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>

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>