commit 94916a5c06065bca0b232106bd4ae68f9986b7b0
author Teresa Charlin <teresa.charlinreyes@arm.com> Wed Oct 19 08:48:07 2022 +0100
committer TeresaARM <teresa.charlinreyes@arm.com> Mon Jan 09 15:09:46 2023 +0000
tree 1510f0e09f48305ce3a2c32e7adfddcc2da7b39a
parent c998108fbd4c134286b481768a873c54ae744a70

IVGCVSW-6493 Add GpuAcc Batch MatMul workload Fp32

* GpuAcc only supports up to 3D, so no 4D test have been added

Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: Ie926cd45c350be624cbdc6cb27c89d2d3f60884b

src/backends/cl/ClLayerSupport.cpp

diff --git a/src/backends/cl/ClLayerSupport.cpp b/src/backends/cl/ClLayerSupport.cpp
index a61a5bb..cb2d756 100644
--- a/src/backends/cl/ClLayerSupport.cpp
+++ b/src/backends/cl/ClLayerSupport.cpp
@@ -22,6 +22,7 @@
 #include "workloads/ClAdditionWorkload.hpp"
 #include "workloads/ClActivationWorkload.hpp"
 #include "workloads/ClArgMinMaxWorkload.hpp"
+#include "workloads/ClBatchMatMulWorkload.hpp"
 #include "workloads/ClBatchNormalizationFloatWorkload.hpp"
 #include "workloads/ClBatchToSpaceNdWorkload.hpp"
 #include "workloads/ClCastWorkload.hpp"
@@ -201,6 +202,12 @@
                                         infos[1],
                                         *(PolymorphicDowncast<const ArgMinMaxDescriptor*>(&descriptor)),
                                         reasonIfUnsupported);
+        case LayerType::BatchMatMul:
+            return IsBatchMatMulSupported(infos[0],
+                                          infos[1],
+                                          infos[2],
+                                          *(PolymorphicDowncast<const BatchMatMulDescriptor*>(&descriptor)),
+                                          reasonIfUnsupported);
         case LayerType::BatchNormalization:
             return IsBatchNormalizationSupported(infos[0],
                                                  infos[1],
@@ -640,6 +647,20 @@
                                    descriptor);
 }
 
+bool ClLayerSupport::IsBatchMatMulSupported(const TensorInfo& inputX,
+                                            const TensorInfo& inputY,
+                                            const TensorInfo& output,
+                                            const BatchMatMulDescriptor& descriptor,
+                                            Optional<std::string&> reasonIfUnsupported) const
+{
+    FORWARD_WORKLOAD_VALIDATE_FUNC(ClBatchMatMulValidate,
+                                   reasonIfUnsupported,
+                                   inputX,
+                                   inputY,
+                                   output,
+                                   descriptor);
+}
+
 bool ClLayerSupport::IsBatchNormalizationSupported(const TensorInfo& input,
                                                    const TensorInfo& output,
                                                    const TensorInfo& mean,