commit c186b574e52b81c75e551cee46a6c4cc7d500c90
author Moritz Pflanzer <moritz.pflanzer@arm.com> Thu Sep 07 09:48:04 2017 +0100
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:35:24 2018 +0000
tree 682add7d1b705eb6d0b79b1f9a7a039dbaa4287b
parent 3e05e4e85912e745b8555102e1bcef13478d2ceb

COMPMID-481: Add thread info parameter

Change-Id: Iebb50a88d017445b6b37a86563ebd4abd86c5cf5
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/86788
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp

diff --git a/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp b/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp
index 2b7b391..a02ebf6 100644
--- a/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp
+++ b/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp
@@ -49,7 +49,7 @@
 
 namespace
 {
-void vector_matrix_multiply_f16(const ITensor *input0, const ITensor *input1, ITensor *output, const Window &window)
+void vector_matrix_multiply_f16(const ITensor *input0, const ITensor *input1, ITensor *output, const Window &window, const ThreadInfo &info)
 {
 #ifdef ARM_COMPUTE_ENABLE_FP16
     const auto width_matrix_b  = static_cast<int>(output->info()->dimension(0));
@@ -57,8 +57,8 @@
     const auto num_elems_vec_a = static_cast<int>(input0->info()->dimension(0));
 
     // The implementation computes 16 elements per iteration
-    const int window_start_x = 16 * window.thread_id();
-    const int window_step_x  = 16 * window.num_threads();
+    const int window_start_x = 16 * info.thread_id;
+    const int window_step_x  = 16 * info.num_threads;
     // Make sure (window_end_x - window_start_x) is a multiple of window_step_x
     const int window_end_x = ceil_to_multiple(width_matrix_b - window_start_x, window_step_x) + window_start_x;
 
@@ -169,15 +169,15 @@
 #endif /* ARM_COMPUTE_ENABLE_FP16 */
 }
 
-void vector_matrix_multiply_f32(const ITensor *input0, const ITensor *input1, ITensor *output, const Window &window)
+void vector_matrix_multiply_f32(const ITensor *input0, const ITensor *input1, ITensor *output, const Window &window, const ThreadInfo &info)
 {
     const auto width_matrix_b  = static_cast<int>(output->info()->dimension(0));
     const auto in_b_stride     = static_cast<int>(input1->info()->strides_in_bytes()[1] / data_size_from_type(input1->info()->data_type()));
     const auto num_elems_vec_a = static_cast<int>(input0->info()->dimension(0));
 
     // The implementation computes 16 elements per iteration
-    const int window_start_x = 16 * window.thread_id();
-    const int window_step_x  = 16 * window.num_threads();
+    const int window_start_x = 16 * info.thread_id;
+    const int window_step_x  = 16 * info.num_threads;
     // Make sure (window_end_x - window_start_x) is a multiple of window_step_x
     const int window_end_x = ceil_to_multiple(width_matrix_b - window_start_x, window_step_x) + window_start_x;
 
@@ -337,7 +337,7 @@
     INEKernel::configure(win);
 }
 
-void NELocallyConnectedMatrixMultiplyKernel::run(const Window &window)
+void NELocallyConnectedMatrixMultiplyKernel::run(const Window &window, const ThreadInfo &info)
 {
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
@@ -346,12 +346,12 @@
     {
         case DataType::F16:
         {
-            vector_matrix_multiply_f16(_input0, _input1, _output, window);
+            vector_matrix_multiply_f16(_input0, _input1, _output, window, info);
             break;
         }
         case DataType::F32:
         {
-            vector_matrix_multiply_f32(_input0, _input1, _output, window);
+            vector_matrix_multiply_f32(_input0, _input1, _output, window, info);
             break;
         }
         default: