Emil Ohlsson authored Feb 17, 2025 and Jakub Sujak committed Feb 17, 2025

Add support for GEMV like kernel for producing QAI8 from QAI8 LHS and
QSI8CXP packed RHS.

Update unit tests to include support for new kernel

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

Reviewed-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Reviewed-by: Anton Bondarenko <anton.bondarenko@arm.com>
Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Michael Kozlov authored Feb 12, 2025 and Emil Ohlsson committed Feb 12, 2025

- Update build script
- Add to unit test

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

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

Anitha Raj authored Feb 11, 2025 and Anton Bondarenko committed Feb 11, 2025

Resolves: #KLEIDIAI-485, #COMPMID-8024

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

Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Approved-by: Anton Bondarenko <anton.bondarenko@arm.com>

Michael Kozlov authored Feb 06, 2025 and Jakub Sujak committed Feb 06, 2025

- Update build script
- Add to unit test

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

Approved-by: Anitha Raj <anitha.raj@arm.com>

Suhail M authored Jan 31, 2025 and Felix Johnny Thomasmathibalan committed Jan 31, 2025

- Added demonstration of klediai integration as external dependency 

Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>

Reviewed-by: Mohammed Suhail Munshi <mohammedsuhail.munshi@arm.com>
Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Approved-by: Emil Ohlsson <emil.ohlsson@arm.com>

Anton Bondarenko authored Jan 30, 2025 and Felix Johnny Thomasmathibalan committed Jan 30, 2025

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

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

Emil Ohlsson authored Jan 29, 2025 and Anton Bondarenko committed Jan 29, 2025

Update all version indications

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

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

Michael Kozlov authored Jan 28, 2025 and Felix Johnny Thomasmathibalan committed Jan 28, 2025

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

Signed-off-by: Anitha Raj <anitha.raj@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>

Anitha Raj authored Jan 27, 2025 and Jakub Sujak committed Jan 27, 2025

In kai_matmul_clamp_f32_qai8dxp1vlx8_qsi4cxp4vlx8_1vlx4vl_sme2_mopa:
 * Fix the offset calculation
 * Fix pointer increments in the matmul

Add new shapes to unit tests, to test n > 64

Resolves: #KLEIDIAI-405, #COMPMID-7918

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

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

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

Downstream projects might not build with same toolchain, which might
trigger warnings. Instead of treating these warnings as errors, using
this patch they are only warnings.

Also, add `-Werror` to the build flags to our testing pipeline, because
we do not want to accidentally introduce new warnings by mistake

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

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

Mention recent fixes for compiler warnings produced by enabling the `-Wcast-qual -Wmissing-prototypes -Wstrict-prototypes -Woverlength-strings` compiler options.

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

Approved-by: Emil Ohlsson <emil.ohlsson@arm.com>

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

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

Gian Marco Iodice authored Jan 16, 2025 and Emil Ohlsson committed Jan 16, 2025

- Previously, we were dividing m_idx/n_idx by m_step/n_step. However, since m_step/n_step can be different from mr/nr, we should divide by mr/nr instead

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

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

Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Approved-by: Emil Ohlsson <emil.ohlsson@arm.com>

Describe changes since 1.1.0

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

Reviewed-by: Emil Ohlsson <emil.ohlsson@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>

Michael Kozlov authored Dec 24, 2024 and Jakub Sujak committed Dec 24, 2024

- Add unit test

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

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Anitha Raj authored Dec 24, 2024 and Jakub Sujak committed Dec 24, 2024

* GEMM and GEMV micro-kernels to compute the matrix multiplication of dynamically quantized 8-bit integer (QAI8DX) LHS matrix and quantized 4-bit integer (QSI4CX) RHS matrix and the accumulation of the result into a single-precision (F32) output, optimized for SME2 technology.

