Skip to content

perf: Add FP16 GEMM MMUL Reshaped Only Rhs Kernel#1181

Merged
ArmDude merged 1 commit into
mainfrom
pr/g1-kernel
Sep 23, 2025
Merged

perf: Add FP16 GEMM MMUL Reshaped Only Rhs Kernel#1181
ArmDude merged 1 commit into
mainfrom
pr/g1-kernel

Conversation

@ArmDude
Copy link
Copy Markdown
Contributor

@ArmDude ArmDude commented Sep 16, 2025

This patch introduces a GEMM routine that is optimized for Arm(R) Mali(TM)-G1

Resolves: [COMPMID-8311], [COMPMID-8312]

Change-Id: I84e685f0314da9af1c3fbb50d83e68b355727770

@Dongsung-arm Dongsung-arm self-requested a review September 16, 2025 13:26
@Dongsung-arm
Copy link
Copy Markdown

Looks fine to me

Dongsung-arm
Dongsung-arm previously approved these changes Sep 16, 2025
Comment thread src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsMMULKernel.cpp Outdated
Comment thread src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsMMULKernel.cpp Outdated
Comment thread src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsMMULKernel.cpp Outdated
Comment thread src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsMMULKernel.cpp Outdated
Comment thread tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRhsMMUL.cpp Outdated
Comment thread tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRhsMMUL.cpp Outdated
Comment thread tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRhsMMUL.cpp Outdated
Comment thread tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRhsMMUL.cpp Outdated
Comment thread tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRhsMMUL.cpp
Comment thread arm_compute/core/CL/cl_definitions.h
@ArmDude ArmDude force-pushed the pr/g1-kernel branch 3 times, most recently from 5e9d919 to c65ce4f Compare September 22, 2025 10:35
@gunes-arm
Copy link
Copy Markdown
Contributor

I suggest to change the title to: "Add FP16 GEMM MMUL Reshaped Only Rhs Kernel"

Comment thread src/core/CL/cl_kernels/common/gemm_reshaped_only_rhs_mmul.cl Outdated
Comment thread src/core/CL/cl_kernels/common/gemm_reshaped_only_rhs_mmul.cl Outdated
Comment thread src/core/CL/cl_kernels/common/gemm_reshaped_only_rhs_mmul.cl
Comment thread src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsMMULKernel.cpp Outdated
Comment thread src/gpu/cl/kernels/gemm/ClGemmHelpers.h
Comment thread tests/validation/CL/GEMMMatrixMultiplyReshapedOnlyRhsMMUL.cpp Outdated
@ArmDude ArmDude changed the title perf: Add FP16 GEMM MMUL Reshaped Only Rhs Support perf: Add FP16 GEMM MMUL Reshaped Only Rhs Kernel Sep 22, 2025
@ArmDude ArmDude force-pushed the pr/g1-kernel branch 2 times, most recently from 108b2aa to 961a15c Compare September 22, 2025 15:21
Comment thread src/core/CL/cl_kernels/common/gemm_reshaped_only_rhs_mmul.cl Outdated
This patch introduces a GEMM routine that is optimized for Arm(R) Mali(TM)-G1

Resolves: [COMPMID-8311], [COMPMID-8312]

Signed-off-by: Omar Al Khatib <omar.alkhatib@arm.com>
Change-Id: I84e685f0314da9af1c3fbb50d83e68b355727770
@ArmDude ArmDude merged commit 531a496 into main Sep 23, 2025
2 checks passed
@ArmDude ArmDude deleted the pr/g1-kernel branch September 23, 2025 08:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants