commit e6630e4063fc3aa4312a2c8d094318b09ad2c3f5
author Isabella Gottardi <isabella.gottardi@arm.com> Thu Jan 18 15:50:39 2018 +0000
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:45:00 2018 +0000
tree 39ae08686fc3201fd094e3f84b8dd9abd5bf07ea
parent b99d57df435ec1f2a775b3b06a44a68a2aac8df9

COMPMID-790 - NEON: Add QASYMM8 support to Convolution

Change-Id: Iec82a91ad351cfe8d07d0976a24bd42f4703177a
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/116833
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

src/core/NEON/kernels/NEGEMMLowpOffsetContributionKernel.cpp

diff --git a/src/core/NEON/kernels/NEGEMMLowpOffsetContributionKernel.cpp b/src/core/NEON/kernels/NEGEMMLowpOffsetContributionKernel.cpp
index 3d41548..ee334df 100644
--- a/src/core/NEON/kernels/NEGEMMLowpOffsetContributionKernel.cpp
+++ b/src/core/NEON/kernels/NEGEMMLowpOffsetContributionKernel.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -143,13 +143,10 @@
     // If a_offset == 0, vector_sum_col can be a nullptr
     if(a_offset != 0)
     {
-        TensorShape vector_sum_col_shape = vector_sum_col->info()->tensor_shape(); // NOLINT
-        vector_sum_col_shape.collapse(1);
-
         // Check if vector_sum_col_shape should be slidden or not
         // Don't slide vector_sum_col_shape along the y dimension if vector_sum_col_shape has just 1 dimension and vector_sum_row_shape more than 1
         // This scenario can happen when the the matrix multiplication is used to perform a convolution operation
-        _slide_vector_sum_col = vector_sum_col_shape[1] != 1;
+        _slide_vector_sum_col = vector_sum_col->info()->tensor_shape().num_dimensions() > 1;
     }
 
     // Configure kernel window
@@ -201,7 +198,7 @@
         Iterator vector_sum_row(_vector_sum_row, win_vector_sum_row);
         Iterator mm_result(_mm_result, window);
 
-        execute_window_loop(window, [&](const Coordinates & id)
+        execute_window_loop(collapsed_window, [&](const Coordinates & id)
         {
             // Compute the leftover term due to a_offset.
             int32x4x4_t a_offset_term_s32 =