Signed-off-by: Mohamad Najem <mohamad.najem@arm.com>

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

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

Signed-off-by: Thomas Bamelis <thomas.bamelis@arm.com>

Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Anton Bondarenko <anton.bondarenko@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Changes reported release version, and updates changelog since last
release

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

Emil Ohlsson authored Dec 13, 2024 and Felix Johnny Thomasmathibalan committed Dec 13, 2024

There are a few items missing in the changelog since last relase. This
commit updates the list with recent changes.

Signed-off-by: Emil Ohlsson <emil.ohlsson@arm.com>

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

Gian Marco Iodice authored Dec 13, 2024 and Anton Bondarenko committed Dec 13, 2024

- Optimize the generic RHS packing NxK. The performance improvement is
around ~1.5x

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

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

Anitha Raj authored Dec 13, 2024 and Jakub Sujak committed Dec 13, 2024

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

Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Emil Ohlsson <emil.ohlsson@arm.com>
Reviewed-by: alankelly <alankelly@google.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>

The `__fp16` type is known to be troublesome for certain build environments. Instead use type `float` in the interface, and in the implementation use types `float16_t` (an alias for `__fp16`) for casting or `uint16_t` for pointer arithmetic.

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

Gian Marco Iodice authored Dec 02, 2024 and Jakub Sujak committed Dec 02, 2024

- GEMM and GEMV Micro-kernels to compute the matrix multiplication of dynamically quantized symmetric signed 8-bit integer with per-block quantization (QSI8D32) LHS matrix and quantized symmetric 4-bit signed integer with per-block quantization (QSI4C32) RHS matrix and the accumulation of the result into a single-precision (F32) output, optimized for SME2 technology.

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

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

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

Jakub Sujak authored Nov 29, 2024 and Anton Bondarenko committed Nov 29, 2024

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

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

Gian Marco Iodice authored Nov 28, 2024 and Jakub Sujak committed Nov 28, 2024

- Add GeMM-like micro-kernels
- Add GeMV-like micro-kernels

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

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

Michael Kozlov authored Nov 27, 2024 and Felix Johnny Thomasmathibalan committed Nov 27, 2024

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

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

Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Anton Bondarenko <anton.bondarenko@arm.com>

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

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

Compute the Vector-Matrix multiply of F32 inputs to produce an F32 matrix, optimized using SME instructions.

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

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

Anitha Raj authored Oct 14, 2024 and Felix Johnny Thomasmathibalan committed Oct 14, 2024

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

Reviewed-by: Anitha Raj <anitha.raj@arm.com>
Approved-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

Jakub Sujak authored Sep 27, 2024 and Felix Johnny Thomasmathibalan committed Sep 27, 2024

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

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

Anitha Raj authored Sep 27, 2024 and Felix Johnny Thomasmathibalan committed Sep 27, 2024

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

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

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

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

Jakub Sujak authored Sep 25, 2024 and Felix Johnny Thomasmathibalan committed Sep 25, 2024

This reverts commit f4f59599

The existing FP32 GEMM micro-kernel (matmul_clamp_f32_f32_f32p8x1biasf32_6x8x4_neon_mla) has a dedicated path for M=1 (a "GEMV" operation).

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

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

Gian Marco Iodice authored Sep 12, 2024 and Viet-Hoa Do committed Sep 12, 2024

- The LHS matrix is Quantized (Q) Asymmetric (A) Signed 8-bit (I8) with per-row
(DX) quantization parameters
- The RHS matrix is quantized (Q) Symmetric (S) Signed 4-bit (I4) with
per-block quantization
- The destination is F32
- Implement micro-kernels to perform the matrix multiplication
- Implement a micro-kernel to pack the RHS matrix

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: Anitha Raj <anitha.raj@arm.com>
Reviewed-by: Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: Max Ren <maxren@meta.com>
Approved-by: Viet-Hoa Do <viet-hoa.do@arm.com>

Felix Johnny Thomasmathibalan authored Aug 30, 2024 and Jakub Sujak committed Aug 30, 2024

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

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

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

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

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

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