32 before multiplying by input mult, and therefore as large as
-    // -16 afterwards.  Note that exp(-8) is definitely not insignificant to
-    // accumulation, but exp(-16) definitely is.
-#define ACCUM_BITS  12
-#define DIFF_BITS   5
-
-    const int32_t mask = (1 << shift);
-    int32_t col = 0;
-    const int8_t *in_ptr = input_data;
-    int8_t *out_ptr = output_data;
-
-    for (int row_idx = 0; row_idx < height; row_idx++) {
-        int8_t max_in_row = in_ptr[0];
-        for (col = 1; col < width; col++) {
-            max_in_row = max(max_in_row, in_ptr[col]);
-        }
-
-        int32_t input_diff = 0;
-        int32_t sum_of_exps = 0;
-
-        for (col = 0; col < width; col++) {
-            input_diff = in_ptr[col] - max_in_row;
-            if (input_diff >= diff_min) {
-                const int32_t input_diff_rescaled = SAT_HIGH_MUL(input_diff * mask, mult);
-                const int32_t exp_raw = esp_nn_exp_on_negative_values(input_diff_rescaled);
-                scratch_buf[col] = exp_raw; // store to avoid duplicate calculation later
-                sum_of_exps += DIV_POW2(exp_raw, ACCUM_BITS);
-            }
-        }
-
-        const int32_t headroom_plus1 = esp_nn_clz32((uint32_t) sum_of_exps);
-        const int32_t shifted_scale = ONE_OVER_ONE_X((sum_of_exps << headroom_plus1) - (1 << 31));
-        const int32_t bits_over_unit = ACCUM_BITS - headroom_plus1 + 31 - sizeof(int8_t) * 8;
-
-        for (col = 0; col < width; col++) {
-            input_diff = in_ptr[col] - max_in_row;
-            if (input_diff >= diff_min) {
-                int32_t exp_raw = scratch_buf[col];
-                const int32_t shifted_output = SAT_HIGH_MUL(shifted_scale, exp_raw);
-                const int32_t result = DIV_POW2(shifted_output, bits_over_unit) - 128;
-                out_ptr[col] = (int8_t) esp_nn_saturate8(result);
-            } else {
-                out_ptr[col] = -128;
-            }
-        }
-        in_ptr  += width;
-        out_ptr += width;
-    }
-}

code/components/esp-nn/src/softmax/softmax_common.h

@@ -1,104 +0,0 @@
-// Copyright 2022 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <common_functions.h>
-
-#define MASK_IF_ZERO(x)                 (x) == 0 ? ~0 : 0
-#define MASK_IF_NON_ZERO(x)             (x) != 0 ? ~0 : 0
-#define SELECT_USING_MASK(mask, a, b)   ((mask) & (a)) ^ (~(mask) & (b))
-#define SAT_HIGH_MUL(x, y)              esp_nn_sat_round_doubling_high_mul((x), (y))
-#define DIV_POW2(x,y)                   esp_nn_div_by_power_of_two((x), (y))
-
-__NN_FORCE_INLINE__ int32_t mul_power_of_2(int val, int exp)
-{
-    const int32_t thresh = ((1 << (31 - exp)) - 1);
-    int32_t result = val << exp;
-    result = SELECT_USING_MASK(MASK_IF_NON_ZERO(val > thresh), INT32_MAX, result);
-    result = SELECT_USING_MASK(MASK_IF_NON_ZERO(val < -thresh), INT32_MIN, result);
-    return result;
-}
-
-/**
- * @brief   Calculate `1 / (1 + x)` for x in [0, 1]
- *
- * @param   val     input value to calculate `1/(1+x)` for
- * @return  `int32_t` result
- * @note    Newton-Raphson division
- *
- *          https://en.wikipedia.org/wiki/Division_algorithm#Newton.E2.80.93Raphson_division
- *          Refer to that page for the logic behind the 48/17 and 32/17 constants.
- *          Pseudocode: https://en.wikipedia.org/wiki/Division_algorithm#Pseudocode
- */
-__NN_FORCE_INLINE__ int32_t esp_nn_one_over_one_plus_x_for_x_in_0_1(int32_t val)
-{
-    const int64_t sum = (int64_t) val + INT32_MAX;
-    const int32_t half_denominator = (int32_t) ((sum + (sum >= 0 ? 1 : -1)) / 2L);
-    int32_t constant_48_over_17 = 1515870810;
-    int32_t constant_neg_32_over_17 = -1010580540;
-    int32_t x = constant_48_over_17 + SAT_HIGH_MUL(half_denominator, constant_neg_32_over_17);
-    const int32_t fixed_2_one = (1 << 29);
-
-    x += mul_power_of_2(SAT_HIGH_MUL(x, fixed_2_one - SAT_HIGH_MUL(half_denominator, x)), 2);
-    x += mul_power_of_2(SAT_HIGH_MUL(x, fixed_2_one - SAT_HIGH_MUL(half_denominator, x)), 2);
-    x += mul_power_of_2(SAT_HIGH_MUL(x, fixed_2_one - SAT_HIGH_MUL(half_denominator, x)), 2);
-
-    return mul_power_of_2(x, 1);
-}
-
-#define ONE_OVER_ONE_X(x)   esp_nn_one_over_one_plus_x_for_x_in_0_1((x))
-
-/**
- * @brief   Return exp(x) for x < 0.
- *
- */
-__NN_FORCE_INLINE__ int32_t esp_nn_exp_on_negative_values(int32_t val)
-{
-    int32_t shift = 24;
-
-    const int32_t one_quarter = (1 << shift);
-    int32_t mask = one_quarter - 1;
-    const int32_t val_mod_minus_quarter = (val & mask) - one_quarter;
-    const int32_t remainder             = val_mod_minus_quarter - val;
-
-    // calculate exponent for x in [-1/4, 0) in `result`
-    const int32_t x                     = (val_mod_minus_quarter << 5) + (1 << 28);
-    const int32_t x2                    = SAT_HIGH_MUL(x, x);
-    const int32_t x3                    = SAT_HIGH_MUL(x2, x);
-    const int32_t x4                    = SAT_HIGH_MUL(x2, x2);
-    const int32_t one_over_3            = 715827883;
-    const int32_t one_over_8            = 1895147668;
-
-    const int32_t x4_over_4 = DIV_POW2(x4, 2);
-    const int32_t x4_over_4_plus_x3_over_6_plus_x2_over_2 = DIV_POW2(SAT_HIGH_MUL(x4_over_4 + x3, one_over_3) + x2, 1);
-    int32_t result = one_over_8 + SAT_HIGH_MUL(one_over_8, x + x4_over_4_plus_x3_over_6_plus_x2_over_2);
-
-#define SELECT_IF_NON_ZERO(x) {                                   \
-    mask   = MASK_IF_NON_ZERO(remainder & (1 << shift++));        \
-    result = SELECT_USING_MASK(mask, SAT_HIGH_MUL(result, x), result); \
-}
-
-    SELECT_IF_NON_ZERO(1672461947)
-    SELECT_IF_NON_ZERO(1302514674)
-    SELECT_IF_NON_ZERO(790015084)
-    SELECT_IF_NON_ZERO(290630308)
-    SELECT_IF_NON_ZERO(39332535)
-    SELECT_IF_NON_ZERO(720401)
-    SELECT_IF_NON_ZERO(242)
-
-#undef SELECT_IF_NON_ZERO
-
-    mask = MASK_IF_ZERO(val);
-    return SELECT_USING_MASK(mask, INT32_MAX, result);
-}

code/components/esp-nn/test_app/CMakeLists.txt

@@ -1,9 +0,0 @@
-# The following lines of boilerplate have to be in your project's
-# CMakeLists in this exact order for cmake to work correctly
-cmake_minimum_required(VERSION 3.5)
-
-set(EXTRA_COMPONENT_DIRS "../" "../tests/")
-set(IDF_EXCLUDE_COMPONENTS test test_app)
-
-include($ENV{IDF_PATH}/tools/cmake/project.cmake)
-project(test_app)

code/components/esp-nn/test_app/main/CMakeLists.txt

@@ -1,7 +0,0 @@
-
-set(COMPONENT_SRCS "main.c")
-set(COMPONENT_ADD_INCLUDEDIRS "")
-
-set(COMPONENT_PRIV_REQUIRES tests)
-
-register_component()

code/components/esp-nn/test_app/main/component.mk

@@ -1,8 +0,0 @@
-#
-# Main component makefile.
-#
-# This Makefile can be left empty. By default, it will take the sources in the 
-# src/ directory, compile them and link them into lib(subdirectory_name).a
-# in the build directory. This behaviour is entirely configurable,
-# please read the ESP-IDF documents if you need to do this.
-# 

code/components/esp-nn/test_app/main/main.c

@@ -1,87 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <freertos/FreeRTOS.h>
-#include <freertos/task.h>
-#include <esp_log.h>
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include <test_functions.h>
-#include <esp_timer.h>
-
-static const char *TAG = "test_app";
-static uint32_t start_c, start_opt, total_c, total_opt;
-
-void profile_c_start()
-{
-    /* initiate profiling */
-    start_c = esp_cpu_get_ccount();
-}
-
-void profile_c_end()
-{
-    /* record profile number */
-    total_c = esp_cpu_get_ccount() - start_c;
-}
-
-void profile_opt_start()
-{
-    /* initiate profiling */
-    start_opt = esp_cpu_get_ccount();
-}
-
-void profile_opt_end()
-{
-    /* record profile number */
-    total_opt = esp_cpu_get_ccount() - start_opt;
-}
-
-void app_main()
-{
-    /* s8 tests */
-    ESP_LOGI(TAG, "Running s8 tests...");
-    esp_nn_add_elementwise_s8_test();
-    printf("add, c %u opt %u\n", total_c, total_opt);
-    esp_nn_mul_elementwise_s8_test();
-    printf("mul, c %u opt %u\n", total_c, total_opt);
-    esp_nn_depthwise_conv_s8_test();
-    printf("depthwise, c %u opt %u\n", total_c, total_opt);
-    esp_nn_conv_s8_test();
-    printf("conv2d, c %u opt %u\n", total_c, total_opt);
-
-    esp_nn_relu6_s8_test();
-    printf("relu, c %u opt %u\n", total_c, total_opt);
-    esp_nn_avg_pool_s8_test();
-    printf("avg_pool, c %u opt %u\n", total_c, total_opt);
-    esp_nn_max_pool_s8_test();
-    printf("max_pool, c %u opt %u\n", total_c, total_opt);
-    esp_nn_fully_connected_s8_test();
-    printf("fully_connected, c %u opt %u\n", total_c, total_opt);
-    esp_nn_softmax_s8_test();
-    printf("softmax, c %u opt %u\n", total_c, total_opt);
-    ESP_LOGI(TAG, "s8 tests done!\n");
-
-    /* u8 tests */
-    //ESP_LOGI(TAG, "Running u8 tests...");
-    //esp_nn_add_elementwise_u8_test();
-    //esp_nn_depthwise_conv_u8_test();
-    //esp_nn_conv_u8_test();
-    //esp_nn_avg_pool_u8_test();
-    //esp_nn_max_pool_u8_test();
-    //esp_nn_fully_connected_u8_test();
-    //ESP_LOGI(TAG, "u8 tests done!\n");
-}

code/components/esp-nn/test_app/sdkconfig.defaults

@@ -1,5 +0,0 @@
-
-#
-# esp-nn
-#
-CONFIG_NN_ESP32=y

code/components/esp-nn/tests/CMakeLists.txt

@@ -1,15 +0,0 @@
-
-set(COMPONENT_ADD_INCLUDEDIRS ./include/)
-set(COMPONENT_SRCS "src/basic_math_test.c"
-                   "src/convolution_test.c"
-                   "src/fully_connected_test.c"
-                   "src/pooling_test.c"
-                   "src/relu_test.c"
-                   "src/softmax_test.c")
-
-set(COMPONENT_REQUIRES )
-set(COMPONENT_PRIV_REQUIRES esp-nn)
-
-register_component()
-
-target_compile_options(${COMPONENT_LIB} PRIVATE -Wno-unused-function)

code/components/esp-nn/tests/README.md

@@ -1,4 +0,0 @@
-# Tests for esp_nn library
-
-- Include these in your test framework and run the framework.
-- For IDF test please refer `test_app`

code/components/esp-nn/tests/component.mk

@@ -1,5 +0,0 @@
-#FIXME
-
-COMPONENT_ADD_INCLUDEDIRS := include/
-
-COMPONENT_SRCDIRS :=  src/

code/components/esp-nn/tests/include/test_functions.h

@@ -1,48 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-
-/* int8_t ops tests */
-void esp_nn_add_elementwise_s8_test();
-void esp_nn_mul_elementwise_s8_test();
-
-void esp_nn_depthwise_conv_s8_test();
-void esp_nn_conv_s8_test();
-
-void esp_nn_avg_pool_s8_test();
-void esp_nn_max_pool_s8_test();
-
-void esp_nn_fully_connected_s8_test();
-
-void esp_nn_relu6_s8_test();
-
-void esp_nn_softmax_s8_test();
-
-/* uint8_t ops tests */
-void esp_nn_add_elementwise_u8_test();
-
-void esp_nn_depthwise_conv_u8_test();
-void esp_nn_conv_u8_test();
-
-void esp_nn_avg_pool_u8_test();
-void esp_nn_max_pool_u8_test();
-
-void esp_nn_fully_connected_u8_test();
-
-/* instructions test functions */
-void compare_instructions_test();
-void arith_instructions_test();
-void min_max_instructions_test();
-void bitwise_instructions_test();
-void load_store_instructions_test();

code/components/esp-nn/tests/include/test_utils.h

@@ -1,87 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <common_functions.h>
-#include <stdio.h>
-
-/* mult value range */
-#define MULT_MAX    INT32_MAX
-#define MULT_MIN    0
-
-/* shift value range */
-#define SHIFT_MIN   -31
-#define SHIFT_MAX   30
-
-/**
- * @brief callback function to run before C function
- */
-void profile_c_start();
-
-/**
- * @brief callback function to run after C function
- */
-void profile_c_end();
-
-/**
- * @brief callback function to run before optimized function
- */
-void profile_opt_start();
-
-/**
- * @brief callback function to run after optimized function
- */
-void profile_opt_end();
-
-#define ANSI_COLOR_RED     "\x1b[31m"
-#define ANSI_COLOR_GREEN   "\x1b[32m"
-#define ANSI_COLOR_YELLOW  "\x1b[33m"
-#define ANSI_COLOR_BLUE    "\x1b[34m"
-#define ANSI_COLOR_MAGENTA "\x1b[35m"
-#define ANSI_COLOR_CYAN    "\x1b[36m"
-#define ANSI_COLOR_RESET   "\x1b[0m"
-
-#define CHECK_EQUAL(ARRAY1, ARRAY2, size) ({    \
-    bool res = true;                            \
-    for (int _i = 0; _i < size; _i++) {         \
-        if (ARRAY1[_i] != ARRAY2[_i]) {         \
-            res = false;                        \
-            break;                              \
-        }                                       \
-    }                                           \
-    res;                                        \
-})
-
-#define PRINT_ARRAY_INT(ARRAY, width, height) ({        \
-    int *_array = (int *) ARRAY;                        \
-    for (int _j = 0; _j < height; _j++) {               \
-        for (int _i = 0; _i < width; _i++) {            \
-            printf("%d\t", _array[width * _j + _i]);    \
-        }                                               \
-        printf("\n");                                   \
-    }                                                   \
-    printf("\n");                                       \
-})
-
-#define PRINT_ARRAY_HEX(ARRAY, width, height) ({        \
-    uint8_t *_array = (uint8_t *) ARRAY;                \
-    for (int _j = 0; _j < height; _j++) {               \
-        for (int _i = 0; _i < width; _i++) {            \
-            printf("%02x\t", _array[width * _j + _i]);  \
-        }                                               \
-        printf("\n");                                   \
-    }                                                   \
-    printf("\n");                                       \
-})

code/components/esp-nn/tests/src/basic_math_test.c

@@ -1,343 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <malloc.h>
-
-#include <common_functions.h>
-#include <esp_nn.h>
-#include "test_utils.h"
-
-#if CONFIG_IDF_CMAKE
-#define IDF_HEAP_CAPS 1
-
-#if IDF_HEAP_CAPS
-#include "esp_heap_caps.h"
-#endif
-#endif
-
-void esp_nn_add_elementwise_s8_test()
-{
-    /* prepare data */
-    const int size = 1600 + 8 + 7; /* odd len to test leftover */
-    int8_t *input1;
-    int8_t *input2;
-    int8_t *out_data_c;
-    int8_t *out_data_opt;
-    int8_t *input1_orig = NULL;
-    int8_t *input2_orig = NULL;
-    int8_t *out_c_orig = NULL;
-    int8_t *out_opt_orig = NULL;
-    int32_t input1_offset = 34;
-    int32_t input2_offset = 35;
-    int32_t output_offset = 36;
-    int32_t input1_shift = -8; // right_shift amt always <= 0
-    int32_t input2_shift = -8; // right_shift amt always <= 0
-    int32_t output_shift = -9; // right_shift amt always <= 0
-    int32_t left_shift = 15; // always +ve
-    int32_t input1_mult = INT32_MAX;
-    int32_t input2_mult = INT32_MAX;
-    int32_t output_mult = INT32_MAX;
-    int32_t activation_min = -128;
-    int32_t activation_max = 127;
-
-    for (int itr = 0; itr < 10; itr++) {
-        switch (itr) {
-        case 0: // all zeros
-            input1_offset = 0;
-            input2_offset = 0;
-            output_offset = 0;
-            input1_mult = 0;
-            input2_mult = 0;
-            output_mult = 0;
-            input1_shift = 0;
-            input2_shift = 0;
-            output_shift = 0;
-            left_shift = 0;
-        break;
-        case 1: // hit min
-            input1_offset = -127;
-            input2_offset = -127;
-            output_offset = -128;
-            input1_mult = MULT_MIN;
-            input2_mult = MULT_MIN;
-            output_mult = MULT_MIN;
-            input1_shift = 0;
-            input2_shift = 0;
-            output_shift = 0;
-            left_shift = 0;
-        break;
-        case 2: // hit max
-            input1_offset = 128;
-            input2_offset = 128;
-            output_offset = -127;
-            input1_mult = MULT_MAX;
-            input2_mult = MULT_MAX;
-            output_mult = MULT_MAX;
-            input1_shift = SHIFT_MIN;
-            input2_shift = SHIFT_MIN;
-            output_shift = SHIFT_MIN;
-            left_shift = 30 - 8; // since input is 8 bits
-        break;
-        case 3: // hit extreme max
-            input1_offset = 128;
-            input2_offset = 128;
-            output_offset = -127;
-            input1_mult = MULT_MAX;
-            input2_mult = MULT_MAX;
-            output_mult = MULT_MAX;
-            input1_shift = 0;
-            input2_shift = 0;
-            output_shift = 0;
-            left_shift = 30 - 8; // -8 since input is 8 bit
-        break;
-        default:  // practical random input
-            input1_offset = rand() % 256 - 127; // range [-127, 128]
-            input2_offset = rand() % 256 - 127; // range [-127, 128]
-            output_offset = rand() % 256 - 128; // range [-128, 127]
-            input1_mult = MULT_MAX / 2 + rand() % INT16_MAX;
-            input2_mult = MULT_MAX / 2 + rand() % INT16_MAX;
-            output_mult = MULT_MAX / 2 + rand() % INT16_MAX;
-            input1_shift = -8 + rand() % 4;
-            input2_shift = -8 + rand() % 4;
-            output_shift = -8 + rand() % 4;
-            left_shift = rand() % 15;
-        }
-#if IDF_HEAP_CAPS
-        input1_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        input2_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        out_c_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        out_opt_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-
-        input1 = 16 + input1_orig - ((uint32_t) input1_orig & 0xf);
-        input2 = 16 + input2_orig - ((uint32_t) input2_orig & 0xf);
-        out_data_c = 16 + out_c_orig - ((uint32_t) out_c_orig & 0xf);
-        out_data_opt = 16 + out_opt_orig - ((uint32_t) out_opt_orig & 0xf);
-#else
-        input1 = memalign(16, size);
-        input2 = memalign(16, size);
-        out_data_c = memalign(16, size);
-        out_data_opt = memalign(16, size);
-
-        input1_orig = input1;
-        input2_orig = input2;
-        out_c_orig = out_data_c;
-        out_opt_orig = out_data_opt;
-#endif
-
-        for (int i = 0; i < size; ++i) {
-            input1[i] = rand() % 256 - 128;
-            input2[i] = rand() % 256 - 128;
-        }
-
-        if (itr == 0) {
-            /* enable profiler */
-            profile_c_start();
-        }
-        /* C function */
-        esp_nn_add_elementwise_s8_ansi(input1, input2, input1_offset, input2_offset,
-                                       input1_mult, input2_mult, input1_shift, input2_shift,
-                                       left_shift, out_data_c, output_offset, output_mult,
-                                       output_shift, activation_min, activation_max, size);
-
-        if (itr == 0) {
-            profile_c_end();
-            profile_opt_start();
-        }
-
-        /* Optimized function */
-        esp_nn_add_elementwise_s8(input1, input2, input1_offset, input2_offset,
-                                  input1_mult, input2_mult, input1_shift, input2_shift,
-                                  left_shift, out_data_opt, output_offset, output_mult,
-                                  output_shift, activation_min, activation_max, size);
-        if (itr == 0) {
-            /* disable profiler */
-            profile_opt_end();
-        }
-
-        bool ret = CHECK_EQUAL(out_data_c, out_data_opt, size);
-        if (ret == false) {
-            printf(ANSI_COLOR_RED"%s[%d] failed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-            printf("Output: \n");
-            PRINT_ARRAY_HEX(out_data_opt, size, 1);
-            printf("Expected: \n");
-            PRINT_ARRAY_HEX(out_data_c, size, 1);
-            printf("Input1:\n");
-            PRINT_ARRAY_HEX(input1, size, 1);
-            printf("Input2:\n");
-            PRINT_ARRAY_HEX(input2, size, 1);
-            printf("in1_shift %d, in2_shift %d, left_shift %d, out_shift %d\n",
-                   input1_shift, input2_shift, left_shift, output_shift);
-            printf("in1_mult %d, in2_mult %d, out_mult %d\n", input1_mult, input2_mult, output_mult);
-            goto elementwise_add_test_cleanup;
-        }
-        printf(ANSI_COLOR_GREEN"%s[%d] passed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-
-elementwise_add_test_cleanup:
-        if (input1_orig) {
-            free(input1_orig);
-        }
-        if (input2_orig) {
-            free(input2_orig);
-        }
-        if (out_data_c) {
-            free(out_c_orig);
-        }
-        if (out_data_opt) {
-            free(out_opt_orig);
-        }
-    }
-}
-
-void esp_nn_mul_elementwise_s8_test()
-{
-    /* prepare data */
-    const int size = 1600 + 8 + 7; /* odd len to test leftover */
-    int8_t *input1;
-    int8_t *input2;
-    int8_t *out_data_c;
-    int8_t *out_data_opt;
-    int32_t input1_offset = 34;
-    int32_t input2_offset = 35;
-    int32_t output_offset = 36;
-    int32_t output_shift = -7;
-    int32_t output_mult = MULT_MAX; // max out_mult
-    int32_t activation_min = -128;
-    int32_t activation_max = 127;
-    int8_t *input1_orig = NULL;
-    int8_t *input2_orig = NULL;
-    int8_t *out_c_orig = NULL;
-    int8_t *out_opt_orig = NULL;
-
-    for (int itr = 0; itr < 10; itr++) {
-        switch (itr) {
-        case 0: // all zeros
-            input1_offset = 0;
-            input2_offset = 0;
-            output_offset = 0;
-            output_mult = 0;
-            output_shift = 0;
-        break;
-        case 1: // hit min
-            input1_offset = -127;
-            input2_offset = -127;
-            output_offset = -128;
-            output_mult = MULT_MIN;
-            output_shift = 0;
-        break;
-        case 2: // hit max
-            input1_offset = 128;
-            input2_offset = 128;
-            output_offset = -127;
-            output_mult = MULT_MAX;
-            output_shift = SHIFT_MIN;
-        break;
-        case 3: // hit extreme max
-            input1_offset = 128;
-            input2_offset = 128;
-            output_offset = -127;
-            output_mult = MULT_MAX;
-            output_shift = 0;
-        break;
-        default:  // practical random input
-            input1_offset = rand() % 256 - 127; // range [-127, 128]
-            input2_offset = rand() % 256 - 127; // range [-127, 128]
-            output_offset = rand() % 256 - 128; // range [-128, 127]
-            output_mult = MULT_MAX / 2 + rand() % INT16_MAX;
-            output_shift = -8 + rand() % 4;
-        }
-
-#if IDF_HEAP_CAPS
-        input1_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        input2_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        out_c_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        out_opt_orig = (int8_t *) heap_caps_malloc(size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-
-        input1 = 16 + input1_orig - ((uint32_t) input1_orig & 0xf);
-        input2 = 16 + input2_orig - ((uint32_t) input2_orig & 0xf);
-        out_data_c = 16 + out_c_orig - ((uint32_t) out_c_orig & 0xf);
-        out_data_opt = 16 + out_opt_orig - ((uint32_t) out_opt_orig & 0xf);
-#else
-        input1 = memalign(16, size);
-        input2 = memalign(16, size);
-        out_data_c = memalign(16, size);
-        out_data_opt = memalign(16, size);
-
-        input1_orig = input1;
-        input2_orig = input2;
-        out_c_orig = out_data_c;
-        out_opt_orig = out_data_opt;
-#endif
-
-        for (int i = 0; i < size; ++i) {
-            input1[i] = rand() % 256 - 128;
-            input2[i] = rand() % 256 - 128;
-        }
-
-        if (itr == 0) {
-            /* enable profiler */
-            profile_c_start();
-        }
-        /* C function */
-        esp_nn_mul_elementwise_s8_ansi(input1, input2, input1_offset, input2_offset,
-                                       out_data_c, output_offset, output_mult, output_shift,
-                                       activation_min, activation_max, size);
-
-        if (itr == 0) {
-            profile_c_end();
-            profile_opt_start();
-        }
-        /* Optimized function */
-        esp_nn_mul_elementwise_s8(input1, input2, input1_offset, input2_offset,
-                                  out_data_opt, output_offset, output_mult, output_shift,
-                                  activation_min, activation_max, size);
-
-        if (itr == 0) {
-            /* disable profiler */
-            profile_opt_end();
-        }
-
-        bool ret = CHECK_EQUAL(out_data_c, out_data_opt, size);
-        if (ret == false) {
-            printf(ANSI_COLOR_RED"%s[%d] failed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-            printf("Output: \n");
-            PRINT_ARRAY_HEX(out_data_opt, size, 1);
-            printf("Expected: \n");
-            PRINT_ARRAY_HEX(out_data_c, size, 1);
-            printf("Input1:\n");
-            PRINT_ARRAY_HEX(input1, size, 1);
-            printf("Input2:\n");
-            PRINT_ARRAY_HEX(input2, size, 1);
-            goto elementwise_mult_test_cleanup;
-        }
-        printf(ANSI_COLOR_GREEN"%s[%d] passed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-
-elementwise_mult_test_cleanup:
-        if (input1_orig) {
-            free(input1_orig);
-        }
-        if (input2_orig) {
-            free(input2_orig);
-        }
-        if (out_data_c) {
-            free(out_c_orig);
-        }
-        if (out_data_opt) {
-            free(out_opt_orig);
-        }
-    }
-}

