diff options
author | Michele Di Giorgio <michele.digiorgio@arm.com> | 2019-10-23 10:53:10 +0100 |
---|---|---|
committer | Michele Di Giorgio <michele.digiorgio@arm.com> | 2019-11-28 10:02:15 +0000 |
commit | 14cbfb2921990d8bf125231e350e2ac8dcd95a8b (patch) | |
tree | 9bec073d72c44c480c8807601889481d9b89ee7e /src/runtime | |
parent | ed7b27dd7cbdae57b880029840ad0235523848e0 (diff) | |
download | ComputeLibrary-14cbfb2921990d8bf125231e350e2ac8dcd95a8b.tar.gz |
COMPMID-2609: Enable quantization with multiplier greater than 1 on OpenCL
Change-Id: I050f1f84e214e61f7cbb0197a672b68a4940edae
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/2158
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Manuel Bottini <manuel.bottini@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Diffstat (limited to 'src/runtime')
-rw-r--r-- | src/runtime/CL/functions/CLConvolutionLayer.cpp | 4 | ||||
-rw-r--r-- | src/runtime/CL/functions/CLFullyConnectedLayer.cpp | 2 | ||||
-rw-r--r-- | src/runtime/CL/functions/CLGEMMDeconvolutionLayer.cpp | 17 |
3 files changed, 17 insertions, 6 deletions
diff --git a/src/runtime/CL/functions/CLConvolutionLayer.cpp b/src/runtime/CL/functions/CLConvolutionLayer.cpp index d794cde1f4..c271f502e9 100644 --- a/src/runtime/CL/functions/CLConvolutionLayer.cpp +++ b/src/runtime/CL/functions/CLConvolutionLayer.cpp @@ -34,7 +34,8 @@ #include <memory> #include <tuple> -using namespace arm_compute; +namespace arm_compute +{ using namespace arm_compute::misc::shape_calculator; CLConvolutionLayer::CLConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager) @@ -216,3 +217,4 @@ void CLConvolutionLayer::prepare() { _function->prepare(); } +} // namespace arm_compute diff --git a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp index a8167ce8f7..ad0714ed15 100644 --- a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp +++ b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp @@ -60,7 +60,7 @@ Status construct_gemmlowp_output_stage(const ITensorInfo &input, const ITensorIn const float multiplier = (iq_info.scale * wq_info.scale) / output_quant_info.scale; int output_multiplier = 0; int output_shift = 0; - ARM_COMPUTE_RETURN_ON_ERROR(quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift)); + ARM_COMPUTE_RETURN_ON_ERROR(quantization::calculate_quantized_multiplier(multiplier, &output_multiplier, &output_shift)); // Set the GEMMLowp output stage info gemmlowp_output_stage.gemmlowp_offset = output_quant_info.offset; diff --git a/src/runtime/CL/functions/CLGEMMDeconvolutionLayer.cpp b/src/runtime/CL/functions/CLGEMMDeconvolutionLayer.cpp index 4671be5b61..604147a37a 100644 --- a/src/runtime/CL/functions/CLGEMMDeconvolutionLayer.cpp +++ b/src/runtime/CL/functions/CLGEMMDeconvolutionLayer.cpp @@ -152,9 +152,9 @@ Status CLGEMMDeconvolutionLayer::validate(const ITensorInfo *input, const ITenso } const PadStrideInfo stride_info(deconv_info.stride().first, deconv_info.stride().second); - auto out_dims = deconvolution_output_dimensions(input->dimension(idx_w), input->dimension(idx_h), weights->dimension(idx_w), weights->dimension(idx_h), stride_info); - const TensorShape deconv_shape = misc::shape_calculator::compute_deconvolution_output_shape(out_dims, *input, *weights); - TensorInfo col2im_output_info = gemm_output_info.clone()->set_tensor_shape(deconv_shape).set_is_resizable(true); + auto out_dims = deconvolution_output_dimensions(input->dimension(idx_w), input->dimension(idx_h), weights->dimension(idx_w), weights->dimension(idx_h), stride_info); + const TensorShape deconv_shape = misc::shape_calculator::compute_deconvolution_output_shape(out_dims, *input, *weights); + TensorInfo col2im_output_info = gemm_output_info.clone()->set_tensor_shape(deconv_shape).set_is_resizable(true); if(padded_input && is_quantized) { @@ -173,6 +173,15 @@ Status CLGEMMDeconvolutionLayer::validate(const ITensorInfo *input, const ITenso else if(is_quantized) { ARM_COMPUTE_RETURN_ON_ERROR(CLDeconvolutionReshapeOutputKernel::validate(&gemm_output_info, bias, &col2im_output_info, input, weights, deconv_info)); + + const UniformQuantizationInfo iq_info = input->quantization_info().uniform(); + const UniformQuantizationInfo wq_info = weights->quantization_info().uniform(); + const UniformQuantizationInfo oq_info = output->quantization_info().uniform(); + + float multiplier = iq_info.scale * wq_info.scale / oq_info.scale; + int output_multiplier(0); + int output_shift(0); + ARM_COMPUTE_RETURN_ON_ERROR(quantization::calculate_quantized_multiplier(multiplier, &output_multiplier, &output_shift)); ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint::validate(&col2im_output_info, nullptr, output)); } else @@ -284,7 +293,7 @@ void CLGEMMDeconvolutionLayer::configure(const ICLTensor *input, const ICLTensor float multiplier = iq_info.scale * wq_info.scale / oq_info.scale; int output_multiplier(0); int output_shift(0); - quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); + quantization::calculate_quantized_multiplier(multiplier, &output_multiplier, &output_shift); _gemmlowp_output_stage.configure(&_gemmlowp_final, nullptr, output_stage_output, output_multiplier, output_shift, oq_info.offset); _gemmlowp_final.allocator()->allocate(); } |