Implement Minimum Workload Size (MWS) in all CPPKernels used by small networks
* create get_mws method in ICPPKernel class that retuns default value for all kernels
* overwrite the default value for all the kernels used by small networks (according to banchmark case)
Resolves COMPMID-4648
Change-Id: I46d7cae61217213279d2ee740edc73f600b6d576
Signed-off-by: Dana Zlotnik <dana.zlotnik@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6412
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/cpu/kernels/internal/CpuPool2dAssemblyWrapperKernel.h b/src/cpu/kernels/internal/CpuPool2dAssemblyWrapperKernel.h
index ab3ed25..8625fd9 100644
--- a/src/cpu/kernels/internal/CpuPool2dAssemblyWrapperKernel.h
+++ b/src/cpu/kernels/internal/CpuPool2dAssemblyWrapperKernel.h
@@ -112,6 +112,15 @@
void create_arm_pooling_requant(const ITensorInfo *src, ITensorInfo *dst, const PoolingLayerInfo &info, const CPUInfo &cpu_info);
std::unique_ptr<arm_conv::pooling::IPoolingCommon> _kernel_asm{ nullptr };
+
+ /** Return minimum workload size of the relevant kernel
+ *
+ * @param[in] platform The CPU platform used to create the context.
+ * @param[in] thread_count Number of threads in the execution.
+ *
+ * @return[out] small_network_mws Minimum workload size for requsted configuration.
+ */
+ size_t get_mws(const CPUInfo &platform, size_t thread_count) const override;
};
} // namespace kernels
} // namespace cpu