diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-04-22 21:13:21 +0100 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-05-18 14:48:39 +0000 |
commit | 856f66e6c61b77d03f754cd0fa8439891f0e4aca (patch) | |
tree | f9379cd0853ac407109e54c3d53b385ceee066c2 /arm_compute/runtime/NEON | |
parent | 37f4b2ef1ea225a90ccb563fcb2c08f8fb0fb5d5 (diff) | |
download | ComputeLibrary-856f66e6c61b77d03f754cd0fa8439891f0e4aca.tar.gz |
Port CLGEMM to memory injecting interface
Moves the following kernels:
- CLGEMMMatrixMultiplyKernel
- CLGEMMMatrixMultiplyNativeKernel
- CLGEMMMatrixMultipluReshapedKernel
- CLGEMMMatrixMultiplyReshapedOnlyRHSKernel
Moves the following functions
- CLGEMM
Introduces facilities to easy handling of auxiliary temporary buffers
under then new run interface. Such are:
- CLAuxTensorHandler: That allows wrapping of workspace buffers memory
to CLBuffer objects
- Ability to inject TensorInfo to allocator without transferring
ownership. This reduce the copy overhead if needed.
Resolves: COMPMID-4188
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I7055435d831b05b749b26302082e4ac45f26dfb0
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5498
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute/runtime/NEON')
-rw-r--r-- | arm_compute/runtime/NEON/functions/NESlice.h | 70 | ||||
-rw-r--r-- | arm_compute/runtime/NEON/functions/NEStridedSlice.h | 70 |
2 files changed, 78 insertions, 62 deletions
diