aboutsummaryrefslogtreecommitdiff
path: root/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp')
-rw-r--r--src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp13
1 files changed, 10 insertions, 3 deletions
diff --git a/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp b/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
index a71864c10c..934e38b054 100644
--- a/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
+++ b/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
@@ -357,9 +357,16 @@ const char *CpuDepthwiseConv2dAssemblyWrapperKernel::name() const
size_t CpuDepthwiseConv2dAssemblyWrapperKernel::get_mws(const CPUInfo &platform, size_t thread_count) const
{
- ARM_COMPUTE_UNUSED(platform, thread_count);
-
- return ICPPKernel::small_network_mws;
+ ARM_COMPUTE_UNUSED(thread_count);
+ // Tuning results that gave optimized results in performance investigation
+ if (platform.get_cpu_model() == CPUModel::A73 )
+ {
+ return 10240;
+ }
+ else
+ {
+ return 9216;
+ }
}
} // namespace kernels
} // namespace cpu