commit 181e65145d153210ec5587a42d2938e27e1d5b01
author Pablo Tello <pablo.tello@arm.com> Wed Nov 15 13:28:27 2017 +0000
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:35:24 2018 +0000
tree 70115705382ec4997d2f1ff44a33224f50ace38a
parent bc8fb0634339dfd662f4b2d825f74615b8a69bac

COMPMID-675: NEGEMMLowp Assembly Integration

Added support for S8 input in NEGEMMLowp Matrix Multiply Kernel.
Added a new function to run assembly kernels such that A*B=C (no offsets involved)
Added new tests for the assembly gemmlowp kernels (no offsets)
Integrated the assembly kernel for the A57

Change-Id: Ib3e39c1f3f7f1baa0d39be69485f61cd18e3c9b3
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/95864
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

tests/validation/NEON/GEMMLowp.cpp

diff --git a/tests/validation/NEON/GEMMLowp.cpp b/tests/validation/NEON/GEMMLowp.cpp
index 4407eff..ba91ced 100644
--- a/tests/validation/NEON/GEMMLowp.cpp
+++ b/tests/validation/NEON/GEMMLowp.cpp
@@ -22,7 +22,9 @@
  * SOFTWARE.
  */
 #include "arm_compute/core/NEON/kernels/NEGEMMInterleaveBlockedKernel.h"
+#include "arm_compute/core/NEON/kernels/arm64/NEGEMMLowpAArch64Kernel.h"
 #include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/NEON/functions/NEGEMMLowpAssemblyMatrixMultiplyCore.h"
 #include "arm_compute/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.h"
 #include "arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h"
 #include "arm_compute/runtime/Tensor.h"
@@ -38,6 +40,7 @@
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
 #include "tests/validation/fixtures/GEMMInterleaveBlockedFixture.h"
+#include "tests/validation/fixtures/GEMMLowpAssemblyFixture.h"
 #include "tests/validation/fixtures/GEMMLowpFixture.h"
 
 namespace arm_compute
@@ -48,11 +51,21 @@
 {
 namespace
 {
-const auto data_int_blk    = framework::dataset::make("M", 8, 12) * framework::dataset::make("N", 8, 12) * framework::dataset::make("by", 8, 13) * framework::dataset::make("block", 4, 9);
-const auto data_int_blk_tr = framework::dataset::make("M", 8, 17) * framework::dataset::make("N", 8, 14) * framework::dataset::make("by", 12) * framework::dataset::make("block", 4);
+const auto data_int_blk         = framework::dataset::make("M", 8, 12) * framework::dataset::make("N", 8, 12) * framework::dataset::make("by", 8, 13) * framework::dataset::make("block", 4, 9);
+const auto data_int_blk_tr      = framework::dataset::make("M", 8, 17) * framework::dataset::make("N", 8, 14) * framework::dataset::make("by", 12) * framework::dataset::make("block", 4);
+const auto data_matrix_multiply = framework::dataset::make("M", 12, 20) * framework::dataset::make("N", 12, 20) * framework::dataset::make("K", 16);
 } // namespace
 
 TEST_SUITE(NEON)
+TEST_SUITE(ASSEMBLY_MATRIX_MULTIPLY)
+using NEGEMMAssemblyFixture = GEMMLowpAssemblyFixture<Tensor, Accessor, NEGEMMLowpAssemblyMatrixMultiplyCore>;
+FIXTURE_DATA_TEST_CASE(RunSmall, NEGEMMAssemblyFixture, framework::DatasetMode::PRECOMMIT, data_matrix_multiply)
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END()
+
 TEST_SUITE(GEMMLowp)
 
 TEST_SUITE(INTERLEAVE_BLOCKED)