diff options
| author | Ramy Elgammal <ramelg01@e129512.arm.com> | 2023-01-11 18:48:04 +0000 |
|---|---|---|
| committer | Ramy Elgammal <ramy.elgammal@arm.com> | 2023-01-25 10:24:23 +0000 |
| commit | 002e6530f6218b00a28aef9be8b21efb08cf3602 (patch) | |
| tree | f3e2f9d064b985ffe283512825b34cdd59f29f50 /arm_compute/dynamic_fusion/runtime | |
| parent | cc2877368d5e15d9ea89d31c84ec651fc0fffd13 (diff) | |
| download | ComputeLibrary-002e6530f6218b00a28aef9be8b21efb08cf3602.tar.gz | |
Implement dynamic fusion softmax operator
- Return aux tensorInfo by get_aux_tensors() at runtime to init the aux
tensor with the right size.
- Keep softmax unfusable for this commit
- Hence, added Tensor3D to template writer arguments declaration, for sake of
keeping dynamic fusion softmax componenets' kernels matching their cl
counterparts.
Resolves: COMPMID-5523
Change-Id: I667f39545db925f667036ef448302c79a0330373
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/483924
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8986
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute/dynamic_fusion/runtime')
| -rw-r--r-- | arm_compute/dynamic_fusion/runtime/gpu/cl/ClWorkloadRuntime.h | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/arm_compute/dynamic_fusion/runtime/gpu/cl/ClWorkloadRuntime.h b/arm_compute/dynamic_fusion/runtime/gpu/cl/ClWorkloadRuntime.h index 326880f721..b5af589cd2 100644 --- a/arm_compute/dynamic_fusion/runtime/gpu/cl/ClWorkloadRuntime.h +++ b/arm_compute/dynamic_fusion/runtime/gpu/cl/ClWorkloadRuntime.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2022 Arm Limited. + * Copyright (c) 2022-2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -24,8 +24,8 @@ #ifndef ARM_COMPUTE_DYNAMIC_FUSION_RUNTIME_GPU_CL_CLWORKLOADRUNTIME #define ARM_COMPUTE_DYNAMIC_FUSION_RUNTIME_GPU_CL_CLWORKLOADRUNTIME +#include "arm_compute/core/TensorInfo.h" #include "arm_compute/dynamic_fusion/sketch/MemoryDescriptor.h" - #include <map> #include <memory> @@ -63,7 +63,7 @@ public: Status run(const std::vector<CLTensor *> &tensors); /** Get auxiliary tensors of the workload and their memory requirement */ - std::vector<std::pair<CLTensor *, AuxMemoryInfo>> get_auxiliary_tensors(); + std::vector<std::tuple<CLTensor *, TensorInfo, AuxMemoryInfo>> get_auxiliary_tensors(); private: /** Enqueue prepare workload |