commit 226e4b92b191491ffa57ede66eba1d5d6fcf3b76
author Giorgio Arena <giorgio.arena@arm.com> Thu Aug 23 12:00:02 2018 +0100
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:54:54 2018 +0000
tree 334705a1e743e3465400208d582cf0b25bf950fa
parent 35aea3776449557c438e264bae7af5b1fe0e5ff6

COMPMID-1470 Add auto-init of the output in NECol2im

The output of NECol2Im is already auto-initialized.
This patch is about calling ShapeCalculator instead of computing the shape inside the kernel, adding validate_and_configure_window, and standardize the way convolved dims are passed (now NEON uses Size2D, while CL passes a pair of uint values: using Size2D for both implementations)

Change-Id: I795696e1b6532f57847c3186c1b532c09f5a25da
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/145345
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp

diff --git a/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp b/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp
index 782fe71..c9daea4 100644
--- a/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp
+++ b/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp
@@ -285,7 +285,7 @@
         if(input->info()->data_layout() == DataLayout::NCHW)
         {
             // Configure and tune Col2Im
-            _col2im_kernel.configure(_is_quantized ? gemm_output_staged_to_use : gemm_output_to_use, output, std::make_pair(conv_w, conv_h), num_groups);
+            _col2im_kernel.configure(_is_quantized ? gemm_output_staged_to_use : gemm_output_to_use, output, Size2D(conv_w, conv_h), num_groups);
             CLScheduler::get().tune_kernel_static(_col2im_kernel);
         }
         else
@@ -443,7 +443,7 @@
         {
             ARM_COMPUTE_RETURN_ON_ERROR(CLCol2ImKernel::validate(is_quantized ? gemm_output_staged_to_use : gemm_output_to_use,
                                                                  output,
-                                                                 std::make_pair(conv_w, conv_h), num_groups));
+                                                                 Size2D(conv_w, conv_h), num_groups));
         }
     }
 

src/runtime/CL/functions/CLLocallyConnectedLayer.cpp

diff --git a/src/runtime/CL/functions/CLLocallyConnectedLayer.cpp b/src/runtime/CL/functions/CLLocallyConnectedLayer.cpp
index 40bf032..5c6bef9 100644
--- a/src/runtime/CL/functions/CLLocallyConnectedLayer.cpp
+++ b/src/runtime/CL/functions/CLLocallyConnectedLayer.cpp
@@ -122,7 +122,7 @@
     ARM_COMPUTE_RETURN_ON_ERROR(CLIm2ColKernel::validate(input, &input_im2col_reshaped_info, Size2D(kernel_width, kernel_height), conv_info, has_bias));
     ARM_COMPUTE_RETURN_ON_ERROR(CLWeightsReshapeKernel::validate(weights, biases, &weights_reshaped_info));
     ARM_COMPUTE_RETURN_ON_ERROR(CLLocallyConnectedMatrixMultiplyKernel::validate(&input_im2col_reshaped_info, &weights_reshaped_info, &gemm_output_info));
-    ARM_COMPUTE_RETURN_ON_ERROR(CLCol2ImKernel::validate(&gemm_output_info, output, std::make_pair(conv_w, conv_h)));
+    ARM_COMPUTE_RETURN_ON_ERROR(CLCol2ImKernel::validate(&gemm_output_info, output, Size2D(conv_w, conv_h)));
 
     return Status{};
 }
@@ -163,7 +163,7 @@
     _input_im2col_kernel.configure(input, &_input_im2col_reshaped, Size2D(kernel_width, kernel_height), conv_info, _has_bias);
     _weights_reshape_kernel.configure(weights, biases, &_weights_reshaped);
     _mm_kernel.configure(&_input_im2col_reshaped, &_weights_reshaped, &_gemm_output);
-    _output_col2im_kernel.configure(&_gemm_output, output, std::make_pair(conv_w, conv_h));
+    _output_col2im_kernel.configure(&_gemm_output, output, Size2D(conv_w, conv_h));
 
     // Allocate intermediate tensors
     _input_im2col_reshaped.allocator()->allocate();

src/runtime/CL/tuners/BifrostTuner.cpp

diff --git a/src/runtime/CL/tuners/BifrostTuner.cpp b/src/runtime/CL/tuners/BifrostTuner.cpp
index 2d52f33..59d73b4 100644
--- a/src/runtime/CL/tuners/BifrostTuner.cpp
+++ b/src/runtime/CL/tuners/BifrostTuner.cpp
@@ -134,7 +134,7 @@
     // via exhaustive autotuning over 30 representative tensor shapes.
     if(gpu_target_is_in(gpu_target, GPUTarget::G71, GPUTarget::G72, GPUTarget::G51, GPUTarget::G51BIG, GPUTarget::G51LIT, GPUTarget::G76))
     {
-        if((k._convolved_dims.first == 7) || (k._convolved_dims.first == 14))
+        if((k._convolved_dims.width == 7) || (k._convolved_dims.width == 14))
         {
             lws_hint = cl::NDRange(1, 7, 1);
         }