code/components/esp-nn/tests/src/convolution_test.c

@@ -1,571 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <malloc.h>
-
-#include <esp_nn.h>
-#include "test_utils.h"
-
-#if CONFIG_IDF_CMAKE
-#define IDF_HEAP_CAPS 1
-
-#if IDF_HEAP_CAPS
-#include "esp_heap_caps.h"
-#endif
-#endif
-
-void esp_nn_depthwise_conv_s8_test()
-{
-    int8_t *input = NULL, *filter_data = NULL, *out_data_c = NULL, *out_data_opt = NULL;
-    int32_t *bias = NULL;
-    int32_t input_offset = 5; /* some number in [-128, 127] */
-    int32_t out_offset = 7;
-    int32_t activation_min = -125;
-    int32_t activation_max = 120;
-    void *scratch_buf = NULL;
-
-    /* independent variables */
-    int input_wd, input_ht, channels;
-    uint16_t filter_ht, filter_wd, ch_mult;
-    uint16_t pad_wd, pad_ht, stride_wd, stride_ht;
-
-    // run for 10 iterations
-    for (int itr = 0; itr < 10; itr++) {
-        /* prepare data */
-        switch (itr) {
-        case 0: // (ch_mult 1, (channels % 16) = 0), filter (3,3), pad (0,0)
-            input_wd = 18;
-            input_ht = 18;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 1;
-            channels = 16;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 1: // (ch_mult 1, (channels % 16) = 0), filter (3,3), pad (1,1)
-            input_wd = 10;
-            input_ht = 10;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 1;
-            channels = 16;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 2: // (ch_mult 1, (channels % 8) = 0), filter (3,3), pad (1,1)
-            input_wd = 10;
-            input_ht = 10;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 1;
-            channels = 24;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 3: // other filter sizes (ch_mult 1, (channels % 8) = 0)
-            input_wd = 10;
-            input_ht = 10;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 1;
-            channels = 24;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 4: // other filter sizes (ch_mult 8 = 0)
-            input_wd = 6;
-            input_ht = 6;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 8;
-            channels = 4;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 5: // other filter sizes (ch_mult 8 = 0)
-            input_wd = 12;
-            input_ht = 12;
-            filter_ht = 5;
-            filter_wd = 5;
-            ch_mult = 8;
-            channels = 4;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 6: // other filter sizes (ch_mult 4 = 0)
-            input_wd = 6;
-            input_ht = 6;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 4;
-            channels = 4;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 7: // (ch_mult 1, (channels % 16) = 0), filter (3,3), pad (0,0)  stride (2,2)
-            input_wd = 6;
-            input_ht = 6;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 1;
-            channels = 16;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 2;
-            stride_ht = 2;
-            break;
-        default:
-            input_wd = 4;
-            input_ht = 4;
-            filter_ht = 3;
-            filter_wd = 3;
-            ch_mult = 4;
-            channels = 4;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        }
-
-        uint16_t out_wd = (input_wd - filter_wd + 1) / stride_wd;
-        uint16_t out_ht = (input_ht - filter_ht + 1) / stride_ht;
-        int in_size = input_wd * input_ht * channels;
-        int out_size = out_wd * out_ht * channels * ch_mult;
-        int filter_size = filter_wd * filter_ht * channels * ch_mult + 4;
-        int bias_size = channels * ch_mult + 1;
-        int32_t out_shift[channels * ch_mult];
-        int32_t out_mult[channels * ch_mult];
-
-#if IDF_HEAP_CAPS
-        int8_t *input_orig = (int8_t *) heap_caps_malloc(in_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        int8_t *out_c_orig = (int8_t *) heap_caps_malloc(out_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        int8_t *out_opt_orig = (int8_t *) heap_caps_malloc(out_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        filter_data = (int8_t *) heap_caps_malloc(filter_size, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        bias = (int32_t *) heap_caps_malloc(bias_size * 4, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-
-        input = 16 + input_orig - ((uint32_t) input_orig & 0xf);
-        out_data_c = 16 + out_c_orig - ((uint32_t) out_c_orig & 0xf);
-        out_data_opt = 16 + out_opt_orig - ((uint32_t) out_opt_orig & 0xf);
-#else
-        input = memalign(16, in_size + 16);
-        filter_data = memalign(16, filter_size);
-        out_data_c = memalign(16, out_size + 16);
-        out_data_opt = memalign(16, out_size + 16);
-        bias = memalign(16, bias_size * 4);
-        int8_t *input_orig = input;
-        int8_t *out_c_orig = out_data_c;
-        int8_t *out_opt_orig = out_data_opt;
-#endif
-        if (bias == NULL || input == NULL || filter_data == NULL ||
-                out_data_c == NULL || out_data_opt == NULL || bias == NULL) {
-            printf(ANSI_COLOR_RED"%s[%d] allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-            goto dc_s8_cleanup;
-        }
-
-        /* Generate input data */
-        for (int i = 0; i < in_size; ++i) {
-            input[i] = rand() % 128;
-        }
-
-        /* Generate filter data */
-        for (int i = 0; i < filter_size; ++i) {
-            filter_data[i] = rand() % 256 - 128;
-        }
-
-        /* Generate bias data */
-        for (int i = 0; i < channels * ch_mult; ++i) {
-            bias[i + 1] = rand() % INT16_MAX; //0th index left for unalignment
-            out_shift[i] = -8 + rand() % 3;
-            out_mult[i] = 0x7eb0e200 + rand() % 50;
-        }
-
-        int scratch_buf_size = esp_nn_get_depthwise_conv_scratch_size(input_wd, input_ht,
-                                                                    channels, ch_mult,
-                                                                    filter_wd, filter_ht);
-        if (scratch_buf_size > 0) {
-#if IDF_HEAP_CAPS
-            scratch_buf = heap_caps_malloc(scratch_buf_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-            int align_sz = 16 - (((int32_t) scratch_buf) & 0xf);
-#else
-            scratch_buf = memalign(16, scratch_buf_size);
-            int align_sz = 0;
-#endif
-            if (scratch_buf == NULL) {
-                printf(ANSI_COLOR_RED"%s[%d] scratch_buf alloc failed size %d\n"ANSI_COLOR_RESET,
-                       __FUNCTION__, itr, scratch_buf_size);
-                goto dc_s8_cleanup;
-            }
-            esp_nn_set_depthwise_conv_scratch_buf(scratch_buf + align_sz);
-        }
-        if (itr == 0) {
-            /* enable profiler */
-            profile_c_start();
-        }
-
-        /* C function */
-        esp_nn_depthwise_conv_s8_ansi(input, input_wd, input_ht, channels, input_offset,
-                                    pad_wd, pad_ht, stride_wd, stride_ht, ch_mult,
-                                    filter_data + 4, filter_wd, filter_ht,
-                                    bias + 1, out_data_c, out_wd, out_ht, out_offset, out_shift,
-                                    out_mult, activation_min, activation_max);
-
-        if (itr == 0) {
-            profile_c_end();
-            profile_opt_start();
-        }
-
-        /* Optimized function */
-        esp_nn_depthwise_conv_s8(input, input_wd, input_ht, channels, input_offset,
-                                pad_wd, pad_ht, stride_wd, stride_ht, ch_mult,
-                                filter_data + 4, filter_wd, filter_ht,
-                                bias + 1, out_data_opt, out_wd, out_ht, out_offset, out_shift,
-                                out_mult, activation_min, activation_max);
-
-        if (itr == 0) {
-            /* disable profiler */
-            profile_opt_end();
-        }
-
-        bool ret = CHECK_EQUAL(out_data_c, out_data_opt, out_size);
-        if (ret == false) {
-            printf(ANSI_COLOR_RED"%s[%d] failed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-            printf("Output: \n");
-            PRINT_ARRAY_HEX(out_data_opt, out_size / out_ht, out_ht);
-            printf("Expected: \n");
-            PRINT_ARRAY_HEX(out_data_c, out_size / out_ht, out_ht);
-            printf("Input:\n");
-            PRINT_ARRAY_HEX(input, in_size / input_ht, input_ht);
-            printf("Filter data:\n");
-            PRINT_ARRAY_HEX(filter_data + 4, (filter_size - 4) / filter_ht, filter_ht);
-            printf("bias data:\n");
-            PRINT_ARRAY_INT(bias + 1, ch_mult * channels, 1);
-            goto dc_s8_cleanup;
-        }
-        printf(ANSI_COLOR_GREEN"%s[%d] passed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-
-    dc_s8_cleanup:
-        if (input) {
-            free(input_orig);
-        }
-        if (filter_data) {
-            free(filter_data);
-        }
-        if (out_data_c) {
-            free(out_c_orig);
-        }
-        if (out_data_opt) {
-            free(out_opt_orig);
-        }
-        if (bias) {
-            free(bias);
-        }
-        if (scratch_buf) {
-            free(scratch_buf);
-        }
-    }
-}
-
-void esp_nn_conv_s8_test()
-{
-    const int32_t input_offset = 5; /* some number in [-128, 127] */
-    const int32_t activation_min = -125;
-    const int32_t activation_max = 122;
-    const int32_t out_offset = 3;
-
-    void *scratch_buf = NULL;
-    int8_t *input_orig;
-    int8_t *out_c_orig;
-    int8_t *out_opt_orig;
-    int8_t *filter_data;
-    int32_t *bias;
-
-    /* independent variable */
-    int in_wd, in_ht, in_channels, out_channels;
-    uint16_t filter_ht, filter_wd;
-    uint16_t pad_wd, pad_ht, stride_wd, stride_ht;
-
-    // run for 10 iterations
-    for (int itr = 0; itr < 10; itr++) {
-        switch (itr) {
-        case 0: // ch % 8 == 0 && filter (1,1), padding (0,0)
-            in_wd = 10;
-            in_ht = 10;
-            in_channels = 64;
-            out_channels = 64;
-            filter_ht = 1;
-            filter_wd = 1;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 1: // ch % 4 == 0 && (in_wd * in_ht) % 16 == 0
-            in_wd = 4;
-            in_ht = 4;
-            in_channels = 20;
-            out_channels = 8;
-            filter_ht = 1;
-            filter_wd = 1;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 2: // ch, filter (3x3x3)
-            in_wd = 10;
-            in_ht = 10;
-            in_channels = 3;
-            out_channels = 64;
-            filter_ht = 3;
-            filter_wd = 3;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 3: // remaining pad (0, 0)
-            in_wd = 10;
-            in_ht = 10;
-            in_channels = 3;
-            out_channels = 64;
-            filter_ht = 1;
-            filter_wd = 1;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 4: // unopt case
-            in_wd = 10;
-            in_ht = 10;
-            in_channels = 12;
-            out_channels = 64;
-            filter_ht = 3;
-            filter_wd = 3;
-            pad_wd = 1;
-            pad_ht = 1;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        case 5: // ch % 8 == 0 & stride (2,2)
-            in_wd = 16;
-            in_ht = 16;
-            in_channels = 16;
-            out_channels = 16;
-            filter_ht = 1;
-            filter_wd = 1;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 2;
-            stride_ht = 2;
-            break;
-        case 6: // ch % 8 == 0 && filter (1,1), padding (0,0)
-            in_wd = 2;
-            in_ht = 2;
-            in_channels = 8;
-            out_channels = 8;
-            filter_ht = 1;
-            filter_wd = 1;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        default: // ch % 8 == 0
-            in_wd = 8;
-            in_ht = 8;
-            in_channels = 16;
-            out_channels = 16;
-            filter_ht = 1;
-            filter_wd = 1;
-            pad_wd = 0;
-            pad_ht = 0;
-            stride_wd = 1;
-            stride_ht = 1;
-            break;
-        }
-
-        /* prepare data */
-        uint16_t out_wd = (in_wd - filter_wd + 1) / stride_wd;
-        uint16_t out_ht = (in_ht - filter_ht + 1) / stride_ht;
-
-        int in_size = in_wd * in_ht * in_channels;
-        int filter_size = filter_wd * filter_ht * in_channels * out_channels + 2;
-        int out_size = out_wd * out_ht * out_channels;
-
-#if IDF_HEAP_CAPS
-        input_orig = (int8_t *) heap_caps_malloc(in_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        out_c_orig = (int8_t *) heap_caps_malloc(out_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        out_opt_orig = (int8_t *) heap_caps_malloc(out_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        filter_data = (int8_t *) heap_caps_malloc(filter_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-        bias = (int32_t *) heap_caps_malloc(128 + sizeof (int32_t) * out_channels, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-
-        int8_t *input = 16 + input_orig - ((uint32_t) input_orig & 0xf);
-        int8_t *out_data_c = 16 + out_c_orig - ((uint32_t) out_c_orig & 0xf);
-        int8_t *out_data_opt = 16 + out_opt_orig - ((uint32_t) out_opt_orig & 0xf);
-#else
-        int8_t *input = memalign(16, in_size);
-        int8_t *out_data_c = memalign(16, out_size);
-        int8_t *out_data_opt = memalign(16, out_size);
-        filter_data = memalign(16, filter_size);
-        bias = calloc(1, 128 + sizeof (int32_t) * out_channels);
-        input_orig = input;
-        out_c_orig = out_data_c;
-        out_opt_orig = out_data_opt;
-#endif
-        int32_t *out_shift = calloc(1, 128 + sizeof (int32_t) * out_channels);
-        int32_t *out_mult = calloc(1, 128 + sizeof (int32_t) * out_channels);
-
-        if (input == NULL || filter_data == NULL ||
-                out_data_c == NULL || out_data_opt == NULL) {
-            printf(ANSI_COLOR_RED"%s allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-            goto conv_s8_cleanup;
-        }
-
-        if (bias == NULL || out_shift == NULL || out_mult == NULL) {
-            printf(ANSI_COLOR_RED"%s allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-            goto conv_s8_cleanup;
-        }
-
-        /* Generate input data between -128 -> +127 */
-        for (int i = 0; i < in_size; ++i) {
-            input[i] = rand() % 255 - 128;
-        }
-
-        /* Generate filter data between -128 -> +127 */
-        for (int i = 0; i < filter_size; ++i) {
-            filter_data[i] = rand() % 256 - 128;
-        }
-
-        /* Generate bias data */
-        for (int i = 0; i < out_channels; ++i) {
-            bias[i] = (int32_t)rand() % UINT16_MAX + UINT8_MAX;
-        }
-
-        /* Shift and multiplier */
-        for (int i = 0; i < out_channels; ++i) {
-            out_shift[i] = -10 + rand() % 2;
-            out_mult[i] = 0x7f67f4f8 + rand() % 50;
-        }
-
-        int scratch_buf_size = esp_nn_get_conv_scratch_size(in_wd, in_ht, in_channels,
-                                                            out_channels, filter_wd, filter_ht);
-        if (scratch_buf_size > 0) {
-#if IDF_HEAP_CAPS
-            void *scratch_buf = heap_caps_malloc(scratch_buf_size + 32, MALLOC_CAP_SPIRAM | MALLOC_CAP_8BIT);
-            int align_sz = 16 - (((int32_t) scratch_buf) & 0xf);
-#else
-            void *scratch_buf = memalign(16, scratch_buf_size);
-            int align_sz = 0;
-#endif
-            if (scratch_buf == NULL) {
-                printf(ANSI_COLOR_RED"%s scratch_buf alloc failed size %d\n"ANSI_COLOR_RESET, __FUNCTION__, scratch_buf_size);
-                goto conv_s8_cleanup;
-            }
-            esp_nn_set_conv_scratch_buf(scratch_buf + align_sz);
-        }
-
-        if (itr == 0) {
-            /* enable profiler */
-            profile_c_start();
-        }
-
-        /* C function */
-        esp_nn_conv_s8_ansi(input, in_wd, in_ht, in_channels, input_offset,
-                            pad_wd, pad_ht, stride_wd, stride_ht,
-                            filter_data + 2, filter_wd, filter_ht, bias,
-                            out_data_c, out_wd, out_ht, out_channels, out_offset, out_shift,
-                            out_mult, activation_min, activation_max);
-
-        if (itr == 0) {
-            profile_c_end();
-            profile_opt_start();
-        }
-
-        /* Optimized function */
-        esp_nn_conv_s8(input, in_wd, in_ht, in_channels, input_offset,
-                    pad_wd, pad_ht, stride_wd, stride_ht,
-                    filter_data + 2, filter_wd, filter_ht, bias,
-                    out_data_opt, out_wd, out_ht, out_channels, out_offset, out_shift,
-                    out_mult, activation_min, activation_max);
-
-        if (itr == 0) {
-            /* disable profiler */
-            profile_opt_end();
-        }
-
-        bool ret = CHECK_EQUAL(out_data_c, out_data_opt, out_size);
-        if (ret == false) {
-            printf(ANSI_COLOR_RED"%s[%d] failed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-            printf("Output: \n");
-            PRINT_ARRAY_HEX(out_data_opt, out_size / out_ht, out_ht);
-            printf("Expected: \n");
-            PRINT_ARRAY_HEX(out_data_c, out_size / out_ht, out_ht);
-            printf("Input:\n");
-            PRINT_ARRAY_HEX(input, in_size / in_ht, in_ht);
-            printf("Filter data:\n");
-            PRINT_ARRAY_HEX(filter_data + 2, (filter_size - 2) / filter_ht, filter_ht);
-            printf("bias data:\n");
-            PRINT_ARRAY_INT(bias, out_channels, 1);
-            goto conv_s8_cleanup;
-        }
-        printf(ANSI_COLOR_GREEN"%s[%d] passed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-
-    conv_s8_cleanup:
-        if (input) {
-            free(input_orig);
-        }
-        if (filter_data) {
-            free(filter_data);
-        }
-        if (out_data_c) {
-            free(out_c_orig);
-        }
-        if (out_data_opt) {
-            free(out_opt_orig);
-        }
-        if (bias) {
-            free(bias);
-        }
-        if (out_shift) {
-            free(out_shift);
-        }
-        if (out_mult) {
-            free(out_mult);
-        }
-        if (scratch_buf) {
-            free(scratch_buf);
-        }
-    }
-}

code/components/esp-nn/tests/src/fully_connected_test.c

@@ -1,111 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <stdlib.h>
-
-#include <esp_nn.h>
-#include "test_utils.h"
-
-
-void esp_nn_fully_connected_s8_test()
-{
-    /* prepare data */
-    static uint16_t row_len = 256 + 8 + 7; /* odd len to test unaligned+left-over */
-    static uint16_t out_channels = 3;
-    int8_t input[row_len];
-    int8_t filter_data[row_len * out_channels];
-    int8_t output_c[out_channels], output_opt[out_channels];
-    static int32_t activation_min = -128;
-    static int32_t activation_max = 127;
-    static int32_t input_offset = 0;
-    static int32_t filter_offset = 0;
-    int32_t out_shift = -10;
-    static int32_t out_offset = 127;
-    int32_t out_mult = 0x59e492c4;
-    for (int itr = 0; itr < 5; itr++) {
-        out_mult = INT32_MAX / row_len + rand() % INT16_MAX;
-        switch (itr) {
-        case 0:
-            out_shift = -10;
-            break;
-        case 1:
-            out_shift = SHIFT_MIN;
-            break;
-        case 2:
-            out_shift = SHIFT_MAX;
-            break;
-        case 3:
-            out_shift = 0;
-            break;
-        default:
-            out_shift = -10 + rand() % 5;
-            break;
-        }
-        if (itr == 0) {
-            out_shift = SHIFT_MAX;
-        }
-        /* Generate input and filter data */
-        for (int i = 0; i < row_len; ++i) {
-            input[i] = rand() % 256 - 128;
-        }
-        for (int i = 0; i < row_len * out_channels; ++i) {
-            filter_data[i] = rand() % 256 - 128;
-        }
-
-        if (itr == 0) {
-            /* enable profiler */
-            profile_c_start();
-        }
-
-        /* C function */
-        esp_nn_fully_connected_s8_ansi(input, input_offset, row_len, filter_data, filter_offset,
-                                    NULL, output_c, out_channels, out_offset, out_shift, out_mult,
-                                    activation_min, activation_max);
-
-        if (itr == 0) {
-            profile_c_end();
-            profile_opt_start();
-        }
-
-        /* Optimized function */
-        esp_nn_fully_connected_s8(input, input_offset, row_len, filter_data, filter_offset,
-                                NULL, output_opt, out_channels, out_offset, out_shift, out_mult,
-                                activation_min, activation_max);
-
-        if (itr == 0) {
-            /* disable profiler */
-            profile_opt_end();
-        }
-
-        bool ret = CHECK_EQUAL(output_c, output_opt, out_channels);
-        if (ret == false) {
-            printf(ANSI_COLOR_RED"%s[%d] failed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-            printf("Output: \n");
-            PRINT_ARRAY_HEX(output_opt, out_channels, 1);
-            printf("Expected: \n");
-            PRINT_ARRAY_HEX(output_c, out_channels, 1);
-            printf("Input:\n");
-            PRINT_ARRAY_HEX(input, row_len, 1);
-            printf("Filter data:\n");
-            PRINT_ARRAY_HEX(filter_data, row_len, out_channels);
-            printf("Out shift: %d\n", out_shift);
-            printf("Out mult: %x\n", out_mult);
-            return;
-        }
-        printf(ANSI_COLOR_GREEN"%s[%d] passed\n"ANSI_COLOR_RESET, __FUNCTION__, itr);
-    }
-}

