* Move supported toolchain information from application docs to
README.md to avoid duplicatiing it on all supported examples.
* Move supported platforms description from docs/prerequisites.md to
README.md to list and describe the supported platforms in one place.
* Add supported versions information to supported targets.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

The default target for FRI is Corstone-310. Update the
`docs/prerequisites.md` instructions to download Corstone-310 FVP.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

In the initial days of FRI, vht running in ami was the only supported
platform since it supported virtual socket interface. However, latest
ecosystem FVPs do support virtual socket interface. In addition, FRI now
supports FreeRTOS TCP/IP stack. Therefore, remove vht support.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

Gabor Abonyi authored Feb 22, 2024 and Devaraj Ranganna committed Feb 25, 2024

Add missing PARENT_SCOPE to root folder CMakeLists.txt.
It is needed as the entry point is the application/<example>.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

All the components across the project should use
the `iot_reference_arm_corstone3xx_apply_patches`
function to apply patches which in turn improves
consistency. For more information,
please refer to the commit with SHA:
fa76f068

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Move common settings and values from the applications to the bsp and
root CMakeLists.txt files. These values are common across all apps.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

* ml-eval-kit: Make ETHOSU_ARCH configurable in TensorFlow CMake

This patch removes the hardcoded value of u55 in tesorflow.cmake.
It now uses ETHOS_U_NPU_ID as input and convert it to the format
needed by TensorFlow build.
This also means the ETHOS_U_NPU_ID variable preserve workaround is
no longer needed in SetupMlEmbeddedEvaluationKitLibraries.cmake as
it is no longer overwritten in tensorflow.cmake.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

* ml-eval-kit: Move NPU config from app to bsp

These settings are independent from the applications, to reduce
duplication, configuration of NPU is moved to bsp layer.
Added U65 option and checks for config validity.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

* ml-eval-kit: Move activation_buf_dram to .bss.NoInit region

This is needed to place activation_buf_dram in a no init region, so
it does not require space in flash.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

* build-run: Add support for Ethos NPU ID and MAC cycle configuration

* Add option to the build/run scripts to select different Ethos NPU
* Add option to the buld script to select number of Ethos NPU MAC
  unit per cycle

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

* build-run: Use a shell variable to gather CMake arguments

This is so some CMake arguments can be excluded more easily
depending on the build configuration. It will remove the need
to set default values for CMake cache variables from the scripts
as CMake cache variables already have default values.

This change allows a standard overriding of CMake cache variable
in CMake scripts without a check to see if the user input a value
for the cache variable.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

* build: Unify error report format in build.sh

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

---------

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

Filip Jagodzinski authored Jan 22, 2024 and Devaraj Ranganna committed Feb 19, 2024

Since both the certificate path and the private key path are forwarded
to CMake, the user is asked to specify them as absolute. Build process
fails when either of these is a local path instead. Unfortunately,
although our build script checks for existence of both files, the check
is not done from CMake's point of view. As a result, the build error
seen by the user is hard to understand (no known rule to make a missing
ninja target).

Resolve both AWS certificate path and AWS private key path to prevent
the unhelpful error message from showing up, and make the build script
more flexible by giving users option to use local paths as well as
absolute paths for PEM files.

Signed-off-by: Filip Jagodzinski <filip.jagodzinski@arm.com>

Hugues Kamba Mpiana authored Feb 13, 2024 and Devaraj Ranganna committed Feb 14, 2024

Change the optimization level for the `freertos-iot-libraries-tests`
application to reduce code size whilst still maintaining a
comfortable debugging experience.

Arm Compiler For Embedded optimization level is set to use `-O1` and
Arm GNU Toolchain is set to use `-Og`.
These optimization levels are already used by the other FRI
applications.

The Arm Compiler For Embedded `-O1` optimization level results
in a slower binary execution speed on `FVP_Corstone_SSE-310`.
The MQTT test process loop timeout period
(`MQTT_TEST_PROCESS_LOOP_TIMEOUT_MS`) needs therefore to be
increased from `40000` to `54630` to pass the
`MqttTest, MQTT_Restore_Session_Duplicate_Incoming_Publish_Qos1`
test case.

Signed-off-by: Hugues Kamba Mpiana <hugues.kambampiana@arm.com>

Improvements include:
* Use descriptive names for symbols in lines with other test scripts
* Add docstrings
* Add type annotations
* Use Python assignment operators where appropriate
* Use f-strings to make print statement more readable

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

Instead of repeating the executable name in the application
`CMakeLists.txt`, use `CMAKE_PROJECT_NAME`.
This change ensures that the project name matches the CMake
target executable name.

