diff options
| author | Michalis Spyrou <michalis.spyrou@arm.com> | 2018-02-22 18:07:43 +0000 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:48:35 +0000 |
| commit | a9676118fd2a0e5bc916969af83ecee049bae76b (patch) | |
| tree | f67ff64d962a2f802700f6b26a9ab160c04c721d /src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp | |
| parent | 2bc74410251dcbaf17a7c5447317aa6d0171972a (diff) | |
| download | ComputeLibrary-a9676118fd2a0e5bc916969af83ecee049bae76b.tar.gz | |
COMPMID-886 Don't use LWS hints by default for GPU post Mali-G72
Change-Id: I64cb2d7f9513d69aebd9307a803b1b2c9c0e04c3
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/121929
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp')
| -rw-r--r-- | src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp index d5c93dd24a..3309775c36 100644 --- a/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp +++ b/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -52,7 +52,7 @@ std::pair<Status, Window> validate_and_configure_window(ITensorInfo *accum, ITen unsigned int &num_elems_processed_per_iteration) { // Select the vector size to use (8 for Bifrost; 16 for Midgard). - num_elems_processed_per_iteration = (gpu_target == GPUTarget::BIFROST) ? 8 : 16; + num_elems_processed_per_iteration = gpu_target_is_in(gpu_target, GPUTarget::G71, GPUTarget::G72) ? 8 : 16; // Configure kernel window Window win = calculate_max_window(*accum, Steps(num_elems_processed_per_iteration)); @@ -81,12 +81,12 @@ void CLGEMMMatrixAccumulateBiasesKernel::configure(ICLTensor *accum, const ICLTe _biases = biases; _accum = accum; - // Get the target architecture - GPUTarget arch_target = get_arch_from_target(get_target()); + // Get the target gpu + GPUTarget gpu_target = get_target(); unsigned int vector_size = 0; // Configure kernel window - auto win_config = validate_and_configure_window(accum->info(), biases->info(), arch_target, vector_size); + auto win_config = validate_and_configure_window(accum->info(), biases->info(), gpu_target, vector_size); ARM_COMPUTE_ERROR_THROW_ON(win_config.first); ICLKernel::configure(win_config.second); |