code/components/esp-nn/tests/src/pooling_test.c

@@ -1,184 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <malloc.h>
-
-#include <esp_nn.h>
-#include "test_utils.h"
-
-
-void esp_nn_avg_pool_s8_test()
-{
-    /* prepare data */
-    const uint16_t input_wd = 16;
-    const uint16_t input_ht = 16;
-    const uint16_t channels = 16; /* With TFLite example, I have seen it 256 */
-    const int size = input_wd * input_ht * channels;
-    int8_t *input, *output_c, *output_opt;
-    const int32_t activation_min = -128;
-    const int32_t activation_max = 127;
-    const uint16_t pad_wd = 1;
-    const uint16_t pad_ht = 1;
-    const uint16_t stride_wd = 1;
-    const uint16_t stride_ht = 1;
-    const uint16_t filter_ht = 3;
-    const uint16_t filter_wd = 3;
-    const uint16_t out_wd = input_wd / stride_wd;
-    const uint16_t out_ht = input_ht / stride_ht;
-    const int out_size = out_wd * out_ht * channels;
-
-    input = memalign(16, size);
-    output_c = memalign(16, out_size);
-    output_opt = memalign(16, out_size);
-
-    if (input == NULL || output_c == NULL || output_opt == NULL) {
-        printf(ANSI_COLOR_RED"%s allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        goto avg_pool_s8_cleanup;
-    }
-    /**
-     * width/height, channels etc look suspicious but it it true.
-     * It actually depends upon where in model this is actually placed.
-     * If at the end wd/ht tends to be smaller and depth larger.
-     */
-
-    for (int i = 0; i < size; ++i) {
-        input[i] = rand() % 256 - 128;
-    }
-
-    /* enable profiler */
-    profile_c_start();
-
-    /* C function */
-    esp_nn_avg_pool_s8_ansi(input, input_wd, input_ht, output_c, out_wd, out_ht,
-                              stride_wd, stride_ht, filter_wd, filter_ht, pad_wd, pad_ht,
-                              activation_min, activation_max, channels);
-
-    profile_c_end();
-    profile_opt_start();
-
-    /* Optimized function */
-    esp_nn_avg_pool_s8(input, input_wd, input_ht, output_opt, out_wd, out_ht,
-                         stride_wd, stride_ht, filter_wd, filter_ht, pad_wd, pad_ht,
-                         activation_min, activation_max, channels);
-
-    /* disable profiler */
-    profile_opt_end();
-
-
-    bool ret = CHECK_EQUAL(output_c, output_opt, out_size);
-    if (ret == false) {
-        printf(ANSI_COLOR_RED"%s failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        printf("Output: \n");
-        PRINT_ARRAY_HEX(output_opt, out_wd * channels, out_ht);
-        printf("Expected: \n");
-        PRINT_ARRAY_HEX(output_c, out_wd * channels, out_ht);
-        printf("Input:\n");
-        PRINT_ARRAY_HEX(input, input_wd * channels, input_ht);
-        goto avg_pool_s8_cleanup;
-    }
-    printf(ANSI_COLOR_GREEN"%s passed\n"ANSI_COLOR_RESET, __FUNCTION__);
-
-avg_pool_s8_cleanup:
-    if (input) {
-        free(input);
-    }
-    if (output_c) {
-        free(output_c);
-    }
-    if (output_opt) {
-        free(output_opt);
-    }
-}
-
-void esp_nn_max_pool_s8_test()
-{
-    /* prepare data */
-    const uint16_t input_wd = 16;
-    const uint16_t input_ht = 16;
-    const uint16_t channels = 16; /* With TFLite example, I have seen it 256 */
-    int8_t *input, *output_c, *output_opt;
-    const int size = input_wd * input_ht * channels;
-    const int32_t activation_min = -128;
-    const int32_t activation_max = 127;
-    const uint16_t pad_wd = 1;
-    const uint16_t pad_ht = 1;
-    const uint16_t stride_wd = 1;
-    const uint16_t stride_ht = 1;
-    const uint16_t filter_ht = 3;
-    const uint16_t filter_wd = 3;
-    const uint16_t out_wd = input_wd / stride_wd;
-    const uint16_t out_ht = input_ht / stride_ht;
-    const int out_size = out_wd * out_ht * channels;
-
-    input = memalign(16, size);
-    output_c = memalign(16, out_size);
-    output_opt = memalign(16, out_size);
-
-    if (input == NULL || output_c == NULL || output_opt == NULL) {
-        printf(ANSI_COLOR_RED"%s allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        goto max_pool_s8_cleanup;
-    }
-
-    for (int i = 0; i < size; ++i) {
-        input[i] = rand() % 256 - 128;
-    }
-
-    /* enable profiler */
-    profile_c_start();
-
-    /* C function */
-    esp_nn_max_pool_s8_ansi(input, input_wd, input_ht, output_c, out_wd, out_ht,
-                            stride_wd, stride_ht, filter_wd, filter_ht, pad_wd, pad_ht,
-                            activation_min, activation_max, channels);
-
-    profile_c_end();
-    profile_opt_start();
-
-    /* Optimized function */
-    esp_nn_max_pool_s8(input, input_wd, input_ht, output_opt, out_wd, out_ht,
-                       stride_wd, stride_ht, filter_wd, filter_ht, pad_wd, pad_ht,
-                       activation_min, activation_max, channels);
-
-    /* disable profiler */
-    profile_opt_end();
-
-
-    bool ret = CHECK_EQUAL(output_c, output_opt, out_wd * out_ht * channels);
-    if (ret == false) {
-        printf(ANSI_COLOR_RED"%s failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        printf("Output: \n");
-        PRINT_ARRAY_HEX(output_opt, out_wd * out_ht * channels, 1);
-        printf("Expected: \n");
-        PRINT_ARRAY_HEX(output_c, out_wd * out_ht * channels, 1);
-        printf("Input:\n");
-        PRINT_ARRAY_HEX(input, 8, size / 8);
-        goto max_pool_s8_cleanup;
-    }
-    printf(ANSI_COLOR_GREEN"%s passed\n"ANSI_COLOR_RESET, __FUNCTION__);
-
-max_pool_s8_cleanup:
-    if (input) {
-        free(input);
-    }
-    if (output_c) {
-        free(output_c);
-    }
-    if (output_opt) {
-        free(output_opt);
-    }
-}

code/components/esp-nn/tests/src/relu_test.c

@@ -1,83 +0,0 @@
-// Copyright 2020-2021 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <malloc.h>
-
-#include <esp_nn.h>
-#include "test_utils.h"
-
-void esp_nn_relu6_s8_test()
-{
-    const int size = 1600 + 8 + 7;
-    int8_t *input, *inout_ansi, *inout_opt;
-
-    input = memalign(16, size);
-    inout_ansi = memalign(16, size);
-    inout_opt = memalign(16, size);
-
-    if (input == NULL || inout_ansi == NULL || inout_opt == NULL) {
-        printf(ANSI_COLOR_RED"%s allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        goto relu6_s8_cleanup;
-    }
-    /* Generate filter data between -128 -> +127 */
-    for (int i = 0; i < size; ++i) {
-        input[i] = rand() % 255 - 128;
-        inout_ansi[i] = input[i];
-        inout_opt[i] = input[i];
-    }
-
-    /* enable profiler */
-    profile_c_start();
-
-    /* C function */
-    esp_nn_relu6_s8_ansi(inout_ansi, size);
-
-    profile_c_end();
-    profile_opt_start();
-
-    /* Optimized function */
-    esp_nn_relu6_s8(inout_opt, size);
-
-    /* disable profiler */
-    profile_opt_end();
-
-    bool ret = CHECK_EQUAL(inout_ansi, inout_opt, size);
-    if (ret == false) {
-        printf(ANSI_COLOR_RED"%s failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        printf("Output: \n");
-        PRINT_ARRAY_HEX(inout_opt, size, 1);
-        printf("Expected: \n");
-        PRINT_ARRAY_HEX(inout_ansi, size, 1);
-        printf("Input:\n");
-        PRINT_ARRAY_HEX(input, size, 1);
-        goto relu6_s8_cleanup;
-    }
-    printf(ANSI_COLOR_GREEN"%s passed\n"ANSI_COLOR_RESET, __FUNCTION__);
-
-relu6_s8_cleanup:
-    if (input) {
-        free (input);
-    }
-    if (inout_ansi) {
-        free (inout_ansi);
-    }
-    if (inout_opt) {
-        free (inout_opt);
-    }
-
-}

code/components/esp-nn/tests/src/softmax_test.c

@@ -1,101 +0,0 @@
-// Copyright 2022 Espressif Systems (Shanghai) PTE LTD
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-#include <stdint.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <malloc.h>
-
-#include <esp_nn.h>
-#include "test_utils.h"
-
-void esp_nn_softmax_s8_test()
-{
-    const int32_t height = 8;
-    const int32_t width = 32;
-    const int32_t diff_min = -128;
-    const int32_t mult = INT32_MAX / 2;
-    const int32_t shift = 7;
-    void *scratch_buf = NULL;
-    const int size = width * height;
-    int8_t *input, *out_ansi, *out_opt;
-
-    input = memalign(16, size);
-    out_ansi = memalign(16, size);
-    out_opt = memalign(16, size);
-
-    if (input == NULL || out_ansi == NULL || out_opt == NULL) {
-        printf(ANSI_COLOR_RED"%s buffer allocations failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        goto softmax_s8_cleanup;
-    }
-
-    /* Generate input data between -128 -> +127 */
-    for (int i = 0; i < size; ++i) {
-        input[i] = rand() % 255 - 128;
-    }
-
-    /* enable profiler */
-    profile_c_start();
-
-    /* C function */
-    esp_nn_softmax_s8_ansi(input, height, width, mult, shift, diff_min, out_ansi);
-
-    profile_c_end();
-
-    int32_t scratch_buf_size = esp_nn_get_softmax_scratch_size(width, height);
-    if (scratch_buf_size) {
-        scratch_buf = memalign(4, scratch_buf_size);
-        if (scratch_buf == NULL) {
-            printf(ANSI_COLOR_RED"%s scratch_buf alloc failed size %d\n"ANSI_COLOR_RESET, __FUNCTION__, scratch_buf_size);
-            goto softmax_s8_cleanup;
-        }
-        esp_nn_set_softmax_scratch_buf(scratch_buf);
-    }
-
-    profile_opt_start();
-
-    /* Optimized function */
-    esp_nn_softmax_s8(input, height, width, mult, shift, diff_min, out_opt);
-
-    /* disable profiler */
-    profile_opt_end();
-
-    bool ret = CHECK_EQUAL(out_ansi, out_opt, size);
-    if (ret == false) {
-        printf(ANSI_COLOR_RED"%s failed\n"ANSI_COLOR_RESET, __FUNCTION__);
-        printf("Output: \n");
-        PRINT_ARRAY_HEX(out_opt, width, height);
-        printf("Expected: \n");
-        PRINT_ARRAY_HEX(out_ansi, width, height);
-        printf("Input:\n");
-        PRINT_ARRAY_HEX(input, width, height);
-        goto softmax_s8_cleanup;
-    }
-    printf(ANSI_COLOR_GREEN"%s passed\n"ANSI_COLOR_RESET, __FUNCTION__);
-
-softmax_s8_cleanup:
-    if (input) {
-        free (input);
-    }
-    if (out_ansi) {
-        free (out_ansi);
-    }
-    if (out_opt) {
-        free (out_opt);
-    }
-    if (scratch_buf) {
-        free (scratch_buf);
-    }
-}

code/components/esp-sdmmc/CMakeLists.txt

@@ -0,0 +1,11 @@
+idf_component_register(SRCS "sdmmc_cmd_mh.c"
+                            "sdmmc_common_mh.c"
+                            "sdmmc_init_mh.c"
+                            "sdmmc_io_mh.c"
+                            "sdmmc_mmc_mh.c"
+                            "sdmmc_sd_mh.c"
+                    INCLUDE_DIRS "." "include"
+                    REQUIRES driver esp-fatfs
+                    PRIV_REQUIRES soc)
+
+target_compile_options(${COMPONENT_LIB} PRIVATE "-Wno-format")

code/components/esp-sdmmc/README.md

@@ -0,0 +1,279 @@
+# AIOTED related changes, see https://github.com/jomjol/AI-on-the-edge-device/pull/2781
+These files/folders were copied from `framework-espidf@3.50002.230601/components/` and adapted to our own needs.
+Since not every SD/MMC was recognized and this was due to the implementation of ATA trim support, this was revised.
+Furthermore, files that we don't need were deleted from it.
+
+## The most relevant changes are:
+### fatfs/diskio/diskio_sdmmc.c
+DRESULT ff_sdmmc_ioctl (BYTE pdrv, BYTE cmd, void* buff), at lines 106 to 110 changed from:
+```c
+#if FF_USE_TRIM
+        case CTRL_TRIM:
+            return ff_sdmmc_trim (pdrv, *((DWORD*)buff), //start_sector
+                    (*((DWORD*)buff + 1) - *((DWORD*)buff) + 1)); //sector_count
+#endif //FF_USE_TRIM
+```
+to:
+```c
+#if (FF_USE_TRIM)
+        case CTRL_TRIM:
+            if(FF_CAN_TRIM){
+                return ff_sdmmc_trim (pdrv, *((DWORD*)buff), //start_sector
+                    (*((DWORD*)buff + 1) - *((DWORD*)buff) + 1)); //sector_count
+            }
+            else{
+                return RES_ERROR;
+            }
+#endif //FF_USE_TRIM
+```
+
+### fatfs/src/ff.c
+added:
+```c
+#include "sdmmc_cmd.h"
+```
+
+static FRESULT remove_chain(FFOBJID* obj, DWORD clst, DWORD pclst), at lines 1437 to 1454 changed from:
+```c
+#if FF_FS_EXFAT || FF_USE_TRIM
+		if (ecl + 1 == nxt) {	/* Is next cluster contiguous? */
+			ecl = nxt;
+		} else {				/* End of contiguous cluster block */
+#if FF_FS_EXFAT
+			if (fs->fs_type == FS_EXFAT) {
+				res = change_bitmap(fs, scl, ecl - scl + 1, 0);	/* Mark the cluster block 'free' on the bitmap */
+				if (res != FR_OK) return res;
+			}
+#endif
+#if FF_USE_TRIM
+			rt[0] = clst2sect(fs, scl);					/* Start of data area to be freed */
+			rt[1] = clst2sect(fs, ecl) + fs->csize - 1;	/* End of data area to be freed */
+			disk_ioctl(fs->pdrv, CTRL_TRIM, rt);		/* Inform storage device that the data in the block may be erased */
+#endif
+			scl = ecl = nxt;
+		}
+#endif
+```
+
+to:
+```c
+#if FF_FS_EXFAT || FF_USE_TRIM
+		if(FF_FS_EXFAT || FF_CAN_TRIM){
+			if (ecl + 1 == nxt) {	/* Is next cluster contiguous? */
+				ecl = nxt;
+			} 
+			else {				/* End of contiguous cluster block */
+#if FF_FS_EXFAT
+				if (fs->fs_type == FS_EXFAT) {
+					res = change_bitmap(fs, scl, ecl - scl + 1, 0);	/* Mark the cluster block 'free' on the bitmap */
+					if (res != FR_OK) return res;
+				}
+#endif
+#if FF_USE_TRIM
+				if(FF_CAN_TRIM){
+					rt[0] = clst2sect(fs, scl);					/* Start of data area to be freed */
+					rt[1] = clst2sect(fs, ecl) + fs->csize - 1;	/* End of data area to be freed */
+					disk_ioctl(fs->pdrv, CTRL_TRIM, rt);		/* Inform storage device that the data in the block may be erased */
+				}
+#endif
+				scl = ecl = nxt;
+			}
+		}
+#endif
+```
+
+FRESULT f_mkfs(const TCHAR* path, const MKFS_PARM* opt, void* work, UINT len), at lines 5946 to 5949 changed from:
+```c
+#if FF_USE_TRIM
+		lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+		disk_ioctl(pdrv, CTRL_TRIM, lba);
+#endif
+```
+to:
+```c
+#if FF_USE_TRIM
+		if(FF_CAN_TRIM){
+			lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+			disk_ioctl(pdrv, CTRL_TRIM, lba);
+		}
+#endif
+```
+
+FRESULT f_mkfs(const TCHAR* path, const MKFS_PARM* opt, void* work, UINT len), at lines 6175 to 6178 changed from:
+```c
+#if FF_USE_TRIM
+		lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+		disk_ioctl(pdrv, CTRL_TRIM, lba);
+#endif
+```
+to:
+```c
+#if FF_USE_TRIM
+		if(FF_CAN_TRIM){
+			lba[0] = b_vol; lba[1] = b_vol + sz_vol - 1;	/* Inform storage device that the volume area may be erased */
+			disk_ioctl(pdrv, CTRL_TRIM, lba);
+		}
+#endif
+```
+
+### sdmmc/sdmmc_cmd.c
+added:
+```c
+int FF_CAN_TRIM = 0;
+```
+
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card), at lines 630 to 636 changed from:
+```c
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.sec_feature & EXT_CSD_SEC_GB_CL_EN)) {
+        return ESP_OK;
+    }
+    return ESP_FAIL;
+}
+```
+to:
+```c
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.sec_feature & EXT_CSD_SEC_GB_CL_EN)) {
+        FF_CAN_TRIM = 1;
+        return ESP_OK;
+    }
+    FF_CAN_TRIM = 0;
+    return ESP_FAIL;
+}
+```
+
+### sdmmc/include/sdmmc_cmd.h
+
+added:
+```c
+extern int FF_CAN_TRIM;
+```
+
+
+# Espressif IoT Development Framework
+
+* [中文版](./README_CN.md)
+
+ESP-IDF is the development framework for Espressif SoCs supported on Windows, Linux and macOS.
+
+# ESP-IDF Release Support Schedule
+
+![Support Schedule](https://dl.espressif.com/dl/esp-idf/support-periods.svg)
+
+- Please read [the support policy](SUPPORT_POLICY.md) and [the documentation](https://docs.espressif.com/projects/esp-idf/en/latest/esp32/versions.html) for more information about ESP-IDF versions.
+- Please see the [End-of-Life Advisories](https://www.espressif.com/en/support/documents/advisories?keys=&field_type_of_advisory_tid%5B%5D=817) for information about ESP-IDF releases with discontinued support.
+
+# ESP-IDF Release and SoC Compatibility
+
+The following table shows ESP-IDF support of Espressif SoCs where ![alt text][preview] and ![alt text][supported] denote preview status and support, respectively. The preview support is usually limited in time and intended for beta versions of chips. Please use an ESP-IDF release where the desired SoC is already supported.
+
+|Chip         |          v4.1         |          v4.2          |         v4.3           |          v4.4          |          v5.0          |                                                                                      |
+|:----------- |:---------------------:| :---------------------:| :---------------------:| :---------------------:| :---------------------:|:------------------------------------------------------------------------------------ |
+|ESP32        |![alt text][supported] | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] |                                                                                      |
+|ESP32-S2     |                       | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] |                                                                                      |
+|ESP32-C3     |                       |                        | ![alt text][supported] | ![alt text][supported] | ![alt text][supported] |                                                                                      |
+|ESP32-S3     |                       |                        |                        | ![alt text][supported] | ![alt text][supported] | [Announcement](https://www.espressif.com/en/news/ESP32_S3)                           |
+|ESP32-C2     |                       |                        |                        |                        | ![alt text][supported] | [Announcement](https://blog.espressif.com/esp32-c2-and-why-it-matter-s-bcf4d7d0b2c6) |
+|ESP32-H2     |                       |                        |                        | ![alt text][preview]   | ![alt text][preview]   | [Announcement](https://www.espressif.com/en/news/ESP32_H2)                           |
+
+[supported]: https://img.shields.io/badge/-supported-green "supported"
+[preview]: https://img.shields.io/badge/-preview-orange "preview"
+
+Espressif SoCs released before 2016 (ESP8266 and ESP8285) are supported by [RTOS SDK](https://github.com/espressif/ESP8266_RTOS_SDK) instead.
+
+# Developing With ESP-IDF
+
+## Setting Up ESP-IDF
+
+See https://idf.espressif.com/ for links to detailed instructions on how to set up the ESP-IDF depending on chip you use.
+
+**Note:** Each SoC series and each ESP-IDF release has its own documentation. Please see Section [Versions](https://docs.espressif.com/projects/esp-idf/en/latest/esp32/versions.html) on how to find documentation and how to checkout specific release of ESP-IDF.
+
+### Non-GitHub forks
+
+ESP-IDF uses relative locations as its submodules URLs ([.gitmodules](.gitmodules)). So they link to GitHub. If ESP-IDF is forked to a Git repository which is not on GitHub, you will need to run the script [tools/set-submodules-to-github.sh](tools/set-submodules-to-github.sh) after git clone.
+
+The script sets absolute URLs for all submodules, allowing `git submodule update --init --recursive` to complete. If cloning ESP-IDF from GitHub, this step is not needed.
+
+## Finding a Project
+
+As well as the [esp-idf-template](https://github.com/espressif/esp-idf-template) project mentioned in Getting Started, ESP-IDF comes with some example projects in the [examples](examples) directory.
+
+Once you've found the project you want to work with, change to its directory and you can configure and build it.
+
+To start your own project based on an example, copy the example project directory outside of the ESP-IDF directory.
+
+# Quick Reference
+
+See the Getting Started guide links above for a detailed setup guide. This is a quick reference for common commands when working with ESP-IDF projects:
+
+## Setup Build Environment
+
+(See the Getting Started guide listed above for a full list of required steps with more details.)
+
+* Install host build dependencies mentioned in the Getting Started guide.
+* Run the install script to set up the build environment. The options include `install.bat` or `install.ps1` for Windows, and `install.sh` or `install.fish` for Unix shells.
+* Run the export script on Windows (`export.bat`) or source it on Unix (`source export.sh`) in every shell environment before using ESP-IDF.
+
+## Configuring the Project
+
+* `idf.py set-target <chip_name>` sets the target of the project to `<chip_name>`. Run `idf.py set-target` without any arguments to see a list of supported targets.
+* `idf.py menuconfig` opens a text-based configuration menu where you can configure the project.
+
+## Compiling the Project
+
+`idf.py build`
+
+... will compile app, bootloader and generate a partition table based on the config.
+
+## Flashing the Project
+
+When the build finishes, it will print a command line to use esptool.py to flash the chip. However you can also do this automatically by running:
+
+`idf.py -p PORT flash`
+
+Replace PORT with the name of your serial port (like `COM3` on Windows, `/dev/ttyUSB0` on Linux, or `/dev/cu.usbserial-X` on MacOS. If the `-p` option is left out, `idf.py flash` will try to flash the first available serial port.
+
+This will flash the entire project (app, bootloader and partition table) to a new chip. The settings for serial port flashing can be configured with `idf.py menuconfig`.
+
+You don't need to run `idf.py build` before running `idf.py flash`, `idf.py flash` will automatically rebuild anything which needs it.
+
+## Viewing Serial Output
+
+The `idf.py monitor` target uses the [idf_monitor tool](https://docs.espressif.com/projects/esp-idf/en/latest/get-started/idf-monitor.html) to display serial output from Espressif SoCs. idf_monitor also has a range of features to decode crash output and interact with the device. [Check the documentation page for details](https://docs.espressif.com/projects/esp-idf/en/latest/get-started/idf-monitor.html).
+
+Exit the monitor by typing Ctrl-].
+
+To build, flash and monitor output in one pass, you can run:
+
+`idf.py flash monitor`
+
+## Compiling & Flashing Only the App
+
+After the initial flash, you may just want to build and flash just your app, not the bootloader and partition table:
+
+* `idf.py app` - build just the app.
+* `idf.py app-flash` - flash just the app.
+
+`idf.py app-flash` will automatically rebuild the app if any source files have changed.
+
+(In normal development there's no downside to reflashing the bootloader and partition table each time, if they haven't changed.)
+
+## Erasing Flash
+
+The `idf.py flash` target does not erase the entire flash contents. However it is sometimes useful to set the device back to a totally erased state, particularly when making partition table changes or OTA app updates. To erase the entire flash, run `idf.py erase-flash`.
+
+This can be combined with other targets, ie `idf.py -p PORT erase-flash flash` will erase everything and then re-flash the new app, bootloader and partition table.
+
+# Resources
+
+* Documentation for the latest version: https://docs.espressif.com/projects/esp-idf/. This documentation is built from the [docs directory](docs) of this repository.
+
+* The [esp32.com forum](https://esp32.com/) is a place to ask questions and find community resources.
+
+* [Check the Issues section on github](https://github.com/espressif/esp-idf/issues) if you find a bug or have a feature request. Please check existing Issues before opening a new one.
+
+* If you're interested in contributing to ESP-IDF, please check the [Contributions Guide](https://docs.espressif.com/projects/esp-idf/en/latest/contribute/index.html).

