commit b5311a63d9bc1d8c9265aba5a9e031726cc512b3
author Gian Marco <gianmarco.iodice@arm.com> Wed Dec 13 12:48:03 2017 +0000
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:42:33 2018 +0000
tree 16e15017494450a0d53b740d3c8fd4656d5e578c
parent acacc324216f466a548a69c2d8cc6540dc2769be

COMPMID-556 - Enabled optimized path for Bifrost in CLGEMM

- Enabled optimized path for Bifrost when the input matrices have not
  been reshaped

Change-Id: I92c5591846ce7ea9116fecc434a0e859f3f2d579
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/113083
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>

src/runtime/CL/functions/CLGEMM.cpp

diff --git a/src/runtime/CL/functions/CLGEMM.cpp b/src/runtime/CL/functions/CLGEMM.cpp
index a81d113..ca0228f 100644
--- a/src/runtime/CL/functions/CLGEMM.cpp
+++ b/src/runtime/CL/functions/CLGEMM.cpp
@@ -65,6 +65,9 @@
     const ICLTensor *matrix_a = a;
     const ICLTensor *matrix_b = b;
 
+    // Set the target for the matrix multiply kernel
+    _mm_kernel.set_target(CLScheduler::get().target());
+
     if(_is_interleaved_transposed)
     {
         matrix_a = &_tmp_a;
@@ -95,9 +98,6 @@
 
         // Configure transpose kernel
         _transpose_kernel.configure(b, &_tmp_b);
-
-        // Configure matrix multiply kernel
-        _mm_kernel.set_target(CLScheduler::get().target());
     }
 
     _mm_kernel.configure(matrix_a, matrix_b, output, alpha, _is_interleaved_transposed);