commit c0f54434383f945d95f95549c1c4b0d5f5d2caff
author Giorgio Arena <giorgio.arena@arm.com> Fri Mar 16 14:02:34 2018 +0000
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:50:48 2018 +0000
tree c4dadc7d83fa9dccef8cd7e85b31223266946093
parent 3c520c5a6ca9352560828fdf389d31e38b85afeb

COMPMID-808 Add NHWC data format support for NEON direct convolution

Change-Id: I5d4cc3d5b0d25f3fe4ed998c0f15b1b8e260a43a
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/125697
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp

diff --git a/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp b/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
index 00776d7..445864c 100644
--- a/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
+++ b/src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp
@@ -35,18 +35,22 @@
 
 NEDirectConvolutionLayer::NEDirectConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager)
     : _memory_group(std::move(memory_manager)), _output_stage_kernel(), _conv_kernel(), _input_border_handler(), _activationlayer_function(), _accumulator(), _has_bias(false), _is_fixed_point(false),
-      _is_activationlayer_enabled(false)
+      _is_activationlayer_enabled(false), _dim_split(Window::DimZ)
 {
 }
 
 void NEDirectConvolutionLayer::configure(ITensor *input, const ITensor *weights, const ITensor *bias, ITensor *output, const PadStrideInfo &conv_info, const ActivationLayerInfo &act_info)
 {
+    ARM_COMPUTE_ERROR_ON(input->info()->data_layout() == DataLayout::UNKNOWN);
+
     // Free accumulator
     if(_accumulator.buffer() != nullptr)
     {
         _accumulator.allocator()->free();
     }
 
+    _dim_split = input->info()->data_layout() == DataLayout::NCHW ? Window::DimZ : Window::DimY;
+
     // Check if bias should be added in the convolution result
     _has_bias = (bias != nullptr);
 
@@ -124,7 +128,7 @@
 
     _memory_group.acquire();
 
-    NEScheduler::get().schedule(&_conv_kernel, Window::DimZ);
+    NEScheduler::get().schedule(&_conv_kernel, _dim_split);
     if(_has_bias || _is_fixed_point)
     {
         NEScheduler::get().schedule(&_output_stage_kernel, Window::DimY);