code/components/esp-sdmmc/include/sdmmc_cmd_mh.h

@@ -0,0 +1,357 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2021 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#pragma once
+
+#include <stdio.h>
+#include "esp_err.h"
+#include "driver/sdmmc_types.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern int FF_CAN_TRIM;
+
+/**
+ * Probe and initialize SD/MMC card using given host
+ *
+ * @note Only SD cards (SDSC and SDHC/SDXC) are supported now.
+ *       Support for MMC/eMMC cards will be added later.
+ *
+ * @param host  pointer to structure defining host controller
+ * @param out_card  pointer to structure which will receive information
+ *                  about the card when the function completes
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_card_init(const sdmmc_host_t* host,
+        sdmmc_card_t* out_card);
+
+/**
+ * @brief Print information about the card to a stream
+ * @param stream  stream obtained using fopen or fdopen
+ * @param card  card information structure initialized using sdmmc_card_init
+ */
+void sdmmc_card_print_info(FILE* stream, const sdmmc_card_t* card);
+
+/**
+ * Get status of SD/MMC card
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_get_status(sdmmc_card_t* card);
+
+/**
+ * Write given number of sectors to SD/MMC card
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param src   pointer to data buffer to read data from; data size must be
+ *              equal to sector_count * card->csd.sector_size
+ * @param start_sector  sector where to start writing
+ * @param sector_count  number of sectors to write
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_write_sectors(sdmmc_card_t* card, const void* src,
+        size_t start_sector, size_t sector_count);
+
+/**
+ * Read given number of sectors from the SD/MMC card
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param dst   pointer to data buffer to write into; buffer size must be
+ *              at least sector_count * card->csd.sector_size
+ * @param start_sector  sector where to start reading
+ * @param sector_count  number of sectors to read
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_read_sectors(sdmmc_card_t* card, void* dst,
+        size_t start_sector, size_t sector_count);
+
+/**
+ * Erase given number of sectors from the SD/MMC card
+ *
+ * @note When sdmmc_erase_sectors used with cards in SDSPI mode, it was
+ * observed that card requires re-init after erase operation.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param start_sector  sector where to start erase
+ * @param sector_count  number of sectors to erase
+ * @param arg  erase command (CMD38) argument
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_erase_sectors(sdmmc_card_t* card, size_t start_sector,
+        size_t sector_count, sdmmc_erase_arg_t arg);
+
+/**
+ * Check if SD/MMC card supports discard
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @return
+ *      - ESP_OK if supported by the card/device
+ *      - ESP_FAIL if not supported by the card/device
+ */
+esp_err_t sdmmc_can_discard(sdmmc_card_t* card);
+
+/**
+ * Check if SD/MMC card supports trim
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @return
+ *      - ESP_OK if supported by the card/device
+ *      - ESP_FAIL if not supported by the card/device
+ */
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card);
+
+/**
+ * Check if SD/MMC card supports sanitize
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @return
+ *      - ESP_OK if supported by the card/device
+ *      - ESP_FAIL if not supported by the card/device
+ */
+esp_err_t sdmmc_mmc_can_sanitize(sdmmc_card_t* card);
+
+/**
+ * Sanitize the data that was unmapped by a Discard command
+ *
+ * @note  Discard command has to precede sanitize operation. To discard, use
+ *        MMC_DICARD_ARG with sdmmc_erase_sectors argument
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param timeout_ms timeout value in milliseconds required to sanitize the
+ *                   selected range of sectors.
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_mmc_sanitize(sdmmc_card_t* card, uint32_t timeout_ms);
+
+/**
+ * Erase complete SD/MMC card
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_full_erase(sdmmc_card_t* card);
+
+/**
+ * Read one byte from an SDIO card using IO_RW_DIRECT (CMD52)
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param function  IO function number
+ * @param reg  byte address within IO function
+ * @param[out] out_byte  output, receives the value read from the card
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_io_read_byte(sdmmc_card_t* card, uint32_t function,
+        uint32_t reg, uint8_t *out_byte);
+
+/**
+ * Write one byte to an SDIO card using IO_RW_DIRECT (CMD52)
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param function  IO function number
+ * @param reg  byte address within IO function
+ * @param in_byte  value to be written
+ * @param[out] out_byte  if not NULL, receives new byte value read
+ *                       from the card (read-after-write).
+ * @return
+ *      - ESP_OK on success
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_io_write_byte(sdmmc_card_t* card, uint32_t function,
+        uint32_t reg, uint8_t in_byte, uint8_t* out_byte);
+
+/**
+ * Read multiple bytes from an SDIO card using IO_RW_EXTENDED (CMD53)
+ *
+ * This function performs read operation using CMD53 in byte mode.
+ * For block mode, see sdmmc_io_read_blocks.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param function  IO function number
+ * @param addr  byte address within IO function where reading starts
+ * @param dst  buffer which receives the data read from card
+ * @param size  number of bytes to read
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_SIZE if size exceeds 512 bytes
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_io_read_bytes(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, void* dst, size_t size);
+
+/**
+ * Write multiple bytes to an SDIO card using IO_RW_EXTENDED (CMD53)
+ *
+ * This function performs write operation using CMD53 in byte mode.
+ * For block mode, see sdmmc_io_write_blocks.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param function  IO function number
+ * @param addr  byte address within IO function where writing starts
+ * @param src  data to be written
+ * @param size  number of bytes to write
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_SIZE if size exceeds 512 bytes
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_io_write_bytes(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, const void* src, size_t size);
+
+/**
+ * Read blocks of data from an SDIO card using IO_RW_EXTENDED (CMD53)
+ *
+ * This function performs read operation using CMD53 in block mode.
+ * For byte mode, see sdmmc_io_read_bytes.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param function  IO function number
+ * @param addr  byte address within IO function where writing starts
+ * @param dst  buffer which receives the data read from card
+ * @param size  number of bytes to read, must be divisible by the card block
+ *              size.
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_SIZE if size is not divisible by 512 bytes
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_io_read_blocks(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, void* dst, size_t size);
+
+/**
+ * Write blocks of data to an SDIO card using IO_RW_EXTENDED (CMD53)
+ *
+ * This function performs write operation using CMD53 in block mode.
+ * For byte mode, see sdmmc_io_write_bytes.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param function  IO function number
+ * @param addr  byte address within IO function where writing starts
+ * @param src  data to be written
+ * @param size  number of bytes to read, must be divisible by the card block
+ *              size.
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_INVALID_SIZE if size is not divisible by 512 bytes
+ *      - One of the error codes from SDMMC host controller
+ */
+esp_err_t sdmmc_io_write_blocks(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, const void* src, size_t size);
+
+/**
+ * Enable SDIO interrupt in the SDMMC host
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @return
+ *      - ESP_OK on success
+ *      - ESP_ERR_NOT_SUPPORTED if the host controller does not support
+ *        IO interrupts
+ */
+esp_err_t sdmmc_io_enable_int(sdmmc_card_t* card);
+
+/**
+ * Block until an SDIO interrupt is received
+ *
+ * Slave uses D1 line to signal interrupt condition to the host.
+ * This function can be used to wait for the interrupt.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param timeout_ticks  time to wait for the interrupt, in RTOS ticks
+ * @return
+ *      - ESP_OK if the interrupt is received
+ *      - ESP_ERR_NOT_SUPPORTED if the host controller does not support
+ *        IO interrupts
+ *      - ESP_ERR_TIMEOUT if the interrupt does not happen in timeout_ticks
+ */
+esp_err_t sdmmc_io_wait_int(sdmmc_card_t* card, TickType_t timeout_ticks);
+
+/**
+ * Get the data of CIS region of an SDIO card.
+ *
+ * You may provide a buffer not sufficient to store all the CIS data. In this
+ * case, this function stores as much data into your buffer as possible. Also,
+ * this function will try to get and return the size required for you.
+ *
+ * @param card  pointer to card information structure previously initialized
+ *              using sdmmc_card_init
+ * @param out_buffer Output buffer of the CIS data
+ * @param buffer_size Size of the buffer.
+ * @param inout_cis_size Mandatory, pointer to a size, input and output.
+ *              - input: Limitation of maximum searching range, should be 0 or larger than
+ *                      buffer_size. The function searches for CIS_CODE_END until this range. Set to
+ *                      0 to search infinitely.
+ *              - output: The size required to store all the CIS data, if CIS_CODE_END is found.
+ *
+ * @return
+ *      - ESP_OK: on success
+ *      - ESP_ERR_INVALID_RESPONSE: if the card does not (correctly) support CIS.
+ *      - ESP_ERR_INVALID_SIZE: CIS_CODE_END found, but buffer_size is less than
+ *              required size, which is stored in the inout_cis_size then.
+ *      - ESP_ERR_NOT_FOUND: if the CIS_CODE_END not found. Increase input value of
+ *              inout_cis_size or set it to 0, if you still want to search for the end;
+ *              output value of inout_cis_size is invalid in this case.
+ *      - and other error code return from sdmmc_io_read_bytes
+ */
+esp_err_t sdmmc_io_get_cis_data(sdmmc_card_t* card, uint8_t* out_buffer, size_t buffer_size, size_t* inout_cis_size);
+
+/**
+ * Parse and print the CIS information of an SDIO card.
+ *
+ * @note Not all the CIS codes and all kinds of tuples are supported. If you
+ * see some unresolved code, you can add the parsing of these code in
+ * sdmmc_io.c and contribute to the IDF through the Github repository.
+ *
+ *              using sdmmc_card_init
+ * @param buffer Buffer to parse
+ * @param buffer_size Size of the buffer.
+ * @param fp File pointer to print to, set to NULL to print to stdout.
+ *
+ * @return
+ *      - ESP_OK: on success
+ *      - ESP_ERR_NOT_SUPPORTED: if the value from the card is not supported to be parsed.
+ *      - ESP_ERR_INVALID_SIZE: if the CIS size fields are not correct.
+ */
+esp_err_t sdmmc_io_print_cis_info(uint8_t* buffer, size_t buffer_size, FILE* fp);
+
+
+#ifdef __cplusplus
+}
+#endif

code/components/esp-sdmmc/sdmmc_cmd_mh.c

@@ -0,0 +1,701 @@
+/*
+ * SPDX-FileCopyrightText: 2015-2021 Espressif Systems (Shanghai) CO LTD
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include "sdmmc_common_mh.h"
+
+int FF_CAN_TRIM = 0;
+
+static const char* TAG = "sdmmc_cmd";
+
+
+esp_err_t sdmmc_send_cmd(sdmmc_card_t* card, sdmmc_command_t* cmd)
+{
+    if (card->host.command_timeout_ms != 0) {
+        cmd->timeout_ms = card->host.command_timeout_ms;
+    } else if (cmd->timeout_ms == 0) {
+        cmd->timeout_ms = SDMMC_DEFAULT_CMD_TIMEOUT_MS;
+    }
+
+    int slot = card->host.slot;
+    ESP_LOGV(TAG, "sending cmd slot=%d op=%d arg=%x flags=%x data=%p blklen=%d datalen=%d timeout=%d",
+            slot, cmd->opcode, cmd->arg, cmd->flags, cmd->data, cmd->blklen, cmd->datalen, cmd->timeout_ms);
+    esp_err_t err = (*card->host.do_transaction)(slot, cmd);
+    if (err != 0) {
+        ESP_LOGD(TAG, "cmd=%d, sdmmc_req_run returned 0x%x", cmd->opcode, err);
+        return err;
+    }
+    int state = MMC_R1_CURRENT_STATE(cmd->response);
+    ESP_LOGV(TAG, "cmd response %08x %08x %08x %08x err=0x%x state=%d",
+               cmd->response[0],
+               cmd->response[1],
+               cmd->response[2],
+               cmd->response[3],
+               cmd->error,
+               state);
+    return cmd->error;
+}
+
+esp_err_t sdmmc_send_app_cmd(sdmmc_card_t* card, sdmmc_command_t* cmd)
+{
+    sdmmc_command_t app_cmd = {
+        .opcode = MMC_APP_CMD,
+        .flags = SCF_CMD_AC | SCF_RSP_R1,
+        .arg = MMC_ARG_RCA(card->rca),
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &app_cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    // Check APP_CMD status bit (only in SD mode)
+    if (!host_is_spi(card) && !(MMC_R1(app_cmd.response) & MMC_R1_APP_CMD)) {
+        ESP_LOGW(TAG, "card doesn't support APP_CMD");
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    return sdmmc_send_cmd(card, cmd);
+}
+
+
+esp_err_t sdmmc_send_cmd_go_idle_state(sdmmc_card_t* card)
+{
+    sdmmc_command_t cmd = {
+        .opcode = MMC_GO_IDLE_STATE,
+        .flags = SCF_CMD_BC | SCF_RSP_R0,
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (host_is_spi(card)) {
+        /* To enter SPI mode, CMD0 needs to be sent twice (see figure 4-1 in
+         * SD Simplified spec v4.10). Some cards enter SD mode on first CMD0,
+         * so don't expect the above command to succeed.
+         * SCF_RSP_R1 flag below tells the lower layer to expect correct R1
+         * response (in SPI mode).
+         */
+        (void) err;
+        vTaskDelay(SDMMC_GO_IDLE_DELAY_MS / portTICK_PERIOD_MS);
+
+        cmd.flags |= SCF_RSP_R1;
+        err = sdmmc_send_cmd(card, &cmd);
+    }
+    if (err == ESP_OK) {
+        vTaskDelay(SDMMC_GO_IDLE_DELAY_MS / portTICK_PERIOD_MS);
+    }
+    return err;
+}
+
+
+esp_err_t sdmmc_send_cmd_send_if_cond(sdmmc_card_t* card, uint32_t ocr)
+{
+    const uint8_t pattern = 0xaa; /* any pattern will do here */
+    sdmmc_command_t cmd = {
+        .opcode = SD_SEND_IF_COND,
+        .arg = (((ocr & SD_OCR_VOL_MASK) != 0) << 8) | pattern,
+        .flags = SCF_CMD_BCR | SCF_RSP_R7,
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    uint8_t response = cmd.response[0] & 0xff;
+    if (response != pattern) {
+        ESP_LOGD(TAG, "%s: received=0x%x expected=0x%x", __func__, response, pattern);
+        return ESP_ERR_INVALID_RESPONSE;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_send_cmd_send_op_cond(sdmmc_card_t* card, uint32_t ocr, uint32_t *ocrp)
+{
+    esp_err_t err;
+
+    sdmmc_command_t cmd = {
+            .arg = ocr,
+            .flags = SCF_CMD_BCR | SCF_RSP_R3,
+            .opcode = SD_APP_OP_COND
+    };
+    int nretries = SDMMC_SEND_OP_COND_MAX_RETRIES;
+    int err_cnt = SDMMC_SEND_OP_COND_MAX_ERRORS;
+    for (; nretries != 0; --nretries)  {
+        bzero(&cmd, sizeof cmd);
+        cmd.arg = ocr;
+        cmd.flags = SCF_CMD_BCR | SCF_RSP_R3;
+        if (!card->is_mmc) { /* SD mode */
+            cmd.opcode = SD_APP_OP_COND;
+            err = sdmmc_send_app_cmd(card, &cmd);
+        } else { /* MMC mode */
+            cmd.arg &= ~MMC_OCR_ACCESS_MODE_MASK;
+            cmd.arg |= MMC_OCR_SECTOR_MODE;
+            cmd.opcode = MMC_SEND_OP_COND;
+            err = sdmmc_send_cmd(card, &cmd);
+        }
+
+        if (err != ESP_OK) {
+            if (--err_cnt == 0) {
+                ESP_LOGD(TAG, "%s: sdmmc_send_app_cmd err=0x%x", __func__, err);
+                return err;
+            } else {
+                ESP_LOGV(TAG, "%s: ignoring err=0x%x", __func__, err);
+                continue;
+            }
+        }
+        // In SD protocol, card sets MEM_READY bit in OCR when it is ready.
+        // In SPI protocol, card clears IDLE_STATE bit in R1 response.
+        if (!host_is_spi(card)) {
+            if ((MMC_R3(cmd.response) & MMC_OCR_MEM_READY) ||
+                ocr == 0) {
+                break;
+            }
+        } else {
+            if ((SD_SPI_R1(cmd.response) & SD_SPI_R1_IDLE_STATE) == 0) {
+                break;
+            }
+        }
+        vTaskDelay(10 / portTICK_PERIOD_MS);
+    }
+    if (nretries == 0) {
+        return ESP_ERR_TIMEOUT;
+    }
+    if (ocrp) {
+        *ocrp = MMC_R3(cmd.response);
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_send_cmd_read_ocr(sdmmc_card_t *card, uint32_t *ocrp)
+{
+    assert(ocrp);
+    sdmmc_command_t cmd = {
+        .opcode = SD_READ_OCR,
+        .flags = SCF_CMD_BCR | SCF_RSP_R2
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    *ocrp = SD_SPI_R3(cmd.response);
+    return ESP_OK;
+}
+
+
+esp_err_t sdmmc_send_cmd_all_send_cid(sdmmc_card_t* card, sdmmc_response_t* out_raw_cid)
+{
+    assert(out_raw_cid);
+    sdmmc_command_t cmd = {
+            .opcode = MMC_ALL_SEND_CID,
+            .flags = SCF_CMD_BCR | SCF_RSP_R2
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    memcpy(out_raw_cid, &cmd.response, sizeof(sdmmc_response_t));
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_send_cmd_send_cid(sdmmc_card_t *card, sdmmc_cid_t *out_cid)
+{
+    assert(out_cid);
+    assert(host_is_spi(card) && "SEND_CID should only be used in SPI mode");
+    assert(!card->is_mmc && "MMC cards are not supported in SPI mode");
+    sdmmc_response_t buf;
+    sdmmc_command_t cmd = {
+        .opcode = MMC_SEND_CID,
+        .flags = SCF_CMD_READ | SCF_CMD_ADTC,
+        .arg = 0,
+        .data = &buf[0],
+        .datalen = sizeof(buf)
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    sdmmc_flip_byte_order(buf, sizeof(buf));
+    return sdmmc_decode_cid(buf, out_cid);
+}
+
+
+esp_err_t sdmmc_send_cmd_set_relative_addr(sdmmc_card_t* card, uint16_t* out_rca)
+{
+    assert(out_rca);
+    sdmmc_command_t cmd = {
+            .opcode = SD_SEND_RELATIVE_ADDR,
+            .flags = SCF_CMD_BCR | SCF_RSP_R6
+    };
+
+    /* MMC cards expect us to set the RCA.
+     * Set RCA to 1 since we don't support multiple cards on the same bus, for now.
+     */
+    uint16_t mmc_rca = 1;
+    if (card->is_mmc) {
+        cmd.arg = MMC_ARG_RCA(mmc_rca);
+    }
+
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    *out_rca = (card->is_mmc) ? mmc_rca : SD_R6_RCA(cmd.response);
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_send_cmd_set_blocklen(sdmmc_card_t* card, sdmmc_csd_t* csd)
+{
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SET_BLOCKLEN,
+            .arg = csd->sector_size,
+            .flags = SCF_CMD_AC | SCF_RSP_R1
+    };
+    return sdmmc_send_cmd(card, &cmd);
+}
+
+esp_err_t sdmmc_send_cmd_send_csd(sdmmc_card_t* card, sdmmc_csd_t* out_csd)
+{
+    /* The trick with SEND_CSD is that in SPI mode, it acts as a data read
+     * command, while in SD mode it is an AC command with R2 response.
+     */
+    sdmmc_response_t spi_buf;
+    const bool is_spi = host_is_spi(card);
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SEND_CSD,
+            .arg = is_spi ? 0 : MMC_ARG_RCA(card->rca),
+            .flags = is_spi ? (SCF_CMD_READ | SCF_CMD_ADTC | SCF_RSP_R1) :
+                    (SCF_CMD_AC | SCF_RSP_R2),
+            .data = is_spi ? &spi_buf[0] : 0,
+            .datalen = is_spi ? sizeof(spi_buf) : 0,
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    uint32_t* ptr = cmd.response;
+    if (is_spi) {
+        sdmmc_flip_byte_order(spi_buf,  sizeof(spi_buf));
+        ptr = spi_buf;
+    }
+    if (card->is_mmc) {
+        err = sdmmc_mmc_decode_csd(cmd.response, out_csd);
+    } else {
+        err = sdmmc_decode_csd(ptr, out_csd);
+    }
+    return err;
+}
+
+esp_err_t sdmmc_send_cmd_select_card(sdmmc_card_t* card, uint32_t rca)
+{
+    /* Don't expect to see a response when de-selecting a card */
+    uint32_t response = (rca == 0) ? 0 : SCF_RSP_R1;
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SELECT_CARD,
+            .arg = MMC_ARG_RCA(rca),
+            .flags = SCF_CMD_AC | response
+    };
+    return sdmmc_send_cmd(card, &cmd);
+}
+
+esp_err_t sdmmc_send_cmd_send_scr(sdmmc_card_t* card, sdmmc_scr_t *out_scr)
+{
+    size_t datalen = 8;
+    uint32_t* buf = (uint32_t*) heap_caps_malloc(datalen, MALLOC_CAP_DMA);
+    if (buf == NULL) {
+        return ESP_ERR_NO_MEM;
+    }
+    sdmmc_command_t cmd = {
+            .data = buf,
+            .datalen = datalen,
+            .blklen = datalen,
+            .flags = SCF_CMD_ADTC | SCF_CMD_READ | SCF_RSP_R1,
+            .opcode = SD_APP_SEND_SCR
+    };
+    esp_err_t err = sdmmc_send_app_cmd(card, &cmd);
+    if (err == ESP_OK) {
+        err = sdmmc_decode_scr(buf, out_scr);
+    }
+    free(buf);
+    return err;
+}
+
+esp_err_t sdmmc_send_cmd_set_bus_width(sdmmc_card_t* card, int width)
+{
+    sdmmc_command_t cmd = {
+            .opcode = SD_APP_SET_BUS_WIDTH,
+            .flags = SCF_RSP_R1 | SCF_CMD_AC,
+            .arg = (width == 4) ? SD_ARG_BUS_WIDTH_4 : SD_ARG_BUS_WIDTH_1,
+    };
+
+    return sdmmc_send_app_cmd(card, &cmd);
+}
+
+esp_err_t sdmmc_send_cmd_crc_on_off(sdmmc_card_t* card, bool crc_enable)
+{
+    assert(host_is_spi(card) && "CRC_ON_OFF can only be used in SPI mode");
+    sdmmc_command_t cmd = {
+            .opcode = SD_CRC_ON_OFF,
+            .arg = crc_enable ? 1 : 0,
+            .flags = SCF_CMD_AC | SCF_RSP_R1
+    };
+    return sdmmc_send_cmd(card, &cmd);
+}
+
+esp_err_t sdmmc_send_cmd_send_status(sdmmc_card_t* card, uint32_t* out_status)
+{
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SEND_STATUS,
+            .arg = MMC_ARG_RCA(card->rca),
+            .flags = SCF_CMD_AC | SCF_RSP_R1
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        return err;
+    }
+    if (out_status) {
+        if (host_is_spi(card)) {
+            *out_status = SD_SPI_R2(cmd.response);
+        } else {
+            *out_status = MMC_R1(cmd.response);
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_write_sectors(sdmmc_card_t* card, const void* src,
+        size_t start_block, size_t block_count)
+{
+    esp_err_t err = ESP_OK;
+    size_t block_size = card->csd.sector_size;
+    if (esp_ptr_dma_capable(src) && (intptr_t)src % 4 == 0) {
+        err = sdmmc_write_sectors_dma(card, src, start_block, block_count);
+    } else {
+        // SDMMC peripheral needs DMA-capable buffers. Split the write into
+        // separate single block writes, if needed, and allocate a temporary
+        // DMA-capable buffer.
+        void* tmp_buf = heap_caps_malloc(block_size, MALLOC_CAP_DMA);
+        if (tmp_buf == NULL) {
+            return ESP_ERR_NO_MEM;
+        }
+        const uint8_t* cur_src = (const uint8_t*) src;
+        for (size_t i = 0; i < block_count; ++i) {
+            memcpy(tmp_buf, cur_src, block_size);
+            cur_src += block_size;
+            err = sdmmc_write_sectors_dma(card, tmp_buf, start_block + i, 1);
+            if (err != ESP_OK) {
+                ESP_LOGD(TAG, "%s: error 0x%x writing block %d+%d",
+                        __func__, err, start_block, i);
+                break;
+            }
+        }
+        free(tmp_buf);
+    }
+    return err;
+}
+
+esp_err_t sdmmc_write_sectors_dma(sdmmc_card_t* card, const void* src,
+        size_t start_block, size_t block_count)
+{
+    if (start_block + block_count > card->csd.capacity) {
+        return ESP_ERR_INVALID_SIZE;
+    }
+    size_t block_size = card->csd.sector_size;
+    sdmmc_command_t cmd = {
+            .flags = SCF_CMD_ADTC | SCF_RSP_R1,
+            .blklen = block_size,
+            .data = (void*) src,
+            .datalen = block_count * block_size,
+            .timeout_ms = SDMMC_WRITE_CMD_TIMEOUT_MS
+    };
+    if (block_count == 1) {
+        cmd.opcode = MMC_WRITE_BLOCK_SINGLE;
+    } else {
+        cmd.opcode = MMC_WRITE_BLOCK_MULTIPLE;
+    }
+    if (card->ocr & SD_OCR_SDHC_CAP) {
+        cmd.arg = start_block;
+    } else {
+        cmd.arg = start_block * block_size;
+    }
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd returned 0x%x", __func__, err);
+        return err;
+    }
+    uint32_t status = 0;
+    size_t count = 0;
+    /* SD mode: wait for the card to become idle based on R1 status */
+    while (!host_is_spi(card) && !(status & MMC_R1_READY_FOR_DATA)) {
+        // TODO: add some timeout here
+        err = sdmmc_send_cmd_send_status(card, &status);
+        if (err != ESP_OK) {
+            return err;
+        }
+        if (++count % 10 == 0) {
+            ESP_LOGV(TAG, "waiting for card to become ready (%d)", count);
+        }
+    }
+    /* SPI mode: although card busy indication is based on the busy token,
+     * SD spec recommends that the host checks the results of programming by sending
+     * SEND_STATUS command. Some of the conditions reported in SEND_STATUS are not
+     * reported via a data error token.
+     */
+    if (host_is_spi(card)) {
+        err = sdmmc_send_cmd_send_status(card, &status);
+        if (err != ESP_OK) {
+            return err;
+        }
+        if (status & SD_SPI_R2_CARD_LOCKED) {
+            ESP_LOGE(TAG, "%s: write failed, card is locked: r2=0x%04x",
+                     __func__, status);
+            return ESP_ERR_INVALID_STATE;
+        }
+        if (status != 0) {
+            ESP_LOGE(TAG, "%s: card status indicates an error after write operation: r2=0x%04x",
+                     __func__, status);
+            return ESP_ERR_INVALID_RESPONSE;
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_read_sectors(sdmmc_card_t* card, void* dst,
+        size_t start_block, size_t block_count)
+{
+    esp_err_t err = ESP_OK;
+    size_t block_size = card->csd.sector_size;
+    if (esp_ptr_dma_capable(dst) && (intptr_t)dst % 4 == 0) {
+        err = sdmmc_read_sectors_dma(card, dst, start_block, block_count);
+    } else {
+        // SDMMC peripheral needs DMA-capable buffers. Split the read into
+        // separate single block reads, if needed, and allocate a temporary
+        // DMA-capable buffer.
+        void* tmp_buf = heap_caps_malloc(block_size, MALLOC_CAP_DMA);
+        if (tmp_buf == NULL) {
+            return ESP_ERR_NO_MEM;
+        }
+        uint8_t* cur_dst = (uint8_t*) dst;
+        for (size_t i = 0; i < block_count; ++i) {
+            err = sdmmc_read_sectors_dma(card, tmp_buf, start_block + i, 1);
+            if (err != ESP_OK) {
+                ESP_LOGD(TAG, "%s: error 0x%x writing block %d+%d",
+                        __func__, err, start_block, i);
+                break;
+            }
+            memcpy(cur_dst, tmp_buf, block_size);
+            cur_dst += block_size;
+        }
+        free(tmp_buf);
+    }
+    return err;
+}
+
+esp_err_t sdmmc_read_sectors_dma(sdmmc_card_t* card, void* dst,
+        size_t start_block, size_t block_count)
+{
+    if (start_block + block_count > card->csd.capacity) {
+        return ESP_ERR_INVALID_SIZE;
+    }
+    size_t block_size = card->csd.sector_size;
+    sdmmc_command_t cmd = {
+            .flags = SCF_CMD_ADTC | SCF_CMD_READ | SCF_RSP_R1,
+            .blklen = block_size,
+            .data = (void*) dst,
+            .datalen = block_count * block_size
+    };
+    if (block_count == 1) {
+        cmd.opcode = MMC_READ_BLOCK_SINGLE;
+    } else {
+        cmd.opcode = MMC_READ_BLOCK_MULTIPLE;
+    }
+    if (card->ocr & SD_OCR_SDHC_CAP) {
+        cmd.arg = start_block;
+    } else {
+        cmd.arg = start_block * block_size;
+    }
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd returned 0x%x", __func__, err);
+        return err;
+    }
+    uint32_t status = 0;
+    size_t count = 0;
+    while (!host_is_spi(card) && !(status & MMC_R1_READY_FOR_DATA)) {
+        // TODO: add some timeout here
+        err = sdmmc_send_cmd_send_status(card, &status);
+        if (err != ESP_OK) {
+            return err;
+        }
+        if (++count % 10 == 0) {
+            ESP_LOGV(TAG, "waiting for card to become ready (%d)", count);
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_erase_sectors(sdmmc_card_t* card, size_t start_sector,
+        size_t sector_count, sdmmc_erase_arg_t arg)
+{
+    if (start_sector + sector_count > card->csd.capacity) {
+        return ESP_ERR_INVALID_SIZE;
+    }
+
+    uint32_t cmd38_arg;
+    if (arg == SDMMC_ERASE_ARG) {
+        cmd38_arg = card->is_mmc ? SDMMC_MMC_TRIM_ARG : SDMMC_SD_ERASE_ARG;
+    } else {
+        cmd38_arg = card->is_mmc ? SDMMC_MMC_DISCARD_ARG : SDMMC_SD_DISCARD_ARG;
+    }
+
+    /* validate the CMD38 argument against card supported features */
+    if (card->is_mmc) {
+        if ((cmd38_arg == SDMMC_MMC_TRIM_ARG) && (sdmmc_can_trim(card) != ESP_OK)) {
+            return ESP_ERR_NOT_SUPPORTED;
+        }
+        if ((cmd38_arg == SDMMC_MMC_DISCARD_ARG) && (sdmmc_can_discard(card) != ESP_OK)) {
+            return ESP_ERR_NOT_SUPPORTED;
+        }
+    } else { // SD card
+        if ((cmd38_arg == SDMMC_SD_DISCARD_ARG) && (sdmmc_can_discard(card) != ESP_OK)) {
+            return ESP_ERR_NOT_SUPPORTED;
+        }
+    }
+
+    /* default as block unit address */
+    size_t addr_unit_mult = 1;
+
+    if (!(card->ocr & SD_OCR_SDHC_CAP)) {
+        addr_unit_mult = card->csd.sector_size;
+    }
+
+    /* prepare command to set the start address */
+    sdmmc_command_t cmd = {
+            .flags = SCF_CMD_AC | SCF_RSP_R1 | SCF_WAIT_BUSY,
+            .opcode = card->is_mmc ? MMC_ERASE_GROUP_START :
+                    SD_ERASE_GROUP_START,
+            .arg = (start_sector * addr_unit_mult),
+    };
+
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd (ERASE_GROUP_START) returned 0x%x", __func__, err);
+        return err;
+    }
+
+    /* prepare command to set the end address */
+    cmd.opcode = card->is_mmc ? MMC_ERASE_GROUP_END : SD_ERASE_GROUP_END;
+    cmd.arg = ((start_sector + (sector_count - 1)) * addr_unit_mult);
+
+    err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd (ERASE_GROUP_END) returned 0x%x", __func__, err);
+        return err;
+    }
+
+    /* issue erase command */
+    memset((void *)&cmd, 0 , sizeof(sdmmc_command_t));
+    cmd.flags = SCF_CMD_AC | SCF_RSP_R1B | SCF_WAIT_BUSY;
+    cmd.opcode = MMC_ERASE;
+    cmd.arg = cmd38_arg;
+    cmd.timeout_ms = sdmmc_get_erase_timeout_ms(card, cmd38_arg, sector_count * card->csd.sector_size / 1024);
+
+    err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd (ERASE) returned 0x%x", __func__, err);
+        return err;
+    }
+
+    if (host_is_spi(card)) {
+        uint32_t status;
+        err = sdmmc_send_cmd_send_status(card, &status);
+        if (err != ESP_OK) {
+            return err;
+        }
+        if (status != 0) {
+            ESP_LOGE(TAG, "%s: card status indicates an error after erase operation: r2=0x%04x",
+                     __func__, status);
+            return ESP_ERR_INVALID_RESPONSE;
+        }
+    }
+
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_can_discard(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.rev >= EXT_CSD_REV_1_6)) {
+         return ESP_OK;
+    }
+    // SD card
+    if ((!card->is_mmc) && !host_is_spi(card) && (card->ssr.discard_support == 1)) {
+        return ESP_OK;
+    }
+    return ESP_FAIL;
+}
+
+esp_err_t sdmmc_can_trim(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.sec_feature & EXT_CSD_SEC_GB_CL_EN)) {
+        FF_CAN_TRIM = 1;
+        return ESP_OK;
+    }
+    FF_CAN_TRIM = 0;
+    return ESP_FAIL;
+}
+
+esp_err_t sdmmc_mmc_can_sanitize(sdmmc_card_t* card)
+{
+    if ((card->is_mmc) && (card->ext_csd.sec_feature & EXT_CSD_SEC_SANITIZE)) {
+        return ESP_OK;
+    }
+    return ESP_FAIL;
+}
+
+esp_err_t sdmmc_mmc_sanitize(sdmmc_card_t* card, uint32_t timeout_ms)
+{
+    esp_err_t err;
+    uint8_t index = EXT_CSD_SANITIZE_START;
+    uint8_t set = EXT_CSD_CMD_SET_NORMAL;
+    uint8_t value = 0x01;
+
+    if (sdmmc_mmc_can_sanitize(card) != ESP_OK) {
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    /*
+     * A Sanitize operation is initiated by writing a value to the extended
+     * CSD[165] SANITIZE_START. While the device is performing the sanitize
+     * operation, the busy line is asserted.
+     * SWITCH command is used to write the EXT_CSD register.
+     */
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SWITCH,
+            .arg = (MMC_SWITCH_MODE_WRITE_BYTE << 24) | (index << 16) | (value << 8) | set,
+            .flags = SCF_RSP_R1B | SCF_CMD_AC | SCF_WAIT_BUSY,
+            .timeout_ms = timeout_ms,
+    };
+    err = sdmmc_send_cmd(card, &cmd);
+    if (err == ESP_OK) {
+        //check response bit to see that switch was accepted
+        if (MMC_R1(cmd.response) & MMC_R1_SWITCH_ERROR) {
+            err = ESP_ERR_INVALID_RESPONSE;
+        }
+    }
+    return err;
+}
+
+esp_err_t sdmmc_full_erase(sdmmc_card_t* card)
+{
+    sdmmc_erase_arg_t arg = SDMMC_SD_ERASE_ARG; // erase by default for SD card
+    esp_err_t err;
+    if (card->is_mmc) {
+        arg = sdmmc_mmc_can_sanitize(card) == ESP_OK ? SDMMC_MMC_DISCARD_ARG: SDMMC_MMC_TRIM_ARG;
+    }
+    err = sdmmc_erase_sectors(card, 0, card->csd.capacity, arg);
+    if ((err == ESP_OK) && (arg == SDMMC_MMC_DISCARD_ARG)) {
+        uint32_t timeout_ms = sdmmc_get_erase_timeout_ms(card, SDMMC_MMC_DISCARD_ARG, card->csd.capacity * ((uint64_t) card->csd.sector_size) / 1024);
+        return sdmmc_mmc_sanitize(card, timeout_ms);
+    }
+    return err;
+}
+
+esp_err_t sdmmc_get_status(sdmmc_card_t* card)
+{
+    uint32_t stat;
+    return sdmmc_send_cmd_send_status(card, &stat);
+}

code/components/esp-sdmmc/sdmmc_common_mh.c

@@ -0,0 +1,332 @@
+/*
+ * Copyright (c) 2006 Uwe Stuehler <uwe@openbsd.org>
+ * Adaptations to ESP-IDF Copyright (c) 2016-2018 Espressif Systems (Shanghai) PTE LTD
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include "sdmmc_common_mh.h"
+
+bool card_is_mmc = 0;
+
+static const char* TAG = "sdmmc_common";
+
+esp_err_t sdmmc_init_ocr(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    /* In SPI mode, READ_OCR (CMD58) command is used to figure out which voltage
+     * ranges the card can support. This step is skipped since 1.8V isn't
+     * supported on the ESP32.
+     */
+
+    uint32_t host_ocr = get_host_ocr(card->host.io_voltage);
+    if ((card->ocr & SD_OCR_SDHC_CAP) != 0) {
+        host_ocr |= SD_OCR_SDHC_CAP;
+    }
+    /* Send SEND_OP_COND (ACMD41) command to the card until it becomes ready. */
+    err = sdmmc_send_cmd_send_op_cond(card, host_ocr, &card->ocr);
+
+    /* If time-out, re-try send_op_cond as MMC */
+    if (err == ESP_ERR_TIMEOUT && !host_is_spi(card)) {
+        ESP_LOGD(TAG, "send_op_cond timeout, trying MMC");
+        card->is_mmc = 1;
+        card_is_mmc = 1;
+        err = sdmmc_send_cmd_send_op_cond(card, host_ocr, &card->ocr);
+    }
+
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_op_cond (1) returned 0x%x", __func__, err);
+        return err;
+    }
+    if (host_is_spi(card)) {
+        err = sdmmc_send_cmd_read_ocr(card, &card->ocr);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: read_ocr returned 0x%x", __func__, err);
+            return err;
+        }
+    }
+    ESP_LOGD(TAG, "host_ocr=0x%x card_ocr=0x%x", host_ocr, card->ocr);
+
+    /* Clear all voltage bits in host's OCR which the card doesn't support.
+     * Don't touch CCS bit because in SPI mode cards don't report CCS in ACMD41
+     * response.
+     */
+    host_ocr &= (card->ocr | (~SD_OCR_VOL_MASK));
+    ESP_LOGD(TAG, "sdmmc_card_init: host_ocr=%08x, card_ocr=%08x", host_ocr, card->ocr);
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_cid(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    sdmmc_response_t raw_cid;
+    if (!host_is_spi(card)) {
+        err = sdmmc_send_cmd_all_send_cid(card, &raw_cid);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: all_send_cid returned 0x%x", __func__, err);
+            return err;
+        }
+        if (!card->is_mmc) {
+            err = sdmmc_decode_cid(raw_cid, &card->cid);
+            if (err != ESP_OK) {
+                ESP_LOGE(TAG, "%s: decoding CID failed (0x%x)", __func__, err);
+                return err;
+            }
+        } else {
+            /* For MMC, need to know CSD to decode CID. But CSD can only be read
+             * in data transfer mode, and it is not possible to read CID in data
+             * transfer mode. We temporiliy store the raw cid and do the
+             * decoding after the RCA is set and the card is in data transfer
+             * mode.
+             */
+            memcpy(card->raw_cid, raw_cid, sizeof(sdmmc_response_t));
+        }
+    } else {
+        err = sdmmc_send_cmd_send_cid(card, &card->cid);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: send_cid returned 0x%x", __func__, err);
+            return err;
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_rca(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    err = sdmmc_send_cmd_set_relative_addr(card, &card->rca);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: set_relative_addr returned 0x%x", __func__, err);
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_mmc_decode_cid(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    sdmmc_response_t raw_cid;
+    memcpy(raw_cid, card->raw_cid, sizeof(raw_cid));
+    err = sdmmc_mmc_decode_cid(card->csd.mmc_ver, raw_cid, &card->cid);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: decoding CID failed (0x%x)", __func__, err);
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_csd(sdmmc_card_t* card)
+{
+    assert(card->is_mem == 1);
+    /* Get and decode the contents of CSD register. Determine card capacity. */
+    esp_err_t err = sdmmc_send_cmd_send_csd(card, &card->csd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_csd returned 0x%x", __func__, err);
+        return err;
+    }
+    const size_t max_sdsc_capacity = UINT32_MAX / card->csd.sector_size + 1;
+    if (!(card->ocr & SD_OCR_SDHC_CAP) &&
+         card->csd.capacity > max_sdsc_capacity) {
+        ESP_LOGW(TAG, "%s: SDSC card reports capacity=%u. Limiting to %u.",
+                __func__, card->csd.capacity, max_sdsc_capacity);
+        card->csd.capacity = max_sdsc_capacity;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_select_card(sdmmc_card_t* card)
+{
+    assert(!host_is_spi(card));
+    esp_err_t err = sdmmc_send_cmd_select_card(card, card->rca);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: select_card returned 0x%x", __func__, err);
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_card_hs_mode(sdmmc_card_t* card)
+{
+    esp_err_t err = ESP_ERR_NOT_SUPPORTED;
+    if (card->is_mem && !card->is_mmc) {
+        err = sdmmc_enable_hs_mode_and_check(card);
+    } else if (card->is_sdio) {
+        err = sdmmc_io_enable_hs_mode(card);
+    } else if (card->is_mmc){
+        err = sdmmc_mmc_enable_hs_mode(card);
+    }
+    if (err == ESP_ERR_NOT_SUPPORTED) {
+        ESP_LOGD(TAG, "%s: host supports HS mode, but card doesn't", __func__);
+        card->max_freq_khz = SDMMC_FREQ_DEFAULT;
+    } else if (err != ESP_OK) {
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_host_bus_width(sdmmc_card_t* card)
+{
+    int bus_width = 1;
+
+    if ((card->host.flags & SDMMC_HOST_FLAG_4BIT) &&
+        (card->log_bus_width == 2)) {
+        bus_width = 4;
+    } else if ((card->host.flags & SDMMC_HOST_FLAG_8BIT) &&
+        (card->log_bus_width == 3)) {
+        bus_width = 8;
+    }
+    ESP_LOGD(TAG, "%s: using %d-bit bus", __func__, bus_width);
+    if (bus_width > 1) {
+        esp_err_t err = (*card->host.set_bus_width)(card->host.slot, bus_width);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "host.set_bus_width failed (0x%x)", err);
+            return err;
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_host_frequency(sdmmc_card_t* card)
+{
+    assert(card->max_freq_khz <= card->host.max_freq_khz);
+
+    /* Find highest frequency in the following list,
+     * which is below card->max_freq_khz.
+     */
+    const uint32_t freq_values[] = {
+            SDMMC_FREQ_52M,
+            SDMMC_FREQ_HIGHSPEED,
+            SDMMC_FREQ_26M,
+            SDMMC_FREQ_DEFAULT
+            //NOTE: in sdspi mode, 20MHz may not work. in that case, add 10MHz here.
+    };
+    const int n_freq_values = sizeof(freq_values) / sizeof(freq_values[0]);
+
+    uint32_t selected_freq = SDMMC_FREQ_PROBING;
+    for (int i = 0; i < n_freq_values; ++i) {
+        uint32_t freq = freq_values[i];
+        if (card->max_freq_khz >= freq) {
+            selected_freq = freq;
+            break;
+        }
+    }
+
+    ESP_LOGD(TAG, "%s: using %d kHz bus frequency", __func__, selected_freq);
+    if (selected_freq > SDMMC_FREQ_PROBING) {
+        esp_err_t err = (*card->host.set_card_clk)(card->host.slot, selected_freq);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "failed to switch bus frequency (0x%x)", err);
+            return err;
+        }
+    }
+
+    if (card->is_ddr) {
+        if (card->host.set_bus_ddr_mode == NULL) {
+            ESP_LOGE(TAG, "host doesn't support DDR mode or voltage switching");
+            return ESP_ERR_NOT_SUPPORTED;
+        }
+        esp_err_t err = (*card->host.set_bus_ddr_mode)(card->host.slot, true);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "failed to switch bus to DDR mode (0x%x)", err);
+            return err;
+        }
+    }
+    return ESP_OK;
+}
+
+void sdmmc_flip_byte_order(uint32_t* response, size_t size)
+{
+    assert(size % (2 * sizeof(uint32_t)) == 0);
+    const size_t n_words = size / sizeof(uint32_t);
+    for (int i = 0; i < n_words / 2; ++i) {
+        uint32_t left = __builtin_bswap32(response[i]);
+        uint32_t right = __builtin_bswap32(response[n_words - i - 1]);
+        response[i] = right;
+        response[n_words - i - 1] = left;
+    }
+}
+
+void sdmmc_card_print_info(FILE* stream, const sdmmc_card_t* card)
+{
+    bool print_scr = false;
+    bool print_csd = false;
+    const char* type;
+    fprintf(stream, "Name: %s\n", card->cid.name);
+    if (card->is_sdio) {
+        type = "SDIO";
+        print_scr = true;
+        print_csd = true;
+    } else if (card->is_mmc) {
+        type = "MMC";
+        print_csd = true;
+    } else {
+        type = (card->ocr & SD_OCR_SDHC_CAP) ? "SDHC/SDXC" : "SDSC";
+        print_csd = true;
+    }
+    fprintf(stream, "Type: %s\n", type);
+    if (card->max_freq_khz < 1000) {
+        fprintf(stream, "Speed: %d kHz\n", card->max_freq_khz);
+    } else {
+        fprintf(stream, "Speed: %d MHz%s\n", card->max_freq_khz / 1000,
+                card->is_ddr ? ", DDR" : "");
+    }
+    fprintf(stream, "Size: %lluMB\n", ((uint64_t) card->csd.capacity) * card->csd.sector_size / (1024 * 1024));
+
+    if (print_csd) {
+        fprintf(stream, "CSD: ver=%d, sector_size=%d, capacity=%d read_bl_len=%d\n",
+                (card->is_mmc ? card->csd.csd_ver : card->csd.csd_ver + 1),
+                card->csd.sector_size, card->csd.capacity, card->csd.read_block_len);
+        if (card->is_mmc) {
+            fprintf(stream, "EXT CSD: bus_width=%d\n", (1 << card->log_bus_width));
+        } else if (!card->is_sdio){ // make sure card is SD
+            fprintf(stream, "SSR: bus_width=%d\n", (card->ssr.cur_bus_width ? 4 : 1));
+        }
+    }
+    if (print_scr) {
+        fprintf(stream, "SCR: sd_spec=%d, bus_width=%d\n", card->scr.sd_spec, card->scr.bus_width);
+    }
+}
+
+esp_err_t sdmmc_fix_host_flags(sdmmc_card_t* card)
+{
+    const uint32_t width_1bit = SDMMC_HOST_FLAG_1BIT;
+    const uint32_t width_4bit = SDMMC_HOST_FLAG_4BIT;
+    const uint32_t width_8bit = SDMMC_HOST_FLAG_8BIT;
+    const uint32_t width_mask = width_1bit | width_4bit | width_8bit;
+
+    int slot_bit_width = card->host.get_bus_width(card->host.slot);
+    if (slot_bit_width == 1 &&
+            (card->host.flags & (width_4bit | width_8bit))) {
+        card->host.flags &= ~width_mask;
+        card->host.flags |= width_1bit;
+    } else if (slot_bit_width == 4 && (card->host.flags & width_8bit)) {
+        if ((card->host.flags & width_4bit) == 0) {
+            ESP_LOGW(TAG, "slot width set to 4, but host flags don't have 4 line mode enabled; using 1 line mode");
+            card->host.flags &= ~width_mask;
+            card->host.flags |= width_1bit;
+        } else {
+            card->host.flags &= ~width_mask;
+            card->host.flags |= width_4bit;
+        }
+    }
+    return ESP_OK;
+}
+
+uint32_t sdmmc_get_erase_timeout_ms(const sdmmc_card_t* card, int arg, size_t erase_size_kb)
+{
+    if (card->is_mmc) {
+        return sdmmc_mmc_get_erase_timeout_ms(card, arg, erase_size_kb);
+    } else {
+        return sdmmc_sd_get_erase_timeout_ms(card, arg, erase_size_kb);
+    }
+}

