From c577f2c6a3b4ddb6ba87a882723c53a248afbeba Mon Sep 17 00:00:00 2001 From: telsoa01 Date: Fri, 31 Aug 2018 09:22:23 +0100 Subject: Release 18.08 --- .../ClMultiplicationFloat32Workload.cpp | 26 +++++++++++++++++++--- 1 file changed, 23 insertions(+), 3 deletions(-) (limited to 'src/armnn/backends/ClWorkloads/ClMultiplicationFloat32Workload.cpp') diff --git a/src/armnn/backends/ClWorkloads/ClMultiplicationFloat32Workload.cpp b/src/armnn/backends/ClWorkloads/ClMultiplicationFloat32Workload.cpp index 405d109aa1..7aa33146f3 100644 --- a/src/armnn/backends/ClWorkloads/ClMultiplicationFloat32Workload.cpp +++ b/src/armnn/backends/ClWorkloads/ClMultiplicationFloat32Workload.cpp @@ -10,9 +10,29 @@ namespace armnn { +arm_compute::Status ClMultiplicationWorkloadValidate(const TensorInfo& input0, + const TensorInfo& input1, + const TensorInfo& output) +{ + const arm_compute::TensorInfo aclInput1 = armcomputetensorutils::BuildArmComputeTensorInfo(input0); + const arm_compute::TensorInfo aclInput2 = armcomputetensorutils::BuildArmComputeTensorInfo(input1); + const arm_compute::TensorInfo aclOutput = armcomputetensorutils::BuildArmComputeTensorInfo(output); + + // At the time of writing, configure() will fail if a rounding policy other than TO_ZERO is supplied to it, + // when providing a scale of 1.0 for F32 tensors, even though the provided rounding policy appears to be + // ignored for F32 tensors. + return arm_compute::CLPixelWiseMultiplication::validate(&aclInput1, + &aclInput2, + &aclOutput, + 1.0f, + arm_compute::ConvertPolicy::SATURATE, + arm_compute::RoundingPolicy::TO_ZERO); +} + + ClMultiplicationFloat32Workload::ClMultiplicationFloat32Workload(const MultiplicationQueueDescriptor& descriptor, const WorkloadInfo& info) - : Float32Workload(descriptor, info) + : FloatWorkload(descriptor, info) { m_Data.ValidateInputsOutputs("ClMultiplicationFloat32Workload", 2, 1); @@ -30,9 +50,9 @@ ClMultiplicationFloat32Workload::ClMultiplicationFloat32Workload(const Multiplic void ClMultiplicationFloat32Workload::Execute() const { - ARMNN_SCOPED_PROFILING_EVENT(Compute::GpuAcc, "ClMultiplicationFloat32Workload_Execute"); + ARMNN_SCOPED_PROFILING_EVENT_CL("ClMultiplicationFloat32Workload_Execute"); - // Execute the layer + // Executes the layer. m_PixelWiseMultiplication.run(); } -- cgit v1.2.1