Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Jakub Sujak authored Jul 04, 2024 and Felix Johnny Thomasmathibalan committed Jul 04, 2024

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Reviewed-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

This example demonstrates the usage of the FP16 packing and matmul routines which:

1. Packs the bias and the weights together into a single tensor.

2. Performs a matrix multiplication of the activations and the packed tensor.

All tensors are in half precision floating-point (FP16) data type.

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Felix Johnny Thomasmathibalan authored Jul 02, 2024 and Jakub Sujak committed Jul 02, 2024

- Adjust the int4 matmul micro-kernels (qsi4cx) to skip the bias

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Jens Elofsson authored Jul 02, 2024 and Jakub Sujak committed Jul 02, 2024

Signed-off-by: Jens Elofsson <jens.elofsson@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Anton Bondarenko <anton.bondarenko@arm.com>

Add file and line information to KAI_ERROR-macro.

Signed-off-by: Jens Elofsson <jens.elofsson@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Prepare and upload JUnit report for tests execution
* Simplify pre-commit job
* Replace dependencies with needs to unblock jobs

Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

New hook would be able to automatically download specified version
of the Buildifier tool instead of relying on local one.

Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Add compiler version checks in CMake before adding
feature specific (e.g. FEAT_DOTPROD) micro-kernels.

* Enable compiling scalar routines using MSVC.

Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Approved-by: Jens Elofsson <jens.elofsson@arm.com>

The _dst_lhs_rhs_ naming is used instead of
hybrid for the test function name to follow
the format used in the main repository.

Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Add SME2 GeMM kernels in FP32.
  - Kernels: LHS packing, RHS packing, interleaved GeMM.
* Update the build script and the test framework.
* Fix issues in the API.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Use Bazelisk to perform Bazel builds instead of relying directly
on specific Bazel version. At the moment latest LTS is chosen, but
could be later updated to a specific version if required

Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Add more checks in the test to verify all utility functions
  provided by GeMM kernels.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

"Hello World"-style example of the microbenchmark tool that pipecleans
the cmake-build.

Signed-off-by: Jens Elofsson <jens.elofsson@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Anton Bondarenko <anton.bondarenko@arm.com>

Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Reorder the arguments of getting RHS offset function.
* Regenerate the assembly kernel.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Run full Linux in FVP to test features without silicon.
* Collect coverage report from multiple runs.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Update the base image for the Docker image to Ubuntu 24.04
  which is the latest LTS version of Ubuntu.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

* Add code coverage build and test.
* Add coverage report and visualization.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

stddef.h is included for size_t

Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Gian Marco Iodice authored Jun 07, 2024 and Jakub Sujak committed Jun 07, 2024

- Add the free memory section
- Fix the input arguments for the matmul micro-kernel

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

* Explicitly build NEON-only files with the base architecture.
* Remove redundant compiler flags.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Gian Marco Iodice authored Jun 05, 2024 and Viet-Hoa Do committed Jun 05, 2024

- Move lhs_stride position in the list of input arguments
- Add dst_stride_col in the list of input arguments
- Add the interface header file

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Gian Marco Iodice authored Jun 03, 2024 and Viet-Hoa Do committed Jun 03, 2024

- Add guide to explain how to invoke the int4 matmul micro-kernel

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Gian Marco Iodice authored Jun 03, 2024 and Viet-Hoa Do committed Jun 03, 2024

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Gian Marco Iodice authored May 31, 2024 and Viet-Hoa Do committed May 31, 2024

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>

* The library is move into kai directory (previously src).
* All microkernels are in kai/ukernels directory.
* Packing kernels for matmul will be in kai/ukernels/matmul/pack
  directory.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

* Add GeMM hybrid kernels.
  - Block size: 6x16.
  - Data type: FP16.
* Update test framework and add tests for new kernels.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Approved-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Gian Marco Iodice authored May 29, 2024 and Viet-Hoa Do committed May 29, 2024

- Make RHS packing function accessing in-bound elements when N is not
multiple of nr
- Make LHS packing function accessing in-bound elements when K is not
multiple of kr * sr

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Gian Marco Iodice authored May 23, 2024 and Felix Johnny Thomasmathibalan committed May 23, 2024

- The LHS matrix is quantized (Q) Asymmetric (A) 8-bit (8) with per-row (DX) quantization parameters
- The RHS matrix is quantized (Q) Symmetric (S) 4-bit (4) with per-channel (cx) quantization parameters
- The destination is F32
- Implement matmul int4 micro-kernels with intrinsics by using the dotprod and i8mm extensions
- Implement a micro-kernel to pack the RHS matrix
- Implement two micro-kernels to dynamically quantize and pack the LHS matrix
- Add README.md
- No test added into this PR. Test will be added in a separate PR

Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>

* Also remove trailing whitespace.

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>

Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>

Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>