diff options
author | Gunes Bayir <gunes.bayir@arm.com> | 2024-01-17 16:07:03 +0000 |
---|---|---|
committer | Viet-Hoa Do <viet-hoa.do@arm.com> | 2024-02-01 16:00:34 +0000 |
commit | 2b9fa593a0a172bf36a02b5cdb840c6b9b361d7c (patch) | |
tree | a4e2d5ce46443a79a0778e4960462ce3edf106ec /compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h | |
parent | 7ab7fca87cca8775f82b0e9efec6a40975910c17 (diff) | |
download | ComputeLibrary-2b9fa593a0a172bf36a02b5cdb840c6b9b361d7c.tar.gz |
Use the stable CKW API in the GPU dynamic fusion backend
- Refactor all kernels to work with the CKW stable API
- Add support for sub-tile in the op_load/op_store CKW operator
- Fix mismatch in resize
- Add comments in all kernels written with CKW to help developers
understand the structure of the code
- Add texture image support in depthwise convolution written with CKW
- Add support for different block sizes in depthwise convolution
- Remove the use of the dynamic fusion helper functions.
- Add support for floor in the op_unary() of CKW
Resolves: COMPMID-6708, COMPMID-6743, COMPMID-6530
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Signed-off-by: Jakub Sujak <jakub.sujak@arm.com>
Change-Id: I8104ce4d04a3138a1aeb0b84940e1f1c89e76069
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10914
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h')
-rw-r--r-- | compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h | 24 |
1 files changed, 15 insertions, 9 deletions
diff --git a/compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h b/compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h index f46fee9750..a5b679ac03 100644 --- a/compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h +++ b/compute_kernel_writer/src/cl/helpers/ICLMemoryOpHelper.h @@ -28,6 +28,7 @@ #include "ckw/TensorSampler.h" #include "src/Tensor3dMapper.h" +#include "src/TileView.h" #include <cstdint> #include <memory> @@ -55,18 +56,24 @@ public: * @param[in] tensor @ref ckw::ITensor object to perform the memory operation on * @param[in] sampler @ref ckw::TensorSampler object that tells how to sample a tensor * @param[in] op The memory operation to be done (e.g. Load/Store) + * @param[in] dst The tile to perform the memory operation on */ - ICLMemoryOpHelper(CLKernelWriter *writer, ITensor *tensor, TensorSampler *sampler, MemoryOperation op) - : _writer(writer), _tensor(tensor), _sampler(sampler), _op(op) + ICLMemoryOpHelper(CLKernelWriter *writer, + ITensor *tensor, + TensorSampler *sampler, + MemoryOperation op, + const TileView<CLTile> &dst) + : _writer(writer), _tensor(tensor), _sampler(sampler), _op(op), _dst(dst) { - _mapper = std::make_unique<Tensor3dMapper>(tensor, sampler->format()); + _mapper = std::make_unique<Tensor3dMapper>(tensor, sampler->format()); + _ls_width_full = _dst.width(); } /** Copy constructor */ - ICLMemoryOpHelper(const ICLMemoryOpHelper &) = default; + ICLMemoryOpHelper(const ICLMemoryOpHelper &) = delete; /** Assignment operator overload */ - ICLMemoryOpHelper &operator=(const ICLMemoryOpHelper &) = default; + ICLMemoryOpHelper &operator=(const ICLMemoryOpHelper &) = delete; /** Destructor */ virtual ~ICLMemoryOpHelper() = default; @@ -75,12 +82,11 @@ public: * the batch offset as a tile object, and initializes the code inside * the writer object. * - * @param[in] dst tile object to perform the memory operation on * @param[in] x tile object that describes the x-coordinate of the tensor involved * @param[in] z tile object that describes the z-coordinate of the tensor involved * @param[in] b tile object that describes the batch offset of the tensor involved */ - virtual void initialize(const CLTile *dst, const CLTile *x, const CLTile *z, const CLTile *b) = 0; + virtual void initialize(const CLTile *x, const CLTile *z, const CLTile *b) = 0; /** Method that writes the actual code to the writer that performs the mentioned memory * operation on the tile initialized. It writes the code for a specific row given in the @@ -104,7 +110,7 @@ protected: TensorSampler *_sampler{nullptr}; MemoryOperation _op; std::unique_ptr<Tensor3dMapper> _mapper{nullptr}; - const CLTile *_dst{nullptr}; + TileView<CLTile> _dst{}; int32_t _ls_width_full{0}; std::string _coord_x{}; std::string _coord_z{}; @@ -112,4 +118,4 @@ protected: }; } // namespace ckw -#endif /* CKW_SRC_CL_HELPERS_ICLMEMORYOPHELPER_H */ +#endif // CKW_SRC_CL_HELPERS_ICLMEMORYOPHELPER_H |