code/components/esp-sdmmc/sdmmc_common_mh.h

@@ -0,0 +1,154 @@
+/*
+ * Copyright (c) 2006 Uwe Stuehler <uwe@openbsd.org>
+ * Adaptations to ESP-IDF Copyright (c) 2016-2018 Espressif Systems (Shanghai) PTE LTD
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+#pragma once
+
+#include <string.h>
+#include "esp_log.h"
+#include "esp_heap_caps.h"
+#include "freertos/FreeRTOS.h"
+#include "freertos/task.h"
+#include "driver/sdmmc_defs.h"
+#include "driver/sdmmc_types.h"
+#include "sdmmc_cmd_mh.h"
+#include "sys/param.h"
+#include "soc/soc_memory_layout.h"
+
+extern bool card_is_mmc;
+
+#define SDMMC_GO_IDLE_DELAY_MS              20
+#define SDMMC_IO_SEND_OP_COND_DELAY_MS      10
+
+/* These delay values are mostly useful for cases when CD pin is not used, and
+ * the card is removed. In this case, SDMMC peripheral may not always return
+ * CMD_DONE / DATA_DONE interrupts after signaling the error. These timeouts work
+ * as a safety net in such cases.
+ */
+#define SDMMC_DEFAULT_CMD_TIMEOUT_MS  1000   // Max timeout of ordinary commands
+#define SDMMC_WRITE_CMD_TIMEOUT_MS    5000   // Max timeout of write commands
+
+
+#define SDMMC_SD_DISCARD_TIMEOUT  250    // SD erase (discard) timeout
+
+/* Maximum retry/error count for SEND_OP_COND (CMD1).
+ * These are somewhat arbitrary, values originate from OpenBSD driver.
+ */
+#define SDMMC_SEND_OP_COND_MAX_RETRIES  100
+#define SDMMC_SEND_OP_COND_MAX_ERRORS   3
+
+/* supported arguments for erase command 38 */
+#define SDMMC_SD_ERASE_ARG      0
+#define SDMMC_SD_DISCARD_ARG    1
+#define SDMMC_MMC_TRIM_ARG      1
+#define SDMMC_MMC_DISCARD_ARG   3
+
+/* Functions to send individual commands */
+esp_err_t sdmmc_send_cmd(sdmmc_card_t* card, sdmmc_command_t* cmd);
+esp_err_t sdmmc_send_app_cmd(sdmmc_card_t* card, sdmmc_command_t* cmd);
+esp_err_t sdmmc_send_cmd_go_idle_state(sdmmc_card_t* card);
+esp_err_t sdmmc_send_cmd_send_if_cond(sdmmc_card_t* card, uint32_t ocr);
+esp_err_t sdmmc_send_cmd_send_op_cond(sdmmc_card_t* card, uint32_t ocr, uint32_t *ocrp);
+esp_err_t sdmmc_send_cmd_read_ocr(sdmmc_card_t *card, uint32_t *ocrp);
+esp_err_t sdmmc_send_cmd_send_cid(sdmmc_card_t *card, sdmmc_cid_t *out_cid);
+esp_err_t sdmmc_send_cmd_all_send_cid(sdmmc_card_t* card, sdmmc_response_t* out_raw_cid);
+esp_err_t sdmmc_send_cmd_set_relative_addr(sdmmc_card_t* card, uint16_t* out_rca);
+esp_err_t sdmmc_send_cmd_set_blocklen(sdmmc_card_t* card, sdmmc_csd_t* csd);
+esp_err_t sdmmc_send_cmd_switch_func(sdmmc_card_t* card,
+        uint32_t mode, uint32_t group, uint32_t function,
+        sdmmc_switch_func_rsp_t* resp);
+esp_err_t sdmmc_send_cmd_send_csd(sdmmc_card_t* card, sdmmc_csd_t* out_csd);
+esp_err_t sdmmc_send_cmd_select_card(sdmmc_card_t* card, uint32_t rca);
+esp_err_t sdmmc_send_cmd_send_scr(sdmmc_card_t* card, sdmmc_scr_t *out_scr);
+esp_err_t sdmmc_send_cmd_set_bus_width(sdmmc_card_t* card, int width);
+esp_err_t sdmmc_send_cmd_send_status(sdmmc_card_t* card, uint32_t* out_status);
+esp_err_t sdmmc_send_cmd_crc_on_off(sdmmc_card_t* card, bool crc_enable);
+
+/* Higher level functions */
+esp_err_t sdmmc_enable_hs_mode(sdmmc_card_t* card);
+esp_err_t sdmmc_enable_hs_mode_and_check(sdmmc_card_t* card);
+esp_err_t sdmmc_write_sectors_dma(sdmmc_card_t* card, const void* src,
+        size_t start_block, size_t block_count);
+esp_err_t sdmmc_read_sectors_dma(sdmmc_card_t* card, void* dst,
+        size_t start_block, size_t block_count);
+uint32_t sdmmc_get_erase_timeout_ms(const sdmmc_card_t* card, int arg, size_t erase_size_kb);
+
+/* SD specific */
+esp_err_t sdmmc_check_scr(sdmmc_card_t* card);
+esp_err_t sdmmc_decode_cid(sdmmc_response_t resp, sdmmc_cid_t* out_cid);
+esp_err_t sdmmc_decode_csd(sdmmc_response_t response, sdmmc_csd_t* out_csd);
+esp_err_t sdmmc_decode_scr(uint32_t *raw_scr, sdmmc_scr_t* out_scr);
+esp_err_t sdmmc_decode_ssr(uint32_t *raw_ssr, sdmmc_ssr_t* out_ssr);
+uint32_t sdmmc_sd_get_erase_timeout_ms(const sdmmc_card_t* card, int arg, size_t erase_size_kb);
+
+/* SDIO specific */
+esp_err_t sdmmc_io_reset(sdmmc_card_t* card);
+esp_err_t sdmmc_io_enable_hs_mode(sdmmc_card_t* card);
+esp_err_t sdmmc_io_send_op_cond(sdmmc_card_t* card, uint32_t ocr, uint32_t *ocrp);
+esp_err_t sdmmc_io_rw_direct(sdmmc_card_t* card, int function,
+        uint32_t reg, uint32_t arg, uint8_t *byte);
+esp_err_t sdmmc_io_rw_extended(sdmmc_card_t* card, int function,
+        uint32_t reg, int arg, void *data, size_t size);
+
+
+/* MMC specific */
+esp_err_t sdmmc_mmc_send_ext_csd_data(sdmmc_card_t* card, void *out_data, size_t datalen);
+esp_err_t sdmmc_mmc_switch(sdmmc_card_t* card, uint8_t set, uint8_t index, uint8_t value);
+esp_err_t sdmmc_mmc_decode_cid(int mmc_ver, sdmmc_response_t resp, sdmmc_cid_t* out_cid);
+esp_err_t sdmmc_mmc_decode_csd(sdmmc_response_t response, sdmmc_csd_t* out_csd);
+esp_err_t sdmmc_mmc_enable_hs_mode(sdmmc_card_t* card);
+uint32_t sdmmc_mmc_get_erase_timeout_ms(const sdmmc_card_t* card, int arg, size_t erase_size_kb);
+
+/* Parts of card initialization flow */
+esp_err_t sdmmc_init_sd_if_cond(sdmmc_card_t* card);
+esp_err_t sdmmc_init_select_card(sdmmc_card_t* card);
+esp_err_t sdmmc_init_csd(sdmmc_card_t* card);
+esp_err_t sdmmc_init_cid(sdmmc_card_t* card);
+esp_err_t sdmmc_init_rca(sdmmc_card_t* card);
+esp_err_t sdmmc_init_mmc_decode_cid(sdmmc_card_t* card);
+esp_err_t sdmmc_init_ocr(sdmmc_card_t* card);
+esp_err_t sdmmc_init_spi_crc(sdmmc_card_t* card);
+esp_err_t sdmmc_init_io(sdmmc_card_t* card);
+esp_err_t sdmmc_init_sd_blocklen(sdmmc_card_t* card);
+esp_err_t sdmmc_init_sd_scr(sdmmc_card_t* card);
+esp_err_t sdmmc_init_sd_ssr(sdmmc_card_t* card);
+esp_err_t sdmmc_init_sd_wait_data_ready(sdmmc_card_t* card);
+esp_err_t sdmmc_init_mmc_read_ext_csd(sdmmc_card_t* card);
+esp_err_t sdmmc_init_mmc_read_cid(sdmmc_card_t* card);
+esp_err_t sdmmc_init_host_bus_width(sdmmc_card_t* card);
+esp_err_t sdmmc_init_sd_bus_width(sdmmc_card_t* card);
+esp_err_t sdmmc_init_io_bus_width(sdmmc_card_t* card);
+esp_err_t sdmmc_init_mmc_bus_width(sdmmc_card_t* card);
+esp_err_t sdmmc_init_card_hs_mode(sdmmc_card_t* card);
+esp_err_t sdmmc_init_host_frequency(sdmmc_card_t* card);
+esp_err_t sdmmc_init_mmc_check_ext_csd(sdmmc_card_t* card);
+
+/* Various helper functions */
+static inline bool host_is_spi(const sdmmc_card_t* card)
+{
+    return (card->host.flags & SDMMC_HOST_FLAG_SPI) != 0;
+}
+
+static inline uint32_t get_host_ocr(float voltage)
+{
+    // TODO: report exact voltage to the card
+    // For now tell that the host has 2.8-3.6V voltage range
+    (void) voltage;
+    return SD_OCR_VOL_MASK;
+}
+
+void sdmmc_flip_byte_order(uint32_t* response, size_t size);
+
+esp_err_t sdmmc_fix_host_flags(sdmmc_card_t* card);

code/components/esp-sdmmc/sdmmc_init_mh.c

@@ -0,0 +1,128 @@
+/*
+ * Copyright (c) 2006 Uwe Stuehler <uwe@openbsd.org>
+ * Adaptations to ESP-IDF Copyright (c) 2016-2018 Espressif Systems (Shanghai) PTE LTD
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include "sdmmc_common_mh.h"
+
+static const char* TAG = "sdmmc_init";
+
+#define SDMMC_INIT_STEP(condition, function) \
+    do { \
+        if ((condition)) { \
+            esp_err_t err = (function)(card); \
+            if (err != ESP_OK) { \
+                ESP_LOGD(TAG, "%s: %s returned 0x%x", __func__, #function, err); \
+                return err; \
+            } \
+        } \
+    } while(0);
+
+
+esp_err_t sdmmc_card_init(const sdmmc_host_t* config, sdmmc_card_t* card)
+{
+    memset(card, 0, sizeof(*card));
+    memcpy(&card->host, config, sizeof(*config));
+    const bool is_spi = host_is_spi(card);
+    const bool always = true;
+    const bool io_supported = true;
+
+    /* Check if host flags are compatible with slot configuration. */
+    SDMMC_INIT_STEP(!is_spi, sdmmc_fix_host_flags);
+
+    /* Reset SDIO (CMD52, RES) before re-initializing IO (CMD5). */
+    SDMMC_INIT_STEP(io_supported, sdmmc_io_reset);
+
+    /* GO_IDLE_STATE (CMD0) command resets the card */
+    SDMMC_INIT_STEP(always, sdmmc_send_cmd_go_idle_state);
+
+    /* SEND_IF_COND (CMD8) command is used to identify SDHC/SDXC cards. */
+    SDMMC_INIT_STEP(always, sdmmc_init_sd_if_cond);
+
+    /* IO_SEND_OP_COND(CMD5), Determine if the card is an IO card. */
+    SDMMC_INIT_STEP(io_supported, sdmmc_init_io);
+
+    const bool is_mem = card->is_mem;
+    const bool is_sdio = !is_mem;
+
+    /* Enable CRC16 checks for data transfers in SPI mode */
+    SDMMC_INIT_STEP(is_spi, sdmmc_init_spi_crc);
+
+    /* Use SEND_OP_COND to set up card OCR */
+    SDMMC_INIT_STEP(is_mem, sdmmc_init_ocr);
+
+    const bool is_mmc = is_mem && card->is_mmc;
+    const bool is_sdmem = is_mem && !is_mmc;
+
+    ESP_LOGD(TAG, "%s: card type is %s", __func__,
+            is_sdio ? "SDIO" : is_mmc ? "MMC" : "SD");
+
+    /* Read the contents of CID register*/
+    SDMMC_INIT_STEP(is_mem, sdmmc_init_cid);
+
+    /* Assign RCA */
+    SDMMC_INIT_STEP(!is_spi, sdmmc_init_rca);
+
+    /* Read and decode the contents of CSD register */
+    SDMMC_INIT_STEP(is_mem, sdmmc_init_csd);
+
+    /* Decode the contents of mmc CID register */
+    SDMMC_INIT_STEP(is_mmc && !is_spi, sdmmc_init_mmc_decode_cid);
+
+    /* Switch the card from stand-by mode to data transfer mode (not needed if
+     * SPI interface is used). This is needed to issue SET_BLOCKLEN and
+     * SEND_SCR commands.
+     */
+    SDMMC_INIT_STEP(!is_spi, sdmmc_init_select_card);
+
+    /* SD memory cards:
+     * Set block len for SDSC cards to 512 bytes (same as SDHC)
+     * Read SCR
+     * Wait to enter data transfer state
+     */
+    SDMMC_INIT_STEP(is_sdmem, sdmmc_init_sd_blocklen);
+    SDMMC_INIT_STEP(is_sdmem, sdmmc_init_sd_scr);
+    SDMMC_INIT_STEP(is_sdmem, sdmmc_init_sd_wait_data_ready);
+
+    /* MMC cards: read CXD */
+    SDMMC_INIT_STEP(is_mmc, sdmmc_init_mmc_read_ext_csd);
+
+    /* Try to switch card to HS mode if the card supports it.
+     * Set card->max_freq_khz value accordingly.
+     */
+    SDMMC_INIT_STEP(always, sdmmc_init_card_hs_mode);
+
+    /* Set bus width. One call for every kind of card, then one for the host */
+    if (!is_spi) {
+        SDMMC_INIT_STEP(is_sdmem, sdmmc_init_sd_bus_width);
+        SDMMC_INIT_STEP(is_sdio, sdmmc_init_io_bus_width);
+        SDMMC_INIT_STEP(is_mmc, sdmmc_init_mmc_bus_width);
+        SDMMC_INIT_STEP(always, sdmmc_init_host_bus_width);
+    }
+
+    /* SD card: read SD Status register */
+    SDMMC_INIT_STEP(is_sdmem, sdmmc_init_sd_ssr);
+
+    /* Switch to the host to use card->max_freq_khz frequency. */
+    SDMMC_INIT_STEP(always, sdmmc_init_host_frequency);
+
+    /* Sanity check after switching the bus mode and frequency */
+    SDMMC_INIT_STEP(is_sdmem, sdmmc_check_scr);
+    /* Sanity check after eMMC switch to HS mode */
+    SDMMC_INIT_STEP(is_mmc, sdmmc_init_mmc_check_ext_csd);
+    /* TODO: add similar checks for SDIO */
+
+    return ESP_OK;
+}

code/components/esp-sdmmc/sdmmc_io_mh.c

@@ -0,0 +1,636 @@
+/*
+ * Copyright (c) 2006 Uwe Stuehler <uwe@openbsd.org>
+ * Adaptations to ESP-IDF Copyright (c) 2016-2018 Espressif Systems (Shanghai) PTE LTD
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include "sdmmc_common_mh.h"
+#include "esp_attr.h"
+#include "esp_compiler.h"
+
+
+#define CIS_TUPLE(NAME)  (cis_tuple_t) {.code=CISTPL_CODE_##NAME, .name=#NAME, .func=&cis_tuple_func_default, }
+#define CIS_TUPLE_WITH_FUNC(NAME, FUNC)  (cis_tuple_t) {.code=CISTPL_CODE_##NAME, .name=#NAME, .func=&(FUNC), }
+
+#define CIS_CHECK_SIZE(SIZE, MINIMAL) do {int store_size = (SIZE); if((store_size) < (MINIMAL)) return ESP_ERR_INVALID_SIZE;} while(0)
+#define CIS_CHECK_UNSUPPORTED(COND) do {if(!(COND)) return ESP_ERR_NOT_SUPPORTED;} while(0)
+#define CIS_GET_MINIMAL_SIZE    32
+
+typedef esp_err_t (*cis_tuple_info_func_t)(const void* tuple_info, uint8_t* data, FILE* fp);
+
+typedef struct {
+    int code;
+    const char *name;
+    cis_tuple_info_func_t   func;
+} cis_tuple_t;
+
+static const char* TAG = "sdmmc_io";
+
+static esp_err_t cis_tuple_func_default(const void* p, uint8_t* data, FILE* fp);
+static esp_err_t cis_tuple_func_manfid(const void* p, uint8_t* data, FILE* fp);
+static esp_err_t cis_tuple_func_cftable_entry(const void* p, uint8_t* data, FILE* fp);
+static esp_err_t cis_tuple_func_end(const void* p, uint8_t* data, FILE* fp);
+
+static const cis_tuple_t cis_table[] = {
+    CIS_TUPLE(NULL),
+    CIS_TUPLE(DEVICE),
+    CIS_TUPLE(CHKSUM),
+    CIS_TUPLE(VERS1),
+    CIS_TUPLE(ALTSTR),
+    CIS_TUPLE(CONFIG),
+    CIS_TUPLE_WITH_FUNC(CFTABLE_ENTRY, cis_tuple_func_cftable_entry),
+    CIS_TUPLE_WITH_FUNC(MANFID, cis_tuple_func_manfid),
+    CIS_TUPLE(FUNCID),
+    CIS_TUPLE(FUNCE),
+    CIS_TUPLE(VENDER_BEGIN),
+    CIS_TUPLE(VENDER_END),
+    CIS_TUPLE(SDIO_STD),
+    CIS_TUPLE(SDIO_EXT),
+    CIS_TUPLE_WITH_FUNC(END, cis_tuple_func_end),
+};
+
+
+esp_err_t sdmmc_io_reset(sdmmc_card_t* card)
+{
+    uint8_t sdio_reset = CCCR_CTL_RES;
+    esp_err_t err = sdmmc_io_rw_direct(card, 0, SD_IO_CCCR_CTL, SD_ARG_CMD52_WRITE, &sdio_reset);
+    if (err == ESP_ERR_TIMEOUT || (host_is_spi(card) && err == ESP_ERR_NOT_SUPPORTED)) {
+        /* Non-IO cards are allowed to time out (in SD mode) or
+         * return "invalid command" error (in SPI mode).
+         */
+    } else if (err == ESP_ERR_NOT_FOUND) {
+        ESP_LOGD(TAG, "%s: card not present", __func__);
+        return err;
+    } else if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: unexpected return: 0x%x", __func__, err );
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_io(sdmmc_card_t* card)
+{
+    /* IO_SEND_OP_COND(CMD5), Determine if the card is an IO card.
+     * Non-IO cards will not respond to this command.
+     */
+    esp_err_t err = sdmmc_io_send_op_cond(card, 0, &card->ocr);
+    if (err != ESP_OK) {
+        ESP_LOGD(TAG, "%s: io_send_op_cond (1) returned 0x%x; not IO card", __func__, err);
+        card->is_sdio = 0;
+        card->is_mem = 1;
+    } else {
+        card->is_sdio = 1;
+
+        if (card->ocr & SD_IO_OCR_MEM_PRESENT) {
+            ESP_LOGD(TAG, "%s: IO-only card", __func__);
+            card->is_mem = 0;
+        }
+        card->num_io_functions = SD_IO_OCR_NUM_FUNCTIONS(card->ocr);
+        ESP_LOGD(TAG, "%s: number of IO functions: %d", __func__, card->num_io_functions);
+        if (card->num_io_functions == 0) {
+            card->is_sdio = 0;
+        }
+        uint32_t host_ocr = get_host_ocr(card->host.io_voltage);
+        host_ocr &= card->ocr;
+        err = sdmmc_io_send_op_cond(card, host_ocr, &card->ocr);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: sdmmc_io_send_op_cond (1) returned 0x%x", __func__, err);
+            return err;
+        }
+        err = sdmmc_io_enable_int(card);
+        if (err != ESP_OK) {
+            ESP_LOGD(TAG, "%s: sdmmc_enable_int failed (0x%x)", __func__, err);
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_io_bus_width(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    card->log_bus_width = 0;
+    if (card->host.flags & SDMMC_HOST_FLAG_4BIT) {
+        uint8_t card_cap = 0;
+        err = sdmmc_io_rw_direct(card, 0, SD_IO_CCCR_CARD_CAP,
+                SD_ARG_CMD52_READ, &card_cap);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: sdmmc_io_rw_direct (read SD_IO_CCCR_CARD_CAP) returned 0x%0x", __func__, err);
+            return err;
+        }
+        ESP_LOGD(TAG, "IO card capabilities byte: %02x", card_cap);
+        if (!(card_cap & CCCR_CARD_CAP_LSC) ||
+                (card_cap & CCCR_CARD_CAP_4BLS)) {
+            // This card supports 4-bit bus mode
+            uint8_t bus_width = CCCR_BUS_WIDTH_4;
+            err = sdmmc_io_rw_direct(card, 0, SD_IO_CCCR_BUS_WIDTH,
+                                SD_ARG_CMD52_WRITE, &bus_width);
+            if (err != ESP_OK) {
+                ESP_LOGE(TAG, "%s: sdmmc_io_rw_direct (write SD_IO_CCCR_BUS_WIDTH) returned 0x%0x", __func__, err);
+                return err;
+            }
+            card->log_bus_width = 2;
+        }
+    }
+    return ESP_OK;
+}
+
+
+esp_err_t sdmmc_io_enable_hs_mode(sdmmc_card_t* card)
+{
+    /* If the host is configured to use low frequency, don't attempt to switch */
+    if (card->host.max_freq_khz < SDMMC_FREQ_DEFAULT) {
+        card->max_freq_khz = card->host.max_freq_khz;
+        return ESP_OK;
+    } else if (card->host.max_freq_khz < SDMMC_FREQ_HIGHSPEED) {
+        card->max_freq_khz = SDMMC_FREQ_DEFAULT;
+        return ESP_OK;
+    }
+
+    /* For IO cards, do write + read operation on "High Speed" register,
+     * setting EHS bit. If both EHS and SHS read back as set, then HS mode
+     * has been enabled.
+     */
+    uint8_t val = CCCR_HIGHSPEED_ENABLE;
+    esp_err_t err = sdmmc_io_rw_direct(card, 0, SD_IO_CCCR_HIGHSPEED,
+            SD_ARG_CMD52_WRITE | SD_ARG_CMD52_EXCHANGE, &val);
+    if (err != ESP_OK) {
+        ESP_LOGD(TAG, "%s: sdmmc_io_rw_direct returned 0x%x", __func__, err);
+        return err;
+    }
+
+    ESP_LOGD(TAG, "%s: CCCR_HIGHSPEED=0x%02x", __func__, val);
+    const uint8_t hs_mask = CCCR_HIGHSPEED_ENABLE | CCCR_HIGHSPEED_SUPPORT;
+    if ((val & hs_mask) != hs_mask) {
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    card->max_freq_khz = SDMMC_FREQ_HIGHSPEED;
+    return ESP_OK;
+}
+
+
+esp_err_t sdmmc_io_send_op_cond(sdmmc_card_t* card, uint32_t ocr, uint32_t *ocrp)
+{
+    esp_err_t err = ESP_OK;
+    sdmmc_command_t cmd = {
+        .flags = SCF_CMD_BCR | SCF_RSP_R4,
+        .arg = ocr,
+        .opcode = SD_IO_SEND_OP_COND
+    };
+    for (size_t i = 0; i < 100; i++) {
+        err = sdmmc_send_cmd(card, &cmd);
+        if (err != ESP_OK) {
+            break;
+        }
+        if ((MMC_R4(cmd.response) & SD_IO_OCR_MEM_READY) ||
+            ocr == 0) {
+            break;
+        }
+        err = ESP_ERR_TIMEOUT;
+        vTaskDelay(SDMMC_IO_SEND_OP_COND_DELAY_MS / portTICK_PERIOD_MS);
+    }
+    if (err == ESP_OK && ocrp != NULL)
+        *ocrp = MMC_R4(cmd.response);
+
+    return err;
+}
+
+esp_err_t sdmmc_io_rw_direct(sdmmc_card_t* card, int func,
+        uint32_t reg, uint32_t arg, uint8_t *byte)
+{
+    esp_err_t err;
+    sdmmc_command_t cmd = {
+        .flags = SCF_CMD_AC | SCF_RSP_R5,
+        .arg = 0,
+        .opcode = SD_IO_RW_DIRECT
+    };
+
+    arg |= (func & SD_ARG_CMD52_FUNC_MASK) << SD_ARG_CMD52_FUNC_SHIFT;
+    arg |= (reg & SD_ARG_CMD52_REG_MASK) << SD_ARG_CMD52_REG_SHIFT;
+    arg |= (*byte & SD_ARG_CMD52_DATA_MASK) << SD_ARG_CMD52_DATA_SHIFT;
+    cmd.arg = arg;
+
+    err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGV(TAG, "%s: sdmmc_send_cmd returned 0x%x", __func__, err);
+        return err;
+    }
+
+    *byte = SD_R5_DATA(cmd.response);
+
+    return ESP_OK;
+}
+
+
+esp_err_t sdmmc_io_read_byte(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, uint8_t *out_byte)
+{
+    esp_err_t ret = sdmmc_io_rw_direct(card, function, addr, SD_ARG_CMD52_READ, out_byte);
+    if (unlikely(ret != ESP_OK)) {
+        ESP_LOGE(TAG, "%s: sdmmc_io_rw_direct (read 0x%x) returned 0x%x", __func__, addr, ret);
+    }
+    return ret;
+}
+
+esp_err_t sdmmc_io_write_byte(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, uint8_t in_byte, uint8_t* out_byte)
+{
+    uint8_t tmp_byte = in_byte;
+    esp_err_t ret = sdmmc_io_rw_direct(card, function, addr,
+            SD_ARG_CMD52_WRITE | SD_ARG_CMD52_EXCHANGE, &tmp_byte);
+    if (unlikely(ret != ESP_OK)) {
+        ESP_LOGE(TAG, "%s: sdmmc_io_rw_direct (write 0x%x) returned 0x%x", __func__, addr, ret);
+        return ret;
+    }
+    if (out_byte != NULL) {
+        *out_byte = tmp_byte;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_io_rw_extended(sdmmc_card_t* card, int func,
+    uint32_t reg, int arg, void *datap, size_t datalen)
+{
+    esp_err_t err;
+    const size_t max_byte_transfer_size = 512;
+    sdmmc_command_t cmd = {
+        .flags = SCF_CMD_AC | SCF_RSP_R5,
+        .arg = 0,
+        .opcode = SD_IO_RW_EXTENDED,
+        .data = datap,
+        .datalen = datalen,
+        .blklen = max_byte_transfer_size /* TODO: read max block size from CIS */
+    };
+
+    uint32_t count; /* number of bytes or blocks, depending on transfer mode */
+    if (arg & SD_ARG_CMD53_BLOCK_MODE) {
+        if (cmd.datalen % cmd.blklen != 0) {
+            return ESP_ERR_INVALID_SIZE;
+        }
+        count = cmd.datalen / cmd.blklen;
+    } else {
+        if (datalen > max_byte_transfer_size) {
+            /* TODO: split into multiple operations? */
+            return ESP_ERR_INVALID_SIZE;
+        }
+        if (datalen == max_byte_transfer_size) {
+            count = 0;  // See 5.3.1 SDIO simplifed spec
+        } else {
+            count = datalen;
+        }
+        cmd.blklen = datalen;
+    }
+
+    arg |= (func & SD_ARG_CMD53_FUNC_MASK) << SD_ARG_CMD53_FUNC_SHIFT;
+    arg |= (reg & SD_ARG_CMD53_REG_MASK) << SD_ARG_CMD53_REG_SHIFT;
+    arg |= (count & SD_ARG_CMD53_LENGTH_MASK) << SD_ARG_CMD53_LENGTH_SHIFT;
+    cmd.arg = arg;
+
+    if ((arg & SD_ARG_CMD53_WRITE) == 0) {
+        cmd.flags |= SCF_CMD_READ;
+    }
+
+    err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd returned 0x%x", __func__, err);
+        return err;
+    }
+
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_io_read_bytes(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, void* dst, size_t size)
+{
+    /* host quirk: SDIO transfer with length not divisible by 4 bytes
+     * has to be split into two transfers: one with aligned length,
+     * the other one for the remaining 1-3 bytes.
+     */
+    uint8_t *pc_dst = dst;
+    while (size > 0) {
+        size_t size_aligned = size & (~3);
+        size_t will_transfer = size_aligned > 0 ? size_aligned : size;
+
+        esp_err_t err = sdmmc_io_rw_extended(card, function, addr,
+                SD_ARG_CMD53_READ | SD_ARG_CMD53_INCREMENT,
+                pc_dst, will_transfer);
+        if (unlikely(err != ESP_OK)) {
+            return err;
+        }
+        pc_dst += will_transfer;
+        size -= will_transfer;
+        addr += will_transfer;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_io_write_bytes(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, const void* src, size_t size)
+{
+    /* same host quirk as in sdmmc_io_read_bytes */
+    const uint8_t *pc_src = (const uint8_t*) src;
+
+    while (size > 0) {
+        size_t size_aligned = size & (~3);
+        size_t will_transfer = size_aligned > 0 ? size_aligned : size;
+
+        esp_err_t err = sdmmc_io_rw_extended(card, function, addr,
+                SD_ARG_CMD53_WRITE | SD_ARG_CMD53_INCREMENT,
+                (void*) pc_src, will_transfer);
+        if (unlikely(err != ESP_OK)) {
+            return err;
+        }
+        pc_src += will_transfer;
+        size -= will_transfer;
+        addr += will_transfer;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_io_read_blocks(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, void* dst, size_t size)
+{
+    if (unlikely(size % 4 != 0)) {
+        return ESP_ERR_INVALID_SIZE;
+    }
+    return sdmmc_io_rw_extended(card, function, addr,
+            SD_ARG_CMD53_READ | SD_ARG_CMD53_INCREMENT | SD_ARG_CMD53_BLOCK_MODE,
+            dst, size);
+}
+
+esp_err_t sdmmc_io_write_blocks(sdmmc_card_t* card, uint32_t function,
+        uint32_t addr, const void* src, size_t size)
+{
+    if (unlikely(size % 4 != 0)) {
+        return ESP_ERR_INVALID_SIZE;
+    }
+    return sdmmc_io_rw_extended(card, function, addr,
+            SD_ARG_CMD53_WRITE | SD_ARG_CMD53_INCREMENT | SD_ARG_CMD53_BLOCK_MODE,
+            (void*) src, size);
+}
+
+esp_err_t sdmmc_io_enable_int(sdmmc_card_t* card)
+{
+    if (card->host.io_int_enable == NULL) {
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    return (*card->host.io_int_enable)(card->host.slot);
+}
+
+esp_err_t sdmmc_io_wait_int(sdmmc_card_t* card, TickType_t timeout_ticks)
+{
+    if (card->host.io_int_wait == NULL) {
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    return (*card->host.io_int_wait)(card->host.slot, timeout_ticks);
+}
+
+
+/*
+ * Print the CIS information of a CIS card, currently only ESP slave supported.
+ */
+
+static esp_err_t cis_tuple_func_default(const void* p, uint8_t* data, FILE* fp)
+{
+    const cis_tuple_t* tuple = (const cis_tuple_t*)p;
+    uint8_t code = *(data++);
+    int size = *(data++);
+    if (tuple) {
+        fprintf(fp, "TUPLE: %s, size: %d: ", tuple->name, size);
+    } else {
+        fprintf(fp, "TUPLE: unknown(%02X), size: %d: ", code, size);
+    }
+    for (int i = 0; i < size; i++) fprintf(fp, "%02X ", *(data++));
+    fprintf(fp, "\n");
+    return ESP_OK;
+}
+
+static esp_err_t cis_tuple_func_manfid(const void* p, uint8_t* data, FILE* fp)
+{
+    const cis_tuple_t* tuple = (const cis_tuple_t*)p;
+    data++;
+    int size = *(data++);
+    fprintf(fp, "TUPLE: %s, size: %d\n", tuple->name, size);
+    CIS_CHECK_SIZE(size, 4);
+    fprintf(fp, "  MANF: %04X, CARD: %04X\n", *(uint16_t*)(data), *(uint16_t*)(data+2));
+    return ESP_OK;
+}
+
+static esp_err_t cis_tuple_func_end(const void* p, uint8_t* data, FILE* fp)
+{
+    const cis_tuple_t* tuple = (const cis_tuple_t*)p;
+    data++;
+    fprintf(fp, "TUPLE: %s\n", tuple->name);
+    return ESP_OK;
+}
+
+static esp_err_t cis_tuple_func_cftable_entry(const void* p, uint8_t* data, FILE* fp)
+{
+    const cis_tuple_t* tuple = (const cis_tuple_t*)p;
+    data++;
+    int size = *(data++);
+    fprintf(fp, "TUPLE: %s, size: %d\n", tuple->name, size);
+    CIS_CHECK_SIZE(size, 2);
+
+    CIS_CHECK_SIZE(size--, 1);
+    bool interface = data[0] & BIT(7);
+    bool def = data[0] & BIT(6);
+    int conf_ent_num = data[0] & 0x3F;
+    fprintf(fp, "  INDX: %02X, Intface: %d, Default: %d, Conf-Entry-Num: %d\n", *(data++), interface, def, conf_ent_num);
+
+    if (interface) {
+        CIS_CHECK_SIZE(size--, 1);
+        fprintf(fp, "  IF: %02X\n", *(data++));
+    }
+
+    CIS_CHECK_SIZE(size--, 1);
+    bool misc = data[0] & BIT(7);
+    int mem_space = (data[0] >> 5 )&(0x3);
+    bool irq = data[0] & BIT(4);
+    bool io_sp = data[0] & BIT(3);
+    bool timing = data[0] & BIT(2);
+    int power = data[0] & 3;
+    fprintf(fp, "  FS: %02X, misc: %d, mem_space: %d, irq: %d, io_space: %d, timing: %d, power: %d\n", *(data++), misc, mem_space, irq, io_sp, timing, power);
+
+    CIS_CHECK_UNSUPPORTED(power == 0);  //power descriptor is not handled yet
+    CIS_CHECK_UNSUPPORTED(!timing);     //timing descriptor is not handled yet
+    CIS_CHECK_UNSUPPORTED(!io_sp);      //io space descriptor is not handled yet
+
+    if (irq) {
+        CIS_CHECK_SIZE(size--, 1);
+        bool mask = data[0] & BIT(4);
+        fprintf(fp, "  IR: %02X, mask: %d, ",*(data++), mask);
+        if (mask) {
+            CIS_CHECK_SIZE(size, 2);
+            size-=2;
+            fprintf(fp, "  IRQ: %02X %02X\n", data[0], data[1]);
+            data+=2;
+        }
+    }
+
+    if (mem_space) {
+        CIS_CHECK_SIZE(size, 2);
+        size-=2;
+        CIS_CHECK_UNSUPPORTED(mem_space==1); //other cases not handled yet
+        int len = *(uint16_t*)data;
+        fprintf(fp, "  LEN: %04X\n", len);
+        data+=2;
+    }
+
+    CIS_CHECK_UNSUPPORTED(misc==0);    //misc descriptor is not handled yet
+    return ESP_OK;
+}
+
+static const cis_tuple_t* get_tuple(uint8_t code)
+{
+    for (int i = 0; i < sizeof(cis_table)/sizeof(cis_tuple_t); i++) {
+        if (code == cis_table[i].code) return &cis_table[i];
+    }
+    return NULL;
+}
+
+esp_err_t sdmmc_io_print_cis_info(uint8_t* buffer, size_t buffer_size, FILE* fp)
+{
+    ESP_LOG_BUFFER_HEXDUMP("CIS", buffer, buffer_size, ESP_LOG_DEBUG);
+    if (!fp) fp = stdout;
+
+    uint8_t* cis = buffer;
+    do {
+        const cis_tuple_t* tuple = get_tuple(cis[0]);
+        int size = cis[1];
+        esp_err_t ret = ESP_OK;
+        if (tuple) {
+            ret = tuple->func(tuple, cis, fp);
+        } else {
+            ret = cis_tuple_func_default(NULL, cis, fp);
+        }
+        if (ret != ESP_OK) return ret;
+        cis += 2 + size;
+        if (tuple && tuple->code == CISTPL_CODE_END) break;
+    } while (cis < buffer + buffer_size) ;
+    return ESP_OK;
+}
+
+/**
+ * Check tuples in the buffer.
+ *
+ * @param buf Buffer to check
+ * @param buffer_size Size of the buffer
+ * @param inout_cis_offset
+ *          - input: the last cis_offset, relative to the beginning of the buf. -1 if
+ *                      this buffer begin with the tuple length, otherwise should be no smaller than
+ *                      zero.
+ *          - output: when the end tuple found, output offset of the CISTPL_CODE_END
+ *                      byte + 1 (relative to the beginning of the buffer; when not found, output
+ *                      the address of next tuple code.
+ *
+ * @return true if found, false if haven't.
+ */
+static bool check_tuples_in_buffer(uint8_t* buf, int buffer_size, int* inout_cis_offset)
+{
+    int cis_offset = *inout_cis_offset;
+    if (cis_offset == -1) {
+        //the CIS code is checked in the last buffer, skip to next tuple
+        cis_offset += buf[0] + 2;
+    }
+    assert(cis_offset >= 0);
+    while (1) {
+        if (cis_offset < buffer_size) {
+            //A CIS code in the buffer, check it
+            if (buf[cis_offset] == CISTPL_CODE_END) {
+                *inout_cis_offset = cis_offset + 1;
+                return true;
+            }
+        }
+        if (cis_offset + 1 < buffer_size) {
+            cis_offset += buf[cis_offset+1] + 2;
+        } else {
+            break;
+        }
+    }
+    *inout_cis_offset = cis_offset;
+    return false;
+}
+
+esp_err_t sdmmc_io_get_cis_data(sdmmc_card_t* card, uint8_t* out_buffer, size_t buffer_size, size_t* inout_cis_size)
+{
+    esp_err_t ret = ESP_OK;
+    WORD_ALIGNED_ATTR uint8_t buf[CIS_GET_MINIMAL_SIZE];
+
+    /* Pointer to size is a mandatory parameter */
+    assert(inout_cis_size);
+
+    /*
+     * CIS region exist in 0x1000~0x17FFF of FUNC 0, get the start address of it
+     * from CCCR register.
+     */
+    uint32_t addr;
+    ret = sdmmc_io_read_bytes(card, 0, 9, &addr, 3);
+    if (ret != ESP_OK) return ret;
+    //the sdmmc_io driver reads 4 bytes, the most significant byte is not the address.
+    addr &= 0xffffff;
+    if (addr < 0x1000 || addr > 0x17FFF) {
+        return ESP_ERR_INVALID_RESPONSE;
+    }
+
+    /*
+     * To avoid reading too long, take the input value as limitation if
+     * existing.
+     */
+    size_t max_reading = UINT32_MAX;
+    if (*inout_cis_size != 0) {
+        max_reading = *inout_cis_size;
+    }
+
+    /*
+     * Parse the length while reading. If find the end tuple, or reaches the
+     * limitation, read no more and return both the data and the size already
+     * read.
+     */
+    int buffer_offset = 0;
+    int cur_cis_offset = 0;
+    bool end_tuple_found = false;
+    do {
+        ret = sdmmc_io_read_bytes(card, 0, addr + buffer_offset, &buf, CIS_GET_MINIMAL_SIZE);
+        if (ret != ESP_OK) return ret;
+
+        //calculate relative to the beginning of the buffer
+        int offset = cur_cis_offset - buffer_offset;
+        bool finish = check_tuples_in_buffer(buf, CIS_GET_MINIMAL_SIZE, &offset);
+
+        int remain_size = buffer_size - buffer_offset;
+        int copy_len;
+        if (finish) {
+            copy_len = MIN(offset, remain_size);
+            end_tuple_found = true;
+        } else {
+            copy_len = MIN(CIS_GET_MINIMAL_SIZE, remain_size);
+        }
+        if (copy_len > 0) {
+            memcpy(out_buffer + buffer_offset, buf, copy_len);
+        }
+        cur_cis_offset = buffer_offset + offset;
+        buffer_offset += CIS_GET_MINIMAL_SIZE;
+    } while (!end_tuple_found && buffer_offset < max_reading);
+
+    if (end_tuple_found) {
+        *inout_cis_size = cur_cis_offset;
+        if (cur_cis_offset > buffer_size) {
+            return ESP_ERR_INVALID_SIZE;
+        } else {
+            return ESP_OK;
+        }
+    } else {
+        return ESP_ERR_NOT_FOUND;
+    }
+}

code/components/esp-sdmmc/sdmmc_mmc_mh.c

@@ -0,0 +1,300 @@
+/*
+ * Copyright (c) 2006 Uwe Stuehler <uwe@openbsd.org>
+ * Adaptations to ESP-IDF Copyright (c) 2016-2018 Espressif Systems (Shanghai) PTE LTD
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include <unistd.h>
+#include "sdmmc_common_mh.h"
+
+static const char* TAG = "sdmmc_mmc";
+
+
+esp_err_t sdmmc_init_mmc_read_ext_csd(sdmmc_card_t* card)
+{
+    int card_type;
+    esp_err_t err = ESP_OK;
+
+    uint8_t* ext_csd = heap_caps_malloc(EXT_CSD_MMC_SIZE, MALLOC_CAP_DMA);
+    if (!ext_csd) {
+        ESP_LOGE(TAG, "%s: could not allocate ext_csd", __func__);
+        return ESP_ERR_NO_MEM;
+    }
+
+    uint32_t sectors = 0;
+
+    ESP_LOGD(TAG, "MMC version: %d", card->csd.mmc_ver);
+    if (card->csd.mmc_ver < MMC_CSD_MMCVER_4_0) {
+        err = ESP_ERR_NOT_SUPPORTED;
+        goto out;
+    }
+
+    /* read EXT_CSD */
+    err = sdmmc_mmc_send_ext_csd_data(card, ext_csd, EXT_CSD_MMC_SIZE);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_ext_csd_data error 0x%x", __func__, err);
+        goto out;
+    }
+    card_type = ext_csd[EXT_CSD_CARD_TYPE];
+    card->is_ddr = 0;
+    if (card_type & EXT_CSD_CARD_TYPE_F_52M_1_8V) {
+        card->max_freq_khz = SDMMC_FREQ_52M;
+        if ((card->host.flags & SDMMC_HOST_FLAG_DDR) &&
+                card->host.max_freq_khz >= SDMMC_FREQ_26M &&
+                card->host.get_bus_width(card->host.slot) == 4) {
+            ESP_LOGD(TAG, "card and host support DDR mode");
+            card->is_ddr = 1;
+        }
+    } else if (card_type & EXT_CSD_CARD_TYPE_F_52M) {
+        card->max_freq_khz = SDMMC_FREQ_52M;
+    } else if (card_type & EXT_CSD_CARD_TYPE_F_26M) {
+        card->max_freq_khz = SDMMC_FREQ_26M;
+    } else {
+        ESP_LOGW(TAG, "%s: unknown CARD_TYPE 0x%x", __func__, card_type);
+    }
+    /* For MMC cards, use speed value from EXT_CSD */
+    card->csd.tr_speed = card->max_freq_khz * 1000;
+    ESP_LOGD(TAG, "MMC card type %d, max_freq_khz=%d, is_ddr=%d", card_type, card->max_freq_khz, card->is_ddr);
+    card->max_freq_khz = MIN(card->max_freq_khz, card->host.max_freq_khz);
+
+    if (card->host.flags & SDMMC_HOST_FLAG_8BIT) {
+        card->ext_csd.power_class = ext_csd[(card->max_freq_khz > SDMMC_FREQ_26M) ?
+                EXT_CSD_PWR_CL_52_360 : EXT_CSD_PWR_CL_26_360] >> 4;
+        card->log_bus_width = 3;
+    } else if (card->host.flags & SDMMC_HOST_FLAG_4BIT) {
+        card->ext_csd.power_class = ext_csd[(card->max_freq_khz > SDMMC_FREQ_26M) ?
+                EXT_CSD_PWR_CL_52_360 : EXT_CSD_PWR_CL_26_360] & 0x0f;
+        card->log_bus_width = 2;
+    } else {
+        card->ext_csd.power_class = 0; //card must be able to do full rate at powerclass 0 in 1-bit mode
+        card->log_bus_width = 0;
+    }
+
+    sectors = ( ext_csd[EXT_CSD_SEC_COUNT + 0] << 0 )
+        | ( ext_csd[EXT_CSD_SEC_COUNT + 1] << 8 )
+        | ( ext_csd[EXT_CSD_SEC_COUNT + 2] << 16 )
+        | ( ext_csd[EXT_CSD_SEC_COUNT + 3] << 24 );
+
+    if (sectors > (2u * 1024 * 1024 * 1024) / 512) {
+        card->csd.capacity = sectors;
+    }
+
+    /* erased state of a bit, if 1 byte value read is 0xFF else 0x00 */
+    card->ext_csd.erase_mem_state = ext_csd[EXT_CSD_ERASED_MEM_CONT];
+    card->ext_csd.rev = ext_csd[EXT_CSD_REV];
+    card->ext_csd.sec_feature = ext_csd[EXT_CSD_SEC_FEATURE_SUPPORT];
+
+out:
+    free(ext_csd);
+    return err;
+}
+
+esp_err_t sdmmc_init_mmc_bus_width(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    if (card->ext_csd.power_class != 0) {
+        err = sdmmc_mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
+                EXT_CSD_POWER_CLASS, card->ext_csd.power_class);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: can't change power class (%d bit), 0x%x"
+                    , __func__, card->ext_csd.power_class, err);
+            return err;
+        }
+    }
+
+    if (card->log_bus_width > 0) {
+        int csd_bus_width_value = EXT_CSD_BUS_WIDTH_1;
+        int bus_width = 1;
+        if (card->log_bus_width == 2) {
+            if (card->is_ddr) {
+                csd_bus_width_value = EXT_CSD_BUS_WIDTH_4_DDR;
+            } else {
+                csd_bus_width_value = EXT_CSD_BUS_WIDTH_4;
+            }
+            bus_width = 4;
+        } else if (card->log_bus_width == 3) {
+            if (card->is_ddr) {
+                csd_bus_width_value = EXT_CSD_BUS_WIDTH_8_DDR;
+            } else {
+                csd_bus_width_value = EXT_CSD_BUS_WIDTH_8;
+            }
+            bus_width = 8;
+        }
+        err = sdmmc_mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
+                EXT_CSD_BUS_WIDTH, csd_bus_width_value);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: can't change bus width (%d bit), 0x%x",
+                    __func__, bus_width, err);
+            return err;
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_mmc_enable_hs_mode(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    if (card->max_freq_khz > SDMMC_FREQ_26M) {
+        /* switch to high speed timing */
+        err = sdmmc_mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
+                EXT_CSD_HS_TIMING, EXT_CSD_HS_TIMING_HS);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: mmc_switch EXT_CSD_HS_TIMING_HS error 0x%x",
+                    __func__, err);
+            return err;
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_mmc_decode_cid(int mmc_ver, sdmmc_response_t resp, sdmmc_cid_t* out_cid)
+{
+    if (mmc_ver == MMC_CSD_MMCVER_1_0 ||
+            mmc_ver == MMC_CSD_MMCVER_1_4) {
+        out_cid->mfg_id = MMC_CID_MID_V1(resp);
+        out_cid->oem_id = 0;
+        MMC_CID_PNM_V1_CPY(resp, out_cid->name);
+        out_cid->revision = MMC_CID_REV_V1(resp);
+        out_cid->serial = MMC_CID_PSN_V1(resp);
+        out_cid->date = MMC_CID_MDT_V1(resp);
+    } else if (mmc_ver == MMC_CSD_MMCVER_2_0 ||
+            mmc_ver == MMC_CSD_MMCVER_3_1 ||
+            mmc_ver == MMC_CSD_MMCVER_4_0) {
+        out_cid->mfg_id = MMC_CID_MID_V2(resp);
+        out_cid->oem_id = MMC_CID_OID_V2(resp);
+        MMC_CID_PNM_V1_CPY(resp, out_cid->name);
+        out_cid->revision = 0;
+        out_cid->serial = MMC_CID_PSN_V1(resp);
+        out_cid->date = 0;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_mmc_decode_csd(sdmmc_response_t response, sdmmc_csd_t* out_csd)
+{
+    out_csd->csd_ver = MMC_CSD_CSDVER(response);
+    if (out_csd->csd_ver == MMC_CSD_CSDVER_1_0 ||
+            out_csd->csd_ver == MMC_CSD_CSDVER_2_0 ||
+            out_csd->csd_ver == MMC_CSD_CSDVER_EXT_CSD) {
+        out_csd->mmc_ver = MMC_CSD_MMCVER(response);
+        out_csd->capacity = MMC_CSD_CAPACITY(response);
+        out_csd->read_block_len = MMC_CSD_READ_BL_LEN(response);
+    } else {
+        ESP_LOGE(TAG, "unknown MMC CSD structure version 0x%x\n", out_csd->csd_ver);
+        return 1;
+    }
+    int read_bl_size = 1 << out_csd->read_block_len;
+    out_csd->sector_size = MIN(read_bl_size, 512);
+    if (out_csd->sector_size < read_bl_size) {
+        out_csd->capacity *= read_bl_size / out_csd->sector_size;
+    }
+    /* tr_speed will be determined when reading CXD */
+    out_csd->tr_speed = 0;
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_mmc_send_ext_csd_data(sdmmc_card_t* card, void *out_data, size_t datalen)
+{
+    assert(esp_ptr_dma_capable(out_data));
+    sdmmc_command_t cmd = {
+        .data = out_data,
+        .datalen = datalen,
+        .blklen = datalen,
+        .opcode = MMC_SEND_EXT_CSD,
+        .arg = 0,
+        .flags = SCF_CMD_ADTC | SCF_RSP_R1 | SCF_CMD_READ
+    };
+    return sdmmc_send_cmd(card, &cmd);
+}
+
+esp_err_t sdmmc_mmc_switch(sdmmc_card_t* card, uint8_t set, uint8_t index, uint8_t value)
+{
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SWITCH,
+            .arg = (MMC_SWITCH_MODE_WRITE_BYTE << 24) | (index << 16) | (value << 8) | set,
+            .flags = SCF_RSP_R1B | SCF_CMD_AC | SCF_WAIT_BUSY,
+    };
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err == ESP_OK) {
+        //check response bit to see that switch was accepted
+        if (MMC_R1(cmd.response) & MMC_R1_SWITCH_ERROR) {
+            err = ESP_ERR_INVALID_RESPONSE;
+        }
+    }
+
+    return err;
+}
+
+esp_err_t sdmmc_init_mmc_check_ext_csd(sdmmc_card_t* card)
+{
+    assert(card->is_mem == 1 && card->rca != 0);
+
+    /*
+     * Integrity check required if card switched to HS mode
+     * card->max_freq_khz = MIN(card->max_freq_khz, card->host.max_freq_khz)
+     * For 26MHz limit background see sdmmc_mmc_enable_hs_mode()
+     */
+    if (card->max_freq_khz <= SDMMC_FREQ_26M) {
+        return ESP_OK;
+    }
+
+    /* ensure EXT_CSD buffer is available before starting any SD-card operation */
+    uint8_t* ext_csd = heap_caps_malloc(EXT_CSD_MMC_SIZE, MALLOC_CAP_DMA);
+    if (!ext_csd) {
+        ESP_LOGE(TAG, "%s: could not allocate ext_csd", __func__);
+        return ESP_ERR_NO_MEM;
+    }
+
+    /* ensure card is in transfer state before read ext_csd */
+    uint32_t status;
+    esp_err_t err = sdmmc_send_cmd_send_status(card, &status);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_status returned 0x%x", __func__, err);
+        goto out;
+    }
+    status = ((status & MMC_R1_CURRENT_STATE_MASK) >> MMC_R1_CURRENT_STATE_POS);
+    if (status != MMC_R1_CURRENT_STATE_TRAN) {
+        ESP_LOGE(TAG, "%s: card not in transfer state", __func__);
+        err = ESP_ERR_INVALID_STATE;
+        goto out;
+    }
+
+    /* read EXT_CSD to ensure device works fine in HS mode */
+    err = sdmmc_mmc_send_ext_csd_data(card, ext_csd, EXT_CSD_MMC_SIZE);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_ext_csd_data error 0x%x", __func__, err);
+        goto out;
+    }
+
+    /* EXT_CSD static fields should match the previous read values in sdmmc_card_init */
+    if ((card->ext_csd.rev != ext_csd[EXT_CSD_REV]) ||
+            (card->ext_csd.sec_feature != ext_csd[EXT_CSD_SEC_FEATURE_SUPPORT])) {
+        ESP_LOGE(TAG, "%s: Data integrity test fail in HS mode", __func__);
+        err = ESP_FAIL;
+    }
+
+out:
+    free(ext_csd);
+    return err;
+}
+
+uint32_t sdmmc_mmc_get_erase_timeout_ms(const sdmmc_card_t* card, int arg, size_t erase_size_kb)
+{
+    /* TODO: calculate erase timeout based on ext_csd (trim_timeout) */
+    uint32_t timeout_ms = SDMMC_SD_DISCARD_TIMEOUT * erase_size_kb / card->csd.sector_size;
+    timeout_ms = MAX(1000, timeout_ms);
+    ESP_LOGD(TAG, "%s: erase timeout %u s (erasing %u kB, %ums per sector)",
+             __func__, timeout_ms / 1000, erase_size_kb, SDMMC_SD_DISCARD_TIMEOUT);
+    return timeout_ms;
+}

code/components/esp-sdmmc/sdmmc_sd_mh.c

@@ -0,0 +1,449 @@
+/*
+ * Copyright (c) 2006 Uwe Stuehler <uwe@openbsd.org>
+ * Adaptations to ESP-IDF Copyright (c) 2016-2018 Espressif Systems (Shanghai) PTE LTD
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include "sdmmc_common_mh.h"
+
+static const char* TAG = "sdmmc_sd";
+
+esp_err_t sdmmc_init_sd_if_cond(sdmmc_card_t* card)
+{
+    /* SEND_IF_COND (CMD8) command is used to identify SDHC/SDXC cards.
+     * SD v1 and non-SD cards will not respond to this command.
+     */
+    uint32_t host_ocr = get_host_ocr(card->host.io_voltage);
+    esp_err_t err = sdmmc_send_cmd_send_if_cond(card, host_ocr);
+    if (err == ESP_OK) {
+        ESP_LOGD(TAG, "SDHC/SDXC card");
+        host_ocr |= SD_OCR_SDHC_CAP;
+    } else if (err == ESP_ERR_TIMEOUT) {
+        ESP_LOGD(TAG, "CMD8 timeout; not an SD v2.00 card");
+    } else if (host_is_spi(card) && err == ESP_ERR_NOT_SUPPORTED) {
+        ESP_LOGD(TAG, "CMD8 rejected; not an SD v2.00 card");
+    } else {
+        ESP_LOGE(TAG, "%s: send_if_cond (1) returned 0x%x", __func__, err);
+        return err;
+    }
+    card->ocr = host_ocr;
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_sd_blocklen(sdmmc_card_t* card)
+{
+    /* SDSC cards support configurable data block lengths.
+     * We don't use this feature and set the block length to 512 bytes,
+     * same as the block length for SDHC cards.
+     */
+    if ((card->ocr & SD_OCR_SDHC_CAP) == 0) {
+        esp_err_t err = sdmmc_send_cmd_set_blocklen(card, &card->csd);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: set_blocklen returned 0x%x", __func__, err);
+            return err;
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_sd_scr(sdmmc_card_t* card)
+{
+    esp_err_t err;
+    /* Get the contents of SCR register: bus width and the version of SD spec
+     * supported by the card.
+     * In SD mode, this is the first command which uses D0 line. Errors at
+     * this step usually indicate connection issue or lack of pull-up resistor.
+     */
+    err = sdmmc_send_cmd_send_scr(card, &card->scr);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_scr (1) returned 0x%x", __func__, err);
+        return err;
+    }
+
+    if ((card->scr.bus_width & SCR_SD_BUS_WIDTHS_4BIT)
+            && (card->host.flags & SDMMC_HOST_FLAG_4BIT)) {
+        card->log_bus_width = 2;
+    } else {
+        card->log_bus_width = 0;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_sd_ssr(sdmmc_card_t* card)
+{
+    esp_err_t err = ESP_OK;
+    /* Get the contents of SSR register: SD additional information
+     * ACMD13 to read 512byte SD status information
+     */
+    uint32_t* sd_ssr = heap_caps_calloc(1, SD_SSR_SIZE, MALLOC_CAP_DMA);
+    if (!sd_ssr) {
+        ESP_LOGE(TAG, "%s: could not allocate sd_ssr", __func__);
+        return ESP_ERR_NO_MEM;
+    }
+
+    sdmmc_command_t cmd = {
+        .data = sd_ssr,
+        .datalen = SD_SSR_SIZE,
+        .blklen = SD_SSR_SIZE,
+        .opcode = SD_APP_SD_STATUS,
+        .arg = 0,
+        .flags = SCF_CMD_ADTC | SCF_RSP_R1 | SCF_CMD_READ
+    };
+
+    // read SD status register
+    err = sdmmc_send_app_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        free(sd_ssr);
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd returned 0x%x", __func__, err);
+        return err;
+    }
+
+    err = sdmmc_decode_ssr(sd_ssr, &card->ssr);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: error sdmmc_decode_scr returned 0x%x", __func__, err);
+    }
+    free(sd_ssr);
+    return err;
+}
+
+esp_err_t sdmmc_init_sd_bus_width(sdmmc_card_t* card)
+{
+    int width = 1;
+    if (card->log_bus_width == 2) {
+        width = 4;
+    } else if (card->log_bus_width == 3) {
+        width = 8;
+    }
+    esp_err_t err = sdmmc_send_cmd_set_bus_width(card, width);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "set_bus_width failed (0x%x)", err);
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_sd_wait_data_ready(sdmmc_card_t* card)
+{
+    /* Wait for the card to be ready for data transfers */
+    uint32_t status = 0;
+    uint32_t count = 0;
+    while (!host_is_spi(card) && !(status & MMC_R1_READY_FOR_DATA)) {
+        // TODO: add some timeout here
+        esp_err_t err = sdmmc_send_cmd_send_status(card, &status);
+        if (err != ESP_OK) {
+            return err;
+        }
+        if (++count % 16 == 0) {
+            ESP_LOGV(TAG, "waiting for card to become ready (%d)", count);
+        }
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_send_cmd_switch_func(sdmmc_card_t* card,
+        uint32_t mode, uint32_t group, uint32_t function,
+        sdmmc_switch_func_rsp_t* resp)
+{
+    if (card->scr.sd_spec < SCR_SD_SPEC_VER_1_10 ||
+        ((card->csd.card_command_class & SD_CSD_CCC_SWITCH) == 0)) {
+            return ESP_ERR_NOT_SUPPORTED;
+    }
+
+    if (group == 0 ||
+        group > SD_SFUNC_GROUP_MAX ||
+        function > SD_SFUNC_FUNC_MAX) {
+        return ESP_ERR_INVALID_ARG;
+    }
+
+    if (mode > 1) {
+        return ESP_ERR_INVALID_ARG;
+    }
+
+    uint32_t group_shift = (group - 1) << 2;
+    /* all functions which should not be affected are set to 0xf (no change) */
+    uint32_t other_func_mask = (0x00ffffff & ~(0xf << group_shift));
+    uint32_t func_val = (function << group_shift) | other_func_mask;
+
+    sdmmc_command_t cmd = {
+            .opcode = MMC_SWITCH,
+            .flags = SCF_CMD_ADTC | SCF_CMD_READ | SCF_RSP_R1,
+            .blklen = sizeof(sdmmc_switch_func_rsp_t),
+            .data = resp->data,
+            .datalen = sizeof(sdmmc_switch_func_rsp_t),
+            .arg = (!!mode << 31) | func_val
+    };
+
+    esp_err_t err = sdmmc_send_cmd(card, &cmd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd returned 0x%x", __func__, err);
+        return err;
+    }
+    sdmmc_flip_byte_order(resp->data, sizeof(sdmmc_switch_func_rsp_t));
+    uint32_t resp_ver = SD_SFUNC_VER(resp->data);
+    if (resp_ver == 0) {
+        /* busy response is never sent */
+    } else if (resp_ver == 1) {
+        if (SD_SFUNC_BUSY(resp->data, group) & (1 << function)) {
+            ESP_LOGD(TAG, "%s: response indicates function %d:%d is busy",
+                    __func__, group, function);
+            return ESP_ERR_INVALID_STATE;
+        }
+    } else {
+        ESP_LOGD(TAG, "%s: got an invalid version of SWITCH_FUNC response: 0x%02x",
+                __func__, resp_ver);
+        return ESP_ERR_INVALID_RESPONSE;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_enable_hs_mode(sdmmc_card_t* card)
+{
+    /* This will determine if the card supports SWITCH_FUNC command,
+     * and high speed mode. If the cards supports both, this will enable
+     * high speed mode at the card side.
+     */
+    if (card->scr.sd_spec < SCR_SD_SPEC_VER_1_10 ||
+        ((card->csd.card_command_class & SD_CSD_CCC_SWITCH) == 0)) {
+            return ESP_ERR_NOT_SUPPORTED;
+    }
+    sdmmc_switch_func_rsp_t* response = (sdmmc_switch_func_rsp_t*)
+            heap_caps_malloc(sizeof(*response), MALLOC_CAP_DMA);
+    if (response == NULL) {
+        return ESP_ERR_NO_MEM;
+    }
+
+    esp_err_t err = sdmmc_send_cmd_switch_func(card, 0, SD_ACCESS_MODE, 0, response);
+    if (err != ESP_OK) {
+        ESP_LOGD(TAG, "%s: sdmmc_send_cmd_switch_func (1) returned 0x%x", __func__, err);
+        goto out;
+    }
+    uint32_t supported_mask = SD_SFUNC_SUPPORTED(response->data, 1);
+    if ((supported_mask & BIT(SD_ACCESS_MODE_SDR25)) == 0) {
+        err = ESP_ERR_NOT_SUPPORTED;
+        goto out;
+    }
+    err = sdmmc_send_cmd_switch_func(card, 1, SD_ACCESS_MODE, SD_ACCESS_MODE_SDR25, response);
+    if (err != ESP_OK) {
+        ESP_LOGD(TAG, "%s: sdmmc_send_cmd_switch_func (2) returned 0x%x", __func__, err);
+        goto out;
+    }
+
+out:
+    free(response);
+    return err;
+}
+
+esp_err_t sdmmc_enable_hs_mode_and_check(sdmmc_card_t* card)
+{
+    /* All cards should support at least default speed */
+    card->max_freq_khz = SDMMC_FREQ_DEFAULT;
+    if (card->host.max_freq_khz <= card->max_freq_khz) {
+        /* Host is configured to use low frequency, don't attempt to switch */
+        card->max_freq_khz = card->host.max_freq_khz;
+        return ESP_OK;
+    }
+
+    /* Try to enabled HS mode */
+    esp_err_t err = sdmmc_enable_hs_mode(card);
+    if (err != ESP_OK) {
+        return err;
+    }
+    /* HS mode has been enabled on the card.
+     * Read CSD again, it should now indicate that the card supports
+     * 50MHz clock.
+     * Since SEND_CSD is allowed only in standby mode, and the card is currently in data transfer
+     * mode, deselect the card first, then get the CSD, then select the card again. This step is
+     * not required in SPI mode, since CMD7 (select_card) is not supported.
+     */
+    const bool is_spi = host_is_spi(card);
+    if (!is_spi) {
+        err = sdmmc_send_cmd_select_card(card, 0);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: select_card (1) returned 0x%x", __func__, err);
+            return err;
+        }
+    }
+    err = sdmmc_send_cmd_send_csd(card, &card->csd);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_csd returned 0x%x", __func__, err);
+        return err;
+    }
+    if (!is_spi) {
+        err = sdmmc_send_cmd_select_card(card, card->rca);
+        if (err != ESP_OK) {
+            ESP_LOGE(TAG, "%s: select_card (2) returned 0x%x", __func__, err);
+            return err;
+        }
+    }
+
+    if (card->csd.tr_speed != 50000000) {
+        ESP_LOGW(TAG, "unexpected: after enabling HS mode, tr_speed=%d", card->csd.tr_speed);
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+
+    card->max_freq_khz = SDMMC_FREQ_HIGHSPEED;
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_check_scr(sdmmc_card_t* card)
+{
+    /* If frequency switch has been performed, read SCR register one more time
+     * and compare the result with the previous one. Use this simple check as
+     * an indicator of potential signal integrity issues.
+     */
+    sdmmc_scr_t scr_tmp = { 0 };
+    esp_err_t err = sdmmc_send_cmd_send_scr(card, &scr_tmp);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: send_scr returned 0x%x", __func__, err);
+        return err;
+    }
+    if (memcmp(&card->scr, &scr_tmp, sizeof(scr_tmp)) != 0) {
+        ESP_LOGE(TAG, "got corrupted data after increasing clock frequency");
+        return ESP_ERR_INVALID_RESPONSE;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_init_spi_crc(sdmmc_card_t* card)
+{
+    /* In SD mode, CRC checks of data transfers are mandatory and performed
+     * by the hardware. In SPI mode, CRC16 of data transfers is optional and
+     * needs to be enabled.
+     */
+    assert(host_is_spi(card));
+    esp_err_t err = sdmmc_send_cmd_crc_on_off(card, true);
+    if (err != ESP_OK) {
+        ESP_LOGE(TAG, "%s: sdmmc_send_cmd_crc_on_off returned 0x%x", __func__, err);
+        return err;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_decode_cid(sdmmc_response_t resp, sdmmc_cid_t* out_cid)
+{
+    out_cid->mfg_id = SD_CID_MID(resp);
+    out_cid->oem_id = SD_CID_OID(resp);
+    SD_CID_PNM_CPY(resp, out_cid->name);
+    out_cid->revision = SD_CID_REV(resp);
+    out_cid->serial = SD_CID_PSN(resp);
+    out_cid->date = SD_CID_MDT(resp);
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_decode_csd(sdmmc_response_t response, sdmmc_csd_t* out_csd)
+{
+    out_csd->csd_ver = SD_CSD_CSDVER(response);
+    switch (out_csd->csd_ver) {
+    case SD_CSD_CSDVER_2_0:
+        out_csd->capacity = SD_CSD_V2_CAPACITY(response);
+        out_csd->read_block_len = SD_CSD_V2_BL_LEN;
+        break;
+    case SD_CSD_CSDVER_1_0:
+        out_csd->capacity = SD_CSD_CAPACITY(response);
+        out_csd->read_block_len = SD_CSD_READ_BL_LEN(response);
+        break;
+    default:
+        ESP_LOGE(TAG, "unknown SD CSD structure version 0x%x", out_csd->csd_ver);
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    out_csd->card_command_class = SD_CSD_CCC(response);
+    int read_bl_size = 1 << out_csd->read_block_len;
+    out_csd->sector_size = MIN(read_bl_size, 512);
+    if (out_csd->sector_size < read_bl_size) {
+        out_csd->capacity *= read_bl_size / out_csd->sector_size;
+    }
+    int speed = SD_CSD_SPEED(response);
+    if (speed == SD_CSD_SPEED_50_MHZ) {
+        out_csd->tr_speed = 50000000;
+    } else {
+        out_csd->tr_speed = 25000000;
+    }
+    return ESP_OK;
+}
+
+esp_err_t sdmmc_decode_scr(uint32_t *raw_scr, sdmmc_scr_t* out_scr)
+{
+    sdmmc_response_t resp = { 0 };
+    resp[1] = __builtin_bswap32(raw_scr[0]);
+    resp[0] = __builtin_bswap32(raw_scr[1]);
+    int ver = SCR_STRUCTURE(resp);
+    if (ver != 0) {
+        return ESP_ERR_NOT_SUPPORTED;
+    }
+    out_scr->sd_spec = SCR_SD_SPEC(resp);
+    out_scr->erase_mem_state = SCR_DATA_STAT_AFTER_ERASE(resp);
+    out_scr->bus_width = SCR_SD_BUS_WIDTHS(resp);
+    return ESP_OK;
+}
+
+static const uint32_t s_au_to_size_kb[] = {
+    0, 16, 32, 64,
+    128, 256, 512, 1024,
+    2 * 1024, 4 * 1024,
+    8 * 1024, 12 * 1024,
+    16 * 1024, 24 * 1024,
+    32 * 1024, 64 * 1024
+};
+_Static_assert(sizeof(s_au_to_size_kb)/sizeof(s_au_to_size_kb[0]) == 16, "invalid number of elements in s_au_to_size_kb");
+
+esp_err_t sdmmc_decode_ssr(uint32_t *raw_ssr, sdmmc_ssr_t* out_ssr)
+{
+    uint32_t ssr[(SD_SSR_SIZE/sizeof(uint32_t))] = { 0 };
+    size_t j = (SD_SSR_SIZE/sizeof(uint32_t) - 1);
+
+    for(size_t i = 0; i < (SD_SSR_SIZE/sizeof(uint32_t)); i++) {
+        ssr[j - i] = __builtin_bswap32(raw_ssr[i]);
+    }
+
+    out_ssr->cur_bus_width = SSR_DAT_BUS_WIDTH(ssr);
+    out_ssr->discard_support = SSR_DISCARD_SUPPORT(ssr);
+    out_ssr->fule_support = SSR_FULE_SUPPORT(ssr);
+    uint32_t au = SSR_AU_SIZE(ssr);
+    out_ssr->alloc_unit_kb = s_au_to_size_kb[au];
+    out_ssr->erase_timeout = SSR_ERASE_TIMEOUT(ssr);
+    out_ssr->erase_size_au = SSR_ERASE_SIZE(ssr);
+    out_ssr->erase_offset = SSR_ERASE_OFFSET(ssr);
+
+    return ESP_OK;
+}
+
+uint32_t sdmmc_sd_get_erase_timeout_ms(const sdmmc_card_t* card, int arg, size_t erase_size_kb)
+{
+    if (arg == SDMMC_SD_DISCARD_ARG) {
+        return SDMMC_SD_DISCARD_TIMEOUT;
+    } else if (arg == SDMMC_SD_ERASE_ARG) {
+        if (card->ssr.alloc_unit_kb != 0 &&
+                card->ssr.erase_size_au != 0 &&
+                card->ssr.erase_timeout != 0 &&
+                card->ssr.erase_offset != 0) {
+            /* Card supports erase timeout estimation. See the erase timeout equation in SD spec. */
+            uint32_t timeout_sec = card->ssr.erase_offset +
+                card->ssr.erase_timeout * (erase_size_kb + card->ssr.alloc_unit_kb - 1) /
+                    (card->ssr.erase_size_au * card->ssr.alloc_unit_kb);
+            ESP_LOGD(TAG, "%s: erase timeout %u s (erasing %u kB, ES=%u, ET=%u, EO=%u, AU=%u kB)",
+                     __func__, timeout_sec, erase_size_kb, card->ssr.erase_size_au,
+                     card->ssr.erase_timeout, card->ssr.erase_offset, card->ssr.alloc_unit_kb);
+            return timeout_sec * 1000;
+        } else {
+            uint32_t timeout_ms = SDMMC_SD_DISCARD_TIMEOUT * erase_size_kb / card->csd.sector_size;
+            timeout_ms = MAX(1000, timeout_ms);
+            ESP_LOGD(TAG, "%s: erase timeout %u s (erasing %u kB, %ums per sector)",
+                     __func__, timeout_ms / 1000, erase_size_kb, SDMMC_SD_DISCARD_TIMEOUT);
+            return timeout_ms;
+        }
+    } else {
+        assert(false && "unexpected SD erase argument");
+        return 0;
+    }
+}