commit 0021d750d66d199c411df00cdd8308c325f1fef3
author Diego Lopez Recas <Diego.LopezRecas@arm.com> Mon Dec 18 14:42:56 2017 +0000
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:45:00 2018 +0000
tree b96e618977442a8aab335c136d369a958998d416
parent 5b6904b8d9cb5e8a343cde96fd5a8701f44dff90

IVGCVSW-863 Broadcast support in CL/NEON Arithmetic Add

Also, added instrumentation to support generic tensor broadcasting for
NEON and CL backends.

Change-Id: I1bc5747a286e1a4b464c209067581e103d473b9a
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/114201
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

src/core/CL/kernels/CLPermuteKernel.cpp

diff --git a/src/core/CL/kernels/CLPermuteKernel.cpp b/src/core/CL/kernels/CLPermuteKernel.cpp
index 132de60..1f36445 100644
--- a/src/core/CL/kernels/CLPermuteKernel.cpp
+++ b/src/core/CL/kernels/CLPermuteKernel.cpp
@@ -106,10 +106,10 @@
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(ICLKernel::window(), window);
 
-    Window slice_in = window.first_slice_window_4D();
-    Window slice_out(slice_in);
+    Window slice_in = window.first_slice_window_4D().collapse(ICLKernel::window(), 2, 4);
 
     // Setup output slice
+    Window slice_out(slice_in);
     slice_out.set(Window::DimX, Window::Dimension(0, 0, 0));
     slice_out.set(Window::DimY, Window::Dimension(0, 0, 0));
     slice_out.set(Window::DimZ, Window::Dimension(0, 0, 0));
@@ -117,12 +117,10 @@
 
     do
     {
-        auto         collapsed_slice_in  = slice_in.collapse(ICLKernel::window(), 2);
-        auto         collapsed_slice_out = slice_out.collapse(ICLKernel::window(), 2);
-        unsigned int idx                 = 0;
-        add_4D_tensor_argument(idx, _input, collapsed_slice_in);
-        add_4D_tensor_argument(idx, _output, collapsed_slice_out);
-        enqueue(queue, *this, collapsed_slice_in);
+        unsigned int idx = 0;
+        add_4D_tensor_argument(idx, _input, slice_in);
+        add_4D_tensor_argument(idx, _output, slice_out);
+        enqueue(queue, *this, slice_in);
     }
     while(window.slide_window_slice_4D(slice_in) && window.slide_window_slice_4D(slice_out));
 }