- May 07, 2024
-
-
Ahmed Ismail authoredThese additions are mainly inherited from this document: `https://git.gitlab.arm.com/iot/open-iot-sdk/sdk/-/blob/ main/components/trusted-firmware-m/README.md`. However, some modifications were made to match the FRI's project use case. Signed-off-by:
Ahmed Ismail <Ahmed.Ismail@arm.com>
-
Hugues Kamba Mpiana authoredAdd documentation related to the integration and usage of MbedTLS from within the FRI. Signed-off-by:
Hugues Kamba-Mpiana <hugues.kambampiana@arm.com> -
Hugues Kamba Mpiana authored
Linking the `mbedtls-threading-freertos` CMake library to `mbedtls-config` at the FRI component integration level saves applications from having to include it thus simplifyig the usage of the MbedTLS component. Signed-off-by:Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>
-
- Apr 29, 2024
-
-
Signed-off-by: -
`cspell` is used as the spell checker tool to follow Public GitHub repository spell check process. Signed-off-by:
-
- Apr 22, 2024
-
-
CS315 support have been added to the FreeRTOS-Plus-TCP. Bumped the submodule's commit hash to the one which added the new platform. Signed-off-by:Dávid Házi <david.hazi@arm.com>
-
- Apr 16, 2024
-
-
Devaraj Ranganna authored
Signed-off-by:Devaraj Ranganna <devaraj.ranganna@arm.com>
-
Signed-off-by: -
The idea is to move the ML model to the DDR memory region, generate a separate image for the DDR memory region where the model can be over the air updated without the need to update the NS image. To do so the following changes are needed: * Modify the linker scripts for all platforms to remove the RAM load address of the ML model image as it should be loaded to DDR memory region. * Some linker sections were renamed to be more descriptive. * A new CMake function is introduced to extract sections from input AXF file, this function is used to extract DDR section which contains the ML model out of the NS image to be loaded separately. Signed-off-by: -
It seems like the speech recognition DSP task which is collecting the input data from ROM/VSI audio source needs to be a bit slower as in case of using software to do the ML inference the performance is way slower than using the Ethos-NPU which means the DSP task must be delayed for sometime as to allow some extra time for the software inference to be finished. Signed-off-by:
-
- Apr 10, 2024
-
-
This commit include the following changes: * Move testing ML applications built with GNU toolchain to nightly build scheduled pipeline. * Move testing applications with software inference to nightly build scheduled pipeline. * Move testing applications with VSI as input audio source to nightly build scheduled pipeline. Signed-off-by: -
Pipeline matrices are divided based on the targets to decrease the number of build jobs needed for `test-applications` job as they would exceed the limit (50 jobs) after adding support for software inference with speech-recognition application for all platforms, the limitation of 50 jobs comes from `needs` keyword used in `test-applications` job. Signed-off-by: -
Devaraj Ranganna authored
Incorrect towncrier configuration is causing changelog check job failure. Fixed incorrect towncrier configuration. Also, added `.gitkeep` to `release_changes` folder, so that git won't ignore it during a new release when all files under `release_changes` are deleted. Signed-off-by:
-
- Mar 28, 2024
-
-
Devaraj Ranganna authored
* Updated `CHANGELOG.md` and `manifest.yml` * Fixed incorrect config in `pyproject.toml` * Updated to latest release `v0.3.0` of `arm-corstone-platform-bsp` Signed-off-by: -
Devaraj Ranganna authored
* Install missing `libXcursor` package needed by Fast-models third party add-ons installer * Install missing `libncursesw5` package needed by arm-none-eabi-gdb * Install ninja build tool * Fix incorrect link for `container_post_startup.sh` in devcontainer.json * Improve documentation based on OOB feedback Signed-off-by: -
blinky application doesn't connect to AWS so there is no need to have MQTT and OTA tasks configurations as part of its application configuration. Signed-off-by: -
Signed-off-by: -
Signed-off-by:
-
- Mar 24, 2024
-
-
Devaraj Ranganna authored
The `NPU_ID` defines the NPU to be used and it is defaulted to `U65` which causes issues with Corstone-300 and Corstone-310 which are built with `U55` as default NPU. Therefore, setting `NPU_ID` to null string leads to using `U55` for Corstone-300 and Corstone-310 and `U65` for Corstone-315 as default NPU respectively. Signed-off-by: -
Devaraj Ranganna authored
Signed-off-by: -
Devaraj Ranganna authored
With the introduction of VS Code build and debug support in FRI using devcontainers, the development environment becomes headless (no display). This causes issues with Corstone-315 and object detection example. Until, we find a solution to X11 forwarding with devcontainers, a command-line option is provided to disable HDLCD display in Corstone-315 when running in headless mode. Signed-off-by: -
