diff options
| author | Pablo Tello <pablo.tello@arm.com> | 2019-02-05 16:16:19 +0000 |
|---|---|---|
| committer | Pablo Marquez <pablo.tello@arm.com> | 2019-02-08 11:15:30 +0000 |
| commit | 54e98d98fbe082b265b2c4a384eabe0144866bcc (patch) | |
| tree | f0367171d8dfcad282523efceb709ec523ef9edd /src/runtime/NEON | |
| parent | 03bb550b72ebd107a6cbd994008cdcc00597d822 (diff) | |
| download | ComputeLibrary-54e98d98fbe082b265b2c4a384eabe0144866bcc.tar.gz | |
COMPMID-1918: Different qinfos support in NEConcatLayer.
Added support in NEDepthConcatenateLayerKernel and NEWidthConcatenateLayer for
different quantization arguments both for the input and output.
If input's quantization infos are not homogeneous the input values are requantized using
the output's quantization info.
Change-Id: I2daa638361947eb3ec848d5425d0a5bbfea1936d
Reviewed-on: https://review.mlplatform.org/627
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Isabella Gottardi <isabella.gottardi@arm.com>
Diffstat (limited to 'src/runtime/NEON')
| -rw-r--r-- | src/runtime/NEON/functions/NEWidthConcatenateLayer.cpp | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/src/runtime/NEON/functions/NEWidthConcatenateLayer.cpp b/src/runtime/NEON/functions/NEWidthConcatenateLayer.cpp index 097605c062..7e435c34b1 100644 --- a/src/runtime/NEON/functions/NEWidthConcatenateLayer.cpp +++ b/src/runtime/NEON/functions/NEWidthConcatenateLayer.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018 ARM Limited. + * Copyright (c) 2018-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -66,7 +66,7 @@ void NEWidthConcatenateLayer::configure(std::vector<ITensor *> inputs_vector, IT _num_inputs = inputs_vector.size(); std::vector<ITensorInfo *> inputs_vector_info; - for(unsigned int i = 0; i < _num_inputs; i++) + for(unsigned int i = 0; i < _num_inputs; ++i) { inputs_vector_info.emplace_back(inputs_vector.at(i)->info()); } @@ -80,7 +80,7 @@ void NEWidthConcatenateLayer::configure(std::vector<ITensor *> inputs_vector, IT _concat_kernels_vector = arm_compute::support::cpp14::make_unique<NEWidthConcatenateLayerKernel[]>(_num_inputs); - for(unsigned int i = 0; i < _num_inputs; i++) + for(unsigned int i = 0; i < _num_inputs; ++i) { _concat_kernels_vector[i].configure(inputs_vector.at(i), width_offset, output); width_offset += inputs_vector.at(i)->info()->dimension(0); @@ -89,7 +89,7 @@ void NEWidthConcatenateLayer::configure(std::vector<ITensor *> inputs_vector, IT void NEWidthConcatenateLayer::run() { - for(unsigned i = 0; i < _num_inputs; i++) + for(unsigned i = 0; i < _num_inputs; ++i) { NEScheduler::get().schedule(_concat_kernels_vector.get() + i, Window::DimY); } |