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

Rename the AWS IoT Example application CMake project name from
`aws-iot-example-project` to `aws-iot-example`.
The addition of `-project` suffix was redundant as it is the value of
the `CMAKE_PROJECT_NAME` CMake variable.

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

The `aws-iot-example-tests` executable's source code and
CMake build instructions were contained within the AWS IoT Example
application. It has now been decoupled as a standalone application
named `freertos-iot-libraries-tests`.

The `freertos-iot-libraries-tests` application verifies this FRI's
integration of FreeRTOS IoT libraries running on the supported
hardware platforms for robustness and compatibility with AWS’s
published best practices for AWS IoT Core connectivity.

Other changes include:
* Use `CMAKE_PROJECT_NAME` to refer to the apps executable
* Include header files that declares objects used in source files

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

* Declaring MBEDTLS_ALLOW_PRIVATE_ACCESS allows access
  to mbedtls "private" fields.
* Migration guide:
https://github.com/Mbed-TLS/mbedtls/blob/development/docs/3.0-migration-guide.md



Signed-off-by: Dávid Házi <david.hazi@arm.com>

Ahmed Ismail authored Jan 15, 2024 and Devaraj Ranganna committed Feb 07, 2024

The `otaPal_CloseFile_ValidSignature` test
case was previously removed as it kept failing.
Restore it as it has been fixed.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Ahmed Ismail authored Jan 17, 2024 and Devaraj Ranganna committed Feb 07, 2024

For the integration test app, select the code signing public key .Pem
file within test directory of `FreeRTOS-Libraries-Integration-Tests`
for generating the credentials header file.

This static public key is necessary to verify the test signature
using the same key pair used to generate the signature during
provisioning.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Ahmed Ismail authored Jan 15, 2024 and Devaraj Ranganna committed Feb 07, 2024

The fixes are provided with a patch that does the following:
* Make the binary file path configurable using
  a macro (`OTA_PAL_FIRMWARE_FILE`) instead of
  hardcoding it as `test_happy_path_image.bin`.
  This is so it can be specified as expected by the
  AWS test suite.
* Replace the RSA SHA256 signature as it is invalid.
  The replacement signature has been regenerated using
  a different key pair with the public key also added
  to the `src/ota/test_files` subdirectory.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Gabor Abonyi authored Jan 17, 2024 and Devaraj Ranganna committed Feb 06, 2024

* Improved tensorflow.cmake
* Model URLs file location has changed
* Some patches have been merged upstream

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

Gabor Abonyi authored Jan 17, 2024 and Devaraj Ranganna committed Feb 06, 2024

Instead of building the whole project as an external project,
only build the necessary libraries.

Rename BuildMLEmbeddedEvaluationKit.cmake to
SetupMlEmbeddedEvaluationKitLibraries.cmake

This change also uses the NPU driver component from
ML Evaluation Kit which results in the following changes:
* Modify initialization of NPU in AI applications to use
  API provided by the component
* Add ICache and DCache enablement options as caches are used
  in the NPU component
* Include various patches for the NPU component.
* Remove ML Evaluation Kit device drivers header files
  inclusions as they are no longer used as we are now using
  the NPU component for initializing the NPU.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

Gabor Abonyi authored Jan 18, 2024 and Devaraj Ranganna committed Feb 06, 2024

Use the -additional-ethos-u-config-name option to generate
required NPU optimization instead of patching default list.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

Gabor Abonyi authored Jan 17, 2024 and Devaraj Ranganna committed Feb 06, 2024

This prevents the module from always being included
even when building applications that do not require it.

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

Ahmed Ismail authored Feb 05, 2024 and Devaraj Ranganna committed Feb 06, 2024

This commit is meant to do the following:

* Modify test-application-output job to test blinky
application only as it can be used as a smoke test.

* Test ML inference output for both keyword-detection and
speech-recognition applications as part of the OTA test job
which is now renamed to test_applications as it is testing
OTA update and ML inference outputs as well.

Note:
* The mentioned changes would result in a better reliability
and better CI performance.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Ahmed Ismail authored Feb 02, 2024 and Devaraj Ranganna committed Feb 06, 2024

FVP VSI python workaround shall be done as part
of the docker image generation rather than being
done for each VSI  test which should significantly
reduce CI execution time.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Hugues Kamba Mpiana authored Feb 02, 2024 and Devaraj Ranganna committed Feb 06, 2024

Ensure the URL of the files reported in
the TPIP comment added to the Merge Request
contains the correct branch name.

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

Ahmed Ismail authored Feb 02, 2024 and Devaraj Ranganna committed Feb 06, 2024

For non-ML applications there is no need to build and run for
AUDIO and INFERENCE variants as these configurations should
be relevant for ML applications only. Non-ML applications would
use the default values only. Thus, separating non-ML
applications pipeline configurations from the ML applications
pipeline configurations.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

Gabor Abonyi authored Jan 19, 2024 and Devaraj Ranganna committed Feb 06, 2024

Signed-off-by: Gabor Abonyi <gabor.abonyi@arm.com>

Ahmed Ismail authored Jan 17, 2024 and Devaraj Ranganna committed Feb 05, 2024

Fix firmware update cancellation and cleanup check
by ensuring the result variable is initialized to
the success value. The result variable was only set
in case of failure if the cancellation and/or the
cleanup failed.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

In order for one of the patches to be be cleanly applied, git config
option `core.autocrlf` must be set to value `input`.

The file `src/transport_interface/transport_interface_test.c` has `CRLF`
line ending and patching it fails if `core.autocrlf` is not set to
`input`.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

The `TRANSPORT_TEST_BUFFER_WRITABLE_LENGTH` is set to `2048`. Reduce it
to `1350`, so that the test data can be sent in a single ethernet
packet.

Without this change, it takes a very long time to complete transport
tests and most of the times test fails.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

With a lower timeout value (750), socket connection sporadically fails.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

Use FreeRTOS memory management APIs instead of using C library ones, as
FRI is configured to use FreeRTOS memory management.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

The integration tests validates both secure and unsecure connections
over TCP. However, unsecure TCP connection support was missing from
FreeRTOS TCP/IP integration layer.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

* Change `ECHO_SERVER_ENDPOINT` from `localhost` to `127.0.0.1` as DNS
resolution of `localhost` fails with FreeRTOS-TCP/IP. Therefore
hardcode the numerical value.
* Increase the transport interface test timeout to its maximum value by
overriding the `TRANSPORT_TEST_WAIT_THREAD_TIMEOUT_MS` macro with value
`0xFFFFFFFF`. This is required to ensure the nightly test in the
internal CI passes.
* Increase MQTT process loop timeout, so that the expected packet from
the AWS IoT service is received successfully.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

Provide the device private key and certificate to the build script as
parameters. In addition, provide timeout value to integration test
script as parameter.

Signed-off-by: Devaraj Ranganna <devaraj.ranganna@arm.com>

Ahmed Ismail authored Jan 30, 2024 and Devaraj Ranganna committed Feb 05, 2024

This new way of applying patches is added to fix Corstone300 platform
hardfault exception happening as the TF-M patch
(components/security/trusted_firmware-m/integration/patches
/0001-corstone300-Modify-ISRAM0-to-be-non-secure-partition.patch)
for this platform is sometimes not applied.

Patches are now applied using `git stash` and
`git apply`. This methodology is much better as
it reports whenever the new CMake function fails
to apply a certain patch.

Note:

This solution is based on what TF-M implemented in
`components/security/trusted_firmware-m/library/cmake/
remote_library.cmake`

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

The project is tested in the internal CI/CD with Arm Compiler
For Embedded version 6.21. The previously used version was
version 6.18.

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

Add support for testing ML applications with VSI
as the audio source.

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>

This commit is meant to enable the users to build and run
keyword-detection and speech-recognition ML applications
using VSI(Virtual Streaming Interface) as the input audio
source.

`core_clk.mul=200000000` clock rate multiplier FVP option
is added to increase the clock rate which result in faster
and more precise ML inference.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

It is better to use task notifications in VSI callback
function as this functions is called from an interrupt
service routine where it is not a good practice to execute
functions in its context but rather than that task notifications
should be used to keep the interrupt service routine simple.

Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>

The script is added to provide support for Virtual Streaming
Interface (VSI) as the audio source for ML applications with
audio input files.

Signed-off-by: Hugues Kamba-Mpiana <hugues.kambampiana@arm.com>
Signed-off-by: Ahmed Ismail <Ahmed.Ismail@arm.com>