diff options
author | Jonathan Deakin <jonathan.deakin@arm.com> | 2024-01-24 09:15:38 +0000 |
---|---|---|
committer | Radu Salavat <radu.salavat@arm.com> | 2024-04-15 13:52:31 +0000 |
commit | a668f9f8a4eab405df0fe8dd58e7d9425bcf9640 (patch) | |
tree | db16e6af9289897557a58755b88d2c337dcb8650 /src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp | |
parent | 34bdffb288d6367cb6dca652ebed60c450854039 (diff) | |
download | ComputeLibrary-a668f9f8a4eab405df0fe8dd58e7d9425bcf9640.tar.gz |
Add s8f32 kernels and dynamic QuantizationInfo
- Add support for QASYMM_SIGNED*QASYMM8_SIGNED->F32 in
CpuGemmLowpMatrixMultiplyCore
- Add s8f32 kernel using existing s8->s32 kernels with a new
DequantizeFloat OutputStage, the structure is similar to Requantize32
but the opposite way around.
- Add SME s8f32 kernels with integrated support for DequantizeFloat.
- Add scale to CpuGemmLowpOffsetContributionKernel.
- Add virtual dequantize scale to gemm_common, only implemented for
gemm_interleaved.
- Update year to 2024 in generate_build_files.
- Add dynamic flag to QuantizationInfo which signals to operators that
it can change after configuration
- Add support for dynamic quantization in NEGEMMLowpMatrixMultiplyCore
- Add dynamic quantization fixture by extending
GEMMLowpGenericMatrixMultiplyCoreValidationFixture
- Add GEMMLowpDequantizedMatrixMultiplyValidationFixture
- Store k (number of cols of A) rather than k_offset in the offset
contribution kernels so that we can recompute it when the other
offsets change
relates to: ONCPUML-1444 MLINFSW-439
Co-authored-by: Milos Puzovic <Milos.Puzovic@arm.com>
Co-authored-by: David Mansell <David.Mansell@arm.com>
Change-Id: I58a3acf2c09289a303e52eea6b336a696a5bc8da
Signed-off-by: Jonathan Deakin <jonathan.deakin@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11022
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp')
-rw-r--r-- | src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp | 21 |
1 files changed, 16 insertions, 5 deletions
diff --git a/src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp b/src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp index d008842398..3c113f2828 100644 --- a/src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp +++ b/src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2021, 2023 Arm Limited. + * Copyright (c) 2019-2021, 2023-2024 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -919,7 +919,7 @@ void CpuGemmLowpOffsetContributionOutputStageKernel::configure(const ITensorInfo _a_offset = a_offset; _b_offset = b_offset; - _k_offset = a_offset * b_offset * k; + _k = k; _output_stage = output_stage; // If a_offset == 0, vector_sum_col can be a nullptr @@ -958,6 +958,16 @@ Status CpuGemmLowpOffsetContributionOutputStageKernel::validate(const ITensorInf return Status{}; } +void CpuGemmLowpOffsetContributionOutputStageKernel::set_a_offset(int32_t a_offset) +{ + _a_offset = a_offset; +} + +void CpuGemmLowpOffsetContributionOutputStageKernel::set_b_offset(int32_t b_offset) +{ + _b_offset = b_offset; +} + void CpuGemmLowpOffsetContributionOutputStageKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) @@ -993,10 +1003,11 @@ void CpuGemmLowpOffsetContributionOutputStageKernel::run_op(ITensorPack &te // Check if symmetric per-channel execution const bool is_symm = _output_stage.is_quantized_per_channel; + auto k_offset = _a_offset * _b_offset * _k; if (is_symm) { run_offset_contribution_output_stage_symm(window, mm_result, vector_sum_col, vector_sum_row, bias, dst, - _a_offset, _b_offset, _k_offset, _is_vector_sum_col_batched, + _a_offset, _b_offset, k_offset, _is_vector_sum_col_batched, _output_stage, reinterpret_as_3d, is_bounded_relu, is_fixed_point); } else @@ -1004,13 +1015,13 @@ void CpuGemmLowpOffsetContributionOutputStageKernel::run_op(ITensorPack &te if (is_signed) { run_offset_contribution_output_stage<int8_t>( - window, mm_result, vector_sum_col, vector_sum_row, bias, dst, _a_offset, _b_offset, _k_offset, + window, mm_result, vector_sum_col, vector_sum_row, bias, dst, _a_offset, _b_offset, k_offset, _is_vector_sum_col_batched, _output_stage, reinterpret_as_3d, is_bounded_relu, is_fixed_point); } else { run_offset_contribution_output_stage<uint8_t>( - window, mm_result, vector_sum_col, vector_sum_row, bias, dst, _a_offset, _b_offset, _k_offset, + window, mm_result, vector_sum_col, vector_sum_row, bias, dst, _a_offset, _b_offset, k_offset, _is_vector_sum_col_batched, _output_stage, reinterpret_as_3d, is_bounded_relu, is_fixed_point); } } |