diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-07-23 20:38:47 +0100 |
|---|---|---|
| committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-07-26 08:41:55 +0000 |
| commit | e92c23eb7a91ddd12feeb40cd8bc5d766c6fe5c3 (patch) | |
| tree | 9d9ed4ef73afb88d5951126675ad1bdd405b3469 /src/runtime/gpu/cl/utils/ClAuxTensorHandler.h | |
| parent | 7075fe2c5ee6f7cfe7cfd9454d905235e70b9ac4 (diff) | |
| download | ComputeLibrary-e92c23eb7a91ddd12feeb40cd8bc5d766c6fe5c3.tar.gz | |
Fix allocation of prepare tensor on ClWinogradConv2d
Preparation tensor was allocated during execution. Avoid if GEMM used by
the Winograd function is reshaped.
Resolves: COMPMID-4640
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Id5b7e6bbfe9f69661baa9d42698cdc2cc1b422b2
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5991
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/runtime/gpu/cl/utils/ClAuxTensorHandler.h')
| -rw-r--r-- | src/runtime/gpu/cl/utils/ClAuxTensorHandler.h | 10 |
1 files changed, 8 insertions, 2 deletions
diff --git a/src/runtime/gpu/cl/utils/ClAuxTensorHandler.h b/src/runtime/gpu/cl/utils/ClAuxTensorHandler.h index 1cf717cf6f..af383489a1 100644 --- a/src/runtime/gpu/cl/utils/ClAuxTensorHandler.h +++ b/src/runtime/gpu/cl/utils/ClAuxTensorHandler.h @@ -28,6 +28,7 @@ #include "arm_compute/core/TensorInfo.h" #include "arm_compute/runtime/CL/CLTensor.h" +#include "src/common/utils/Log.h" #include "support/Cast.h" namespace arm_compute @@ -38,7 +39,7 @@ namespace opencl class CLAuxTensorHandler { public: - CLAuxTensorHandler(int slot_id, TensorInfo &info, ITensorPack &pack, bool pack_inject = false) + CLAuxTensorHandler(int slot_id, TensorInfo &info, ITensorPack &pack, bool pack_inject = false, bool bypass_alloc = false) : _tensor() { if(info.total_size() == 0) @@ -50,7 +51,12 @@ public: ICLTensor *packed_tensor = utils::cast::polymorphic_downcast<ICLTensor *>(pack.get_tensor(slot_id)); if((packed_tensor == nullptr) || (info.total_size() > packed_tensor->info()->total_size())) { - _tensor.allocator()->allocate(); + if(!bypass_alloc) + { + _tensor.allocator()->allocate(); + ARM_COMPUTE_LOG_INFO_WITH_FUNCNAME_ACL("Allocating auxiliary tensor"); + } + if(pack_inject) { pack.add_tensor(slot_id, &_tensor); |