From f1f1f87132690a8061801ef1a4638d637c780df7 Mon Sep 17 00:00:00 2001 From: Radu Salavat Date: Tue, 27 Feb 2024 18:32:26 +0000 Subject: Add in place summation to CPU GEMM kernels Instead of dispatching the sum postop for GEMM kernels to a separate kernel + add, that requires an extra destination sized allocation, plus 3 extra load/stores per element, just do it in the GEMM kernel. Resolves: ONCPUML-1442 Signed-off-by: Radu Salavat Co-authored-by: Milos Puzovic Change-Id: I7a1f2da3300875fa1ac88b705a34390969518077 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11298 Reviewed-by: Gunes Bayir Tested-by: Arm Jenkins Comments-Addressed: Arm Jenkins Benchmark: Arm Jenkins --- docs/user_guide/release_version_and_change_log.dox | 1 + 1 file changed, 1 insertion(+) (limited to 'docs') diff --git a/docs/user_guide/release_version_and_change_log.dox b/docs/user_guide/release_version_and_change_log.dox index aa27c2b44c..b8910c9237 100644 --- a/docs/user_guide/release_version_and_change_log.dox +++ b/docs/user_guide/release_version_and_change_log.dox @@ -48,6 +48,7 @@ v24.04 Public major release - Add support for SoftMax in SME2 for FP32. - Performance optimizations: - Optimize @ref NESoftmaxLayer for axis != 0 by natively supporting higher axes up to axis 3. + - Add support for in place accumulation to CPU GEMM kernels. v24.02.1 Public patch release - Fix performance regression in fixed-format kernels -- cgit v1.2.1