From a668f9f8a4eab405df0fe8dd58e7d9425bcf9640 Mon Sep 17 00:00:00 2001 From: Jonathan Deakin Date: Wed, 24 Jan 2024 09:15:38 +0000 Subject: 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 Co-authored-by: David Mansell Change-Id: I58a3acf2c09289a303e52eea6b336a696a5bc8da Signed-off-by: Jonathan Deakin Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11022 Reviewed-by: Gunes Bayir Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins Benchmark: Arm Jenkins --- src/core/NEON/kernels/arm_gemm/quantized.hpp | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src/core/NEON/kernels/arm_gemm/quantized.hpp') diff --git a/src/core/NEON/kernels/arm_gemm/quantized.hpp b/src/core/NEON/kernels/arm_gemm/quantized.hpp index 31dd65b397..bc64fd967b 100644 --- a/src/core/NEON/kernels/arm_gemm/quantized.hpp +++ b/src/core/NEON/kernels/arm_gemm/quantized.hpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019, 2023 Arm Limited. + * Copyright (c) 2019, 2023-2024 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -45,4 +45,8 @@ template void row_sums_indirect(size_t num_strings, const unsigned int *string_lengths, IndirectInputArg A_arg, size_t M, int32_t *output_ptr, const Requantize32 *qp); +void dequantize_block_32(const DequantizeFloat &qp, unsigned int width, unsigned int height, + const int32_t* input, unsigned int in_stride, float *output, unsigned int out_stride, + const float *row_bias, bool not_first_pass, const Activation &act); + } // namespace arm_gemm -- cgit v1.2.1