diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2019-06-21 18:43:12 +0100 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2019-07-05 15:30:24 +0000 |
commit | 37d080f2f11cfd734104b76512e1fb191486216e (patch) | |
tree | d5df067c826aacc0676e7e9557a54b61a9a3b7eb /src/runtime/NEON/functions/assembly | |
parent | 11de30da8a9f79943255ddba7bb70a66b076673b (diff) | |
download | ComputeLibrary-37d080f2f11cfd734104b76512e1fb191486216e.tar.gz |
COMPMID-2378: Sanitize GEMM configuration for NEON
Change-Id: I7859b82b2059e14685f8792424648ac5eacd67f1
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1418
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/runtime/NEON/functions/assembly')
-rw-r--r-- | src/runtime/NEON/functions/assembly/NEGEMMInterleavedWrapper.cpp | 12 |
1 files changed, 6 insertions, 6 deletions
diff --git a/src/runtime/NEON/functions/assembly/NEGEMMInterleavedWrapper.cpp b/src/runtime/NEON/functions/assembly/NEGEMMInterleavedWrapper.cpp index 20aa1496b6..ac809fa142 100644 --- a/src/runtime/NEON/functions/assembly/NEGEMMInterleavedWrapper.cpp +++ b/src/runtime/NEON/functions/assembly/NEGEMMInterleavedWrapper.cpp @@ -339,19 +339,19 @@ void NEGEMMInterleavedWrapper::prepare() } } -void NEGEMMInterleavedWrapper::configure(const ITensor *a, const ITensor *b, ITensor *c, float alpha, float beta, bool pretranspose_b) +void NEGEMMInterleavedWrapper::configure(const ITensor *a, const ITensor *b, ITensor *c, float alpha, float beta, const GEMMInfo &gemm_info) { - _params = INEGEMMWrapperKernel::extract_parameters(a, b, c); + _params = INEGEMMWrapperKernel::extract_parameters(a, b, c, gemm_info); _a = a; _b = b; _c = c; - _pretranspose_b = pretranspose_b; + _pretranspose_b = gemm_info.pretranpose_B(); const DataType input_type = a->info()->data_type(); const CPUInfo &ci = NEScheduler::get().cpu_info(); const unsigned int num_threads = NEScheduler::get().num_threads(); - const arm_gemm::KernelDescription gemm_kernel_info = get_gemm_info(input_type, ci, num_threads, _params, alpha, beta, pretranspose_b); + const arm_gemm::KernelDescription gemm_kernel_info = get_gemm_info(input_type, ci, num_threads, _params, alpha, beta, _pretranspose_b); ARM_COMPUTE_ERROR_ON(gemm_kernel_info.method != arm_gemm::GemmMethod::GEMM_INTERLEAVED); // Forcing 128-byte alignment (required by 32-bit kernels) @@ -411,8 +411,8 @@ void NEGEMMInterleavedWrapper::configure(const ITensor *a, const ITensor *b, ITe _memory_group.manage(&_transformed_a); _memory_group.manage(&_tmp_c); - _transform_a = strategy->instantiate_transformA(_a, &_transformed_a, _block_walker, _params); - _matrix_multiply = strategy->instantiate_matrix_multiply(&_transformed_a, &_transformed_b, &_tmp_c, c, _block_walker, _block_sizes, _params, alpha, beta, pretranspose_b, num_threads); + _transform_a = strategy->instantiate_transformA(_a, &_transformed_a, _block_walker, _params, gemm_info); + _matrix_multiply = strategy->instantiate_matrix_multiply(&_transformed_a, &_transformed_b, &_tmp_c, c, _block_walker, _block_sizes, _params, alpha, beta, gemm_info, num_threads); ARM_COMPUTE_ERROR_ON(_transform_a == nullptr); ARM_COMPUTE_ERROR_ON(_matrix_multiply == nullptr); |