aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL/kernels/CLPoolingLayerKernel.cpp
diff options
context:
space:
mode:
authorMichalis Spyrou <michalis.spyrou@arm.com>2018-02-22 18:07:43 +0000
committerAnthony Barbier <anthony.barbier@arm.com>2018-11-02 16:48:35 +0000
commita9676118fd2a0e5bc916969af83ecee049bae76b (patch)
treef67ff64d962a2f802700f6b26a9ab160c04c721d /src/core/CL/kernels/CLPoolingLayerKernel.cpp
parent2bc74410251dcbaf17a7c5447317aa6d0171972a (diff)
downloadComputeLibrary-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/CLPoolingLayerKernel.cpp')
-rw-r--r--src/core/CL/kernels/CLPoolingLayerKernel.cpp4
1 files changed, 2 insertions, 2 deletions
diff --git a/src/core/CL/kernels/CLPoolingLayerKernel.cpp b/src/core/CL/kernels/CLPoolingLayerKernel.cpp
index b3034e10cc..d7b86e78f6 100644
--- a/src/core/CL/kernels/CLPoolingLayerKernel.cpp
+++ b/src/core/CL/kernels/CLPoolingLayerKernel.cpp
@@ -190,7 +190,7 @@ void CLPoolingLayerKernel::configure(const ICLTensor *input, ICLTensor *output,
_output = output;
_pool_info = pool_info;
- const GPUTarget gpu_target = get_arch_from_target(get_target());
+ const GPUTarget gpu_target = get_target();
const DataType data_type = input->info()->data_type();
// Set build options
@@ -240,7 +240,7 @@ void CLPoolingLayerKernel::configure(const ICLTensor *input, ICLTensor *output,
// On Bifrost, this works for up to 35x35xC filters, for which the pooling_layer_3_optimized
// kernel is launched with gws=(9, 33, C). In any case, the hint will be ignored if it is
// invalid (e.g. exceeds the maximum workgroup size that the kernel can be launched with).
- if(gpu_target == GPUTarget::BIFROST)
+ if(gpu_target_is_in(gpu_target, GPUTarget::G71, GPUTarget::G72))
{
cl::NDRange gws = ICLKernel::gws_from_window(std::get<1>(win_config));
_lws_hint = cl::NDRange(gws[0], gws[1], 1);