diff options
| author | Teresa Charlin <teresa.charlinreyes@arm.com> | 2022-10-13 15:47:08 +0100 |
|---|---|---|
| committer | Teresa Charlin <teresa.charlinreyes@arm.com> | 2022-10-28 09:15:30 +0100 |
| commit | 0f86ecfce593a302ebd2baf8b70c9f6f50616f81 (patch) | |
| tree | bf312b7f8780ccecef54de782ff965c72cf68778 /src/backends/neon/NeonLayerSupport.cpp | |
| parent | 20508425bb92623862e4b3f0da1adf46fc541f27 (diff) | |
| download | armnn-0f86ecfce593a302ebd2baf8b70c9f6f50616f81.tar.gz | |
IVGCVSW-6494 Add CpuAcc Batch MatMul Workload Fp32
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: I2def6995f81d33e68f1ea45d8d19a1e6294049b1
Diffstat (limited to 'src/backends/neon/NeonLayerSupport.cpp')
| -rw-r--r-- | src/backends/neon/NeonLayerSupport.cpp | 21 |
1 files changed, 21 insertions, 0 deletions
diff --git a/src/backends/neon/NeonLayerSupport.cpp b/src/backends/neon/NeonLayerSupport.cpp index cf541f491b..7f311d8684 100644 --- a/src/backends/neon/NeonLayerSupport.cpp +++ b/src/backends/neon/NeonLayerSupport.cpp @@ -24,6 +24,7 @@ #include "workloads/NeonAdditionWorkload.hpp" #include "workloads/NeonActivationWorkload.hpp" #include "workloads/NeonArgMinMaxWorkload.hpp" +#include "workloads/NeonBatchMatMulWorkload.hpp" #include "workloads/NeonBatchNormalizationWorkload.hpp" #include "workloads/NeonBatchToSpaceNdWorkload.hpp" #include "workloads/NeonCastWorkload.hpp" @@ -171,6 +172,12 @@ bool NeonLayerSupport::IsLayerSupported(const LayerType& type, 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], @@ -627,6 +634,20 @@ bool NeonLayerSupport::IsArgMinMaxSupported(const TensorInfo& input, descriptor); } +bool NeonLayerSupport::IsBatchMatMulSupported(const TensorInfo& inputX, + const TensorInfo& inputY, + const TensorInfo& output, + const BatchMatMulDescriptor& descriptor, + Optional<std::string&> reasonIfUnsupported) const +{ + FORWARD_WORKLOAD_VALIDATE_FUNC(NeonBatchMatMulValidate, + reasonIfUnsupported, + inputX, + inputY, + output, + descriptor); +} + bool NeonLayerSupport::IsBatchNormalizationSupported(const TensorInfo& input, const TensorInfo& output, const TensorInfo& mean, |