- Dec 17, 2024
-
-
Signed-off-by:
Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-by:
Emil Ohlsson <emil.ohlsson@arm.com> Approved-by:
-
- Dec 13, 2024
-
-
Emil Ohlsson authored
Changes reported release version, and updates changelog since last release Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com> -
There are a few items missing in the changelog since last relase. This commit updates the list with recent changes. Signed-off-by:
Felix Johnny Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
-
- Optimize the generic RHS packing NxK. The performance improvement is around ~1.5x Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: Anton Bondarenko <anton.bondarenko@arm.com> -
Jakub Sujak authoredSigned-off-by:
Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: -
Signed-off-by:
Anitha Raj <anitha.raj@arm.com> Reviewed-by:
alankelly <alankelly@google.com> Approved-by:
Jakub Sujak <jakub.sujak@arm.com> -
Original way of organizing matmul files was a bit complicated. A simpler way would reduce maintenance and simplify on how to add new microkernels Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
-
- Dec 11, 2024
-
-
Jakub Sujak authored
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 authored
Indicate transition to following Semantic Versioning for future releases. Signed-off-by:
-
- Dec 10, 2024
-
-
Jakub Sujak authored
Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
Approved-by: -
Anton Bondarenko authored
Version 8 of Bazel introduces explicit requirement to select WORKSPACE file as source of external dependencies. While it is quite new there is still a need to support version 6.5 where new flag cause build failure. Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Reviewed-by:
-
- Dec 09, 2024
-
-
Felix Johnny Thomasmathibalan authored
Signed-off-by:
-
- Add more shapes - Add pretty printer for test suite Signed-off-by:
Michael Kozlov <michael.kozlov@arm.com>
Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: -
Anton Bondarenko authored
Usage of compiled headers in a mix of C and C++ code could lead to compilation errors where C only flags are used. In this case headers should be interpreted together with source files using them. Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: -
Viet-Hoa Do authored
* Only the following parts are included in the MSVC build: - All scalar kernels. - Half-precision floating-point reference implementation. * Add tests. Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
-
- Dec 06, 2024
-
-
Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by:
-
- Dec 05, 2024
-
-
Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com> -
The different kernels does a lot of parameter checking, and one of the checks for the FP16 kernels is actually incorrect. This change addresses one error of this kind in the fp16 lhs packing kernel The same issue is also updated in the fp16 documentation Signed-off-by:
-
Emil Ohlsson authored
The current KAI_ASSERT() function will exit the program using `exit(EXIT_FAILURE)` which is unhelpful when running the program under a debugger, as this will not trap the execution. This change changes the call to `abort()` instead Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
-
- Dec 02, 2024
-
-
Signed-off-by:
-
- 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:
Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com> -
Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
Approved-by:
-
- Nov 29, 2024
-
-
Signed-off-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> -
Anton Bondarenko authored
Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: -
Signed-off-by:
Jens Elofsson <jens.elofsson@arm.com> Approved-by:
-
Suhail M authored
* Add the SME Int8 GEMM set of microkernels: - LHS packing kernel. - Non-transposed RHS packing kernel. - Main kernel. * Update the test framework to support static int8 GEMM. Resolves: KLEIDIAI-171, KLEIDIAI-235, KLEIDIAI-39 Signed-off-by:
Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com> Signed-off-by:
Mohammed Suhail Munshi <mohammedsuhail.munshi@arm.com> Reviewed-by:
-
Anton Bondarenko authored
Current solution for SME microkernels support is to use SME opcodes and SVE instruction in streaming mode. And a precondition for compiler to understand SVE instructions is compilation with -march=...+sve+sve2. However this allows compiler to generate own SVE instructions for normal C/C++ code. And might cause illegal instruction exception on CPUs where SME implemented w/o SVE. In this case we want to disable usage of compiler generated SVE instructions. Test no SVE instructions using FVP with disabled SVE support. Signed-off-by:
Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: -
Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com> -
RHS pack is required. LHS pack is not required Signed-off-by:
-
Jakub Sujak authored
Regenerate the SME2 GEMV micro-kernel assembly so that it is contained within the SMSTART/SMSTOP boundary, preventing illegal instruction faults when attempting to execute streaming SVE code on a system without SVE support. Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
Approved-by: -
This commit * Adds bf16 x bf16 = fp16 matmul microkernel with 8x12 output block size * Lhs/Rhs packing functions that packs and converts the inputs from fp16 to bf16 * Corresponding tests, and modifications to the testing framework, and reference implementation Signed-off-by:
Gunes Bayir <gunes.bayir@arm.com> Reviewed-by:
Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: -
Felix Johnny Thomasmathibalan authored
Affected micro kernel: FP16 GEMM, SME2 Signed-off-by:
-
- Nov 28, 2024
-
-
Jens Elofsson authored
Signed-off-by:
-
- Add GeMM-like micro-kernels - Add GeMV-like micro-kernels Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com> -
This change fixes a minor copy paste error in the kernel interfaces. Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
-
- Nov 27, 2024
-
-
Signed-off-by:
Michael Kozlov <michael.kozlov@arm.com>
Signed-off-by: Anton Bondarenko <anton.bondarenko@arm.com>
Approved-by: Jakub Sujak <jakub.sujak@arm.com>
Approved-by: Anton Bondarenko <anton.bondarenko@arm.com>
-
- Nov 21, 2024
-
-
Add fp16 kernels for LHS and RHS packing, and matmul. Also add related unit tests for said kernels, and extend unit Matmul tests to support calling fp16 kernels. Signed-off-by:
-
- Nov 20, 2024
-
-
Emil Ohlsson authored
KleidiAI is intended to target certain build environments, this means that KleidiAI should be buildable using CMake version 3.16 Signed-off-by:
Jakub Sujak <jakub.sujak@arm.com>
-
- Nov 19, 2024
-
-
* Round off the odd strides for the int4 RHS by padding with 0s Signed-off-by:
-
