diff options
author | Gunes Bayir <gunes.bayir@arm.com> | 2023-09-14 09:13:49 +0100 |
---|---|---|
committer | Gunes Bayir <gunes.bayir@arm.com> | 2023-09-18 13:53:19 +0000 |
commit | 532ce2c84dd24cb0c5064a3d2e5c7b4094df0e01 (patch) | |
tree | 7bae37ad706ff0c3d62e0702061277b16cb7f141 /tests/datasets/system_tests/alexnet | |
parent | a116cd3676796412cd4d9318a6cc1c1eef4c093c (diff) | |
download | ComputeLibrary-532ce2c84dd24cb0c5064a3d2e5c7b4094df0e01.tar.gz |
Separate the output quantization calculation logic from matmul
This patch generalizes the suggested output quantization calculation to any operation that employs a dot product between two vectors, i.e.
c = sum_k(a_k * b_k) + d
It also consider and suggests min/max boundaries for random S32 bias generation, depending on the accumulation result.
MatMulKernelFixture is modified to use this interface.
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Change-Id: Ibb528261bb0310015967e11bd7ccd9ed9cff8479
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10312
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests/datasets/system_tests/alexnet')
0 files changed, 0 insertions, 0 deletions