commit 4406fd6cc4abded564d3791324e1f48bdfd34273
author Frank Lei <frank.lei@arm.com> Thu Feb 01 14:47:14 2018 +0800
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:47:40 2018 +0000
tree 22fe402fe9ac7ca338df49e9eccd6eb1587ae875
parent 898d399a0f62c15612a52df4bff5018e783214e4

APPBROWSER-391: Fix GLES COMPUTE alignment issues

APPBROWSER-402: Performance optimization for squeezenet/xray model

Change-Id: If31b186b99a6d6087164019fe94d3ac9279e3204
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/119526
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

src/core/GLES_COMPUTE/kernels/GCScaleKernel.cpp

diff --git a/src/core/GLES_COMPUTE/kernels/GCScaleKernel.cpp b/src/core/GLES_COMPUTE/kernels/GCScaleKernel.cpp
index f307cfb..46d7ff9 100644
--- a/src/core/GLES_COMPUTE/kernels/GCScaleKernel.cpp
+++ b/src/core/GLES_COMPUTE/kernels/GCScaleKernel.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016, 2018 ARM Limited.
+ * Copyright (c) 2016-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -128,9 +128,34 @@
 
     IGCKernel::configure(win);
 
-    unsigned int idx = 2 * num_arguments_per_2D_tensor(); //Skip the tensor parameters
+    unsigned int idx = 2 * num_arguments_per_3D_tensor(); //Skip the tensor parameters
     _kernel.set_argument<float>(idx++, static_cast<float>(input->info()->dimension(0)));
     _kernel.set_argument<float>(idx++, static_cast<float>(input->info()->dimension(1)));
     _kernel.set_argument<float>(idx++, wr);
     _kernel.set_argument<float>(idx++, hr);
 }
+
+void GCScaleKernel::run(const Window &window)
+{
+    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
+
+    _kernel.use();
+
+    _output->set_needs_shifting(true);
+
+    Window slice    = window.first_slice_window_3D();
+    Window slice_in = window.first_slice_window_3D();
+
+    slice.shift(Window::DimX, -(_output->info()->padding()).left);
+
+    do
+    {
+        unsigned int idx = 0;
+        add_3D_tensor_argument(idx, _input, 1, slice_in);
+        add_3D_tensor_argument(idx, _output, 2, slice);
+        _kernel.update_shader_params();
+        enqueue(*this, slice);
+    }
+    while(window.slide_window_slice_3D(slice) && window.slide_window_slice_3D(slice_in));
+}