commit d39e2b1e0be12420b1e00279ebee0c34bae3dd8c
author Gian Marco Iodice <gianmarco.iodice@arm.com> Mon Aug 06 14:31:15 2018 +0100
committer Anthony Barbier <anthony.barbier@arm.com> Fri Nov 02 16:54:54 2018 +0000
tree dbcb65b0e9d7f884fc340c9aea7355ba0b4cfcfe
parent 20394d526820ca97df4c0db91ec2571b98280d6d

COMPMID-1188 - Fixed performance degradation with GEMM3D

The previous implementation of GEMM3D degradated the performance when the
input had to be reinterpreted as 3D. However if both input and output have to be
reinterpreted as 3D, we can skip the offset calculation for that specific case
and run the multi GEMM approach

Change-Id: I0d5d48add2c6ccdebfbb268ea199dd181101f3aa
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/142872
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

tests/datasets/LargeGEMMDataset.h

diff --git a/tests/datasets/LargeGEMMDataset.h b/tests/datasets/LargeGEMMDataset.h
index 9c45cb0..bbf362c 100644
--- a/tests/datasets/LargeGEMMDataset.h
+++ b/tests/datasets/LargeGEMMDataset.h
@@ -71,11 +71,11 @@
     {
         add_config(TensorShape(923U, 143U, 3U), TensorShape(871U, 923U), TensorShape(871U, 143U, 3U), TensorShape(871U, 143U, 3U), 1.0f, 0.0f);
         add_config(TensorShape(681U, 205U, 5U), TensorShape(213U, 681U), TensorShape(213U, 205U, 5U), TensorShape(213U, 205U, 5U), 1.0f, 0.0f);
-        add_config(TensorShape(364U, 605U, 5U), TensorShape(96U, 364U), TensorShape(96U, 605U, 5U), TensorShape(96U, 605U, 5U), 1.0f, 0.0f);
+        add_config(TensorShape(364U, 605U, 5U), TensorShape(96U, 364U), TensorShape(96U, 605U, 5U), TensorShape(96U, 605U, 5U), 0.2f, 1.2f);
         add_config(TensorShape(1201U, 243U, 3U), TensorShape(128U, 1201U), TensorShape(128U, 243U, 3U), TensorShape(128U, 243U, 3U), 1.0f, 0.0f);
-        add_config(TensorShape(2305U, 13U, 13U), TensorShape(384U, 2305U), TensorShape(384U, 13U, 13U), TensorShape(384U, 13U, 13U), 1.0f, 0.0f);
-        add_config(TensorShape(1729U, 85U, 2U, 2U), TensorShape(192U, 1729U, 2U), TensorShape(192U, 85U, 2U, 2U), TensorShape(192U, 85U, 2U, 2U), 1.0f, 0.0f);
-        add_config(TensorShape(1729U, 17U, 10U, 3U), TensorShape(128U, 1729U, 3U), TensorShape(128U, 17U, 10U, 3U), TensorShape(128U, 17U, 10U, 3U), 1.0f, 0.0f);
+        add_config(TensorShape(2305U, 13U, 13U), TensorShape(384U, 2305U), TensorShape(384U, 13U, 13U), TensorShape(384U, 13U, 13U), 0.4f, 0.7f);
+        add_config(TensorShape(1729U, 85U, 2U, 2U), TensorShape(192U, 1729U), TensorShape(192U, 85U, 2U, 2U), TensorShape(192U, 85U, 2U, 2U), 1.0f, 0.0f);
+        add_config(TensorShape(1729U, 17U, 10U, 3U), TensorShape(128U, 1729U), TensorShape(128U, 17U, 10U, 3U), TensorShape(128U, 17U, 10U, 3U), 1.0f, 0.3f);
     }
 };
 } // namespace datasets