Devaraj Ranganna authored
Signed-off-by: -
Devaraj Ranganna authored
VS Code tasks configuration used for automating building and debugging, doesn't support optional arguments. One of the possible solutions is to pass both `command-line option` (--toolchain) and `value` (GNU) as VS Code tasks `promptString`, but a space must be added between `command-line option` and `value`. VS Code encapsulates this `promptString` with a space in double quotes causing shell not recognising this as a valid command-line option. Other solution is to introduce a short `command-line option` (-T, toolchain), so that both `command-line option` and `value` can be passed in a `promptString` without space. If this is an optional argument, then an empty string can be passed as `promptString`. Signed-off-by: -
Devaraj Ranganna authored
- Remove `docs` from `ignorePaths` to run spell-check on FRI documentation - Remove `History.txt` from `ignorePaths` which is not part of FRI - Fix spell-check errors Signed-off-by: -
Devaraj Ranganna authored
The improvements include: * Correcting spelling mistakes * Structure of the documents for better readability Signed-off-by: -
Devaraj Ranganna authored
With all documentation files added under `docs` folder, finding a relevant document was a time consuming process. Restructure the docs folder to replicate the FRI source structure for easy navigation and better user experience. Signed-off-by: -
Devaraj Ranganna authored
Add VS Code tasks and launch configuration files to enable building and debugging FRI using VS Code. Signed-off-by: -
Devaraj Ranganna authored
Create `devcontainer.json` and the `Dockerfile` to bootstrap the devcontainer when FRI project is cloned using VS Code command palette. Signed-off-by: -
Devaraj Ranganna authored
In order to support VS Code build and debug support, add debug option `-d` to run script. In addition, use `realpath` to get absolute path for build folder. Signed-off-by:
-
- Mar 18, 2024
-
-
When building ML apps, the MLEK downloads model resources for the model specified by a given application. The FRI integration of the MLEK prevents these resources from being re-donwloaded if they already exist to speed up build time. However, in the case of a different application being build, model resources have to be re-downloaded as different ML applications use different models. To allow the re-download of model resources, the generated ML model resources subdirectory is now generated in the CMake build directory. Building a different application requires triggering a clean build which will also delete the resources specific to a model used by another application. Signed-off-by:Hugues Kamba Mpiana <hugues.kambampiana@arm.com>
-
- Mar 17, 2024
-
-
Hugues Kamba-Mpiana authored
The `freertos-iot-libraries-tests` application was mistakenly relying on another app's TF-M configuration file. The other app (`aws-iot-example`) was removed along with its configuration files. The `freertos-iot-libraries-tests` application now uses its own TF-M configuration file. Signed-off-by:Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>
-
- Mar 14, 2024
-
-
Corstone-315 added to the blinky and to the ML/OTA tests too. Signed-off-by: -
Object detection example is an ML vision example on the Cortex-M systems. Currently running only on Corstone-315 to demonstrate the usage of ISP, NPU, HDLCD together. Signed-off-by:Gabor Abonyi <gabor.abonyi@arm.com>
-
Starting bootloader printed by BL2. Corstone-300 and 310 starting the BL2 first, than do the provisioning part. Corstone-315 do the provisioning first, because of the BL1 ROM bootloader, therefore the log messages comes swapped from this platform. We have deleted the Starting bootloader line to unify the pass output file. Signed-off-by: -
The HDLCD helper adds 3 predefined display formats to the project, RGB565, RGB32 and A2R10G10B10. Signed-off-by: -
The tools has been added to convert images to .FRM format, which can be fed into the Corstone-315 FVP's "camera" input. Signed-off-by: -
Generates `You Only Look Once` (YOLO) fastest object detection model for the object detection application Signed-off-by: -
Arm-2D is used by the object detection example to draw frames around detected faces. Signed-off-by: -
Using the latest platform, the Corstone-315 as default. Signed-off-by: -
Arm Mali-C55 image signal processor (ISP) driver The Corstone-315 has an ISP IP, which can be used to create downscaled image for ML vision applications. ISP also can handle camera lens, sensors. The driver comes with a 3A algorithm (proprietary) Signed-off-by:
-
