Anton Bondarenko authored Aug 30, 2024 and Jakub Sujak committed Aug 30, 2024

Code coverage tools used in the project has a strong limitation
for location of source code used to compile with coverage. This
source code must be in exactly same absolute path when running
coverage processing tool. It's not easy to guarantee when several
GitLab runners are used in project CI as these have different paths
for source code location. Improve coverity pipeline where source code,
execution and report generation could be in different folders.

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

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

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

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

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

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

Michael Kozlov authored Aug 28, 2024 and Felix Johnny Thomasmathibalan committed Aug 28, 2024

Signed-off-by: Michael Kozlov <michael.kozlov@arm.com>

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

Examples is a vital part of KleidiAI delivery and need to be checked
in a same way as core library and unit tests.

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

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

Jakub Sujak authored Aug 19, 2024 and Felix Johnny Thomasmathibalan committed Aug 19, 2024

* Add test coverage for the matmul_clamp_f32_qsi8d32p1x8_qsi4c32p4x8_1x4x32_neon_dotprod micro-kernel.

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

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

Jakub Sujak authored Aug 19, 2024 and Felix Johnny Thomasmathibalan committed Aug 19, 2024

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

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

Gian Marco Iodice authored Aug 16, 2024 and Jakub Sujak committed Aug 16, 2024

* The LHS matrix is Quantized (Q) Symmetric (S) Signed 8-bit (I8) with per-block quantization (D32) quantization parameters
* The RHS matrix is Quantized (Q) Symmetric (S) Signed 4-bit (I4) with per-block quantization(C32) F16 scale factors,
* The destination is F32
* Implement micro-kernels to perform the matrix multiplication
* Implement a micro-kernel to pack the LHS and RHS matrices
* Added unit tests

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

Signed-off-by: Anitha <Raj Anitha.Raj@arm.com>

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

* Tested kernels:
  - lhs_quant_pack_qai8dxp_f32
  - rhs_pack_nxk_qsi4cxp_qsu4cxs1s0
  - matmul_clamp_f32_qai8dxp4x8_qsi4cxp8x8_8x8x32_neon_i8mm
* Only entire matrix is tested - support for testing arbitrary
  portion of the output will come later.
* Update test framework to accommodate the new tests.
  - The new test no longer relies on DataFormat concept.
    Now the packing function for each format is explicitly called.

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>

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

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

This reverts commit 94aa71e6.

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

Block size: 6x8

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

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

Update so CMake and Bazel have the same '-march' compiler
option for files.

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

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

* Compute the general matrix-vector (GEMV) multiplication between an FP16 LHS and RHS and accumulate into FP16 output. The RHS packs FP16 weights and biases together.

* Optimized for Arm® Neon™ using MLA instructions.

* Add accompanying tests.

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

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

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

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

Michael Kozlov authored Aug 01, 2024 and Felix Johnny Thomasmathibalan committed Aug 01, 2024

Signed-off-by: Michael Kozlov <michael.kozlov@arm.com>

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

* Make it clearer and remove redundant unnecessary text.

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

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

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

* It informs the user of potential security weaknesses
  that their product must mitigate.

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

* Use baremetal compiler to compile FVP image components
* Add NDK for x86_64 target

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

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

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

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

Curl is an alternative for wget and used by some external tools.
So it's good to have both for flexibility.

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

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

New hook uses Bazel tools v6.4.0 and supports
common development environments.

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

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

Allows to pass external arguments by using kwargs to cc_library
rule.

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

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

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

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

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

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

Bazel is one of the build system used in the modern software world and
especually in ML OSS projects. Support for Bazel build
would allow simpler intergration into such frameworks.

Currently supported:
* Build microkernels with required CPU micro-architecture
* Combine into common library (matmul)
* Tests compilation and execution

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

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

* Fix condition for left-over rows store.

Resolves: COMPMID-7245
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>

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

* Add instructions for building for various platforms using CMake.

* Provide a CMake toolchain file for Arm GNU Toolchain.

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

Approved-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Approved-by: Jens Elofsson <jens.elofsson@arm.com>

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

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

Jens Elofsson authored Jul 05, 2024 and Jakub Sujak committed Jul 05, 2024

Runs on Linux® and Android™.

Tests the different variations of the matmul_clamp_f32_qai8dxp_qsi4cxp
microkernel with user-specified m-, n-, and k-values.

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

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

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

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

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>