diff options
| author | Gunes Bayir <gunes.bayir@arm.com> | 2024-02-12 21:32:51 +0000 |
|---|---|---|
| committer | Gunes Bayir <gunes.bayir@arm.com> | 2024-02-21 10:36:22 +0000 |
| commit | ef637398a8c2060e15de438020c53331da8bd6dd (patch) | |
| tree | b1a1738736c9b6b49e76767e44bf4b77bf732876 /src/core/NEON/kernels/arm_gemm/transform.cpp | |
| parent | 0a48c4c83b598991b4d4235f870c24d9e6634b20 (diff) | |
| download | ComputeLibrary-ef637398a8c2060e15de438020c53331da8bd6dd.tar.gz | |
Integrate new pretranspose_b_array with extra fused transpose of B
This patch fuses the transposition taking place in Acl with the transformations done in arm_gemm (called pretranspose_b_array) if the underlying kernel and transform supports it. This should improve start-up time (as it's for constant Rhs matrices) and memory footprint. The transformations in arm_gemm are kernel specific. The Rhs matrix is transformed into certain layouts to improve the performance.
Resolves: COMPMID-6595
Change-Id: Id2932dd966e59f903c279417bebcea83d9a42464
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11144
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/transform.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/transform.cpp | 11 |
1 files changed, 9 insertions, 2 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/transform.cpp b/src/core/NEON/kernels/arm_gemm/transform.cpp index 5aa62f0fe4..45e4f0e1de 100644 --- a/src/core/NEON/kernels/arm_gemm/transform.cpp +++ b/src/core/NEON/kernels/arm_gemm/transform.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021-2023 Arm Limited. + * Copyright (c) 2021-2024 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -134,7 +134,14 @@ template void Transform<8, 1, true, VLType::None>(float *, const __fp16 *, int, #endif // defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) #ifdef ARM_COMPUTE_ENABLE_BF16 template void Transform<8, 1, true, VLType::None>(float *, const bfloat16 *, int, int, int, int, int); -#endif +#endif // ARM_COMPUTE_ENABLE_BF16 #endif // AArch32 +#if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) +template void Transform<12, 1, false, VLType::None>(float *, const __fp16 *, int, int, int, int, int); +#endif // defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) +#ifdef ARM_COMPUTE_ENABLE_BF16 +template void Transform<12, 1, false, VLType::None>(float *, const bfloat16 *, int, int, int, int, int); +#endif // ARM_COMPUTE_ENABLE_BF16 + } // namespace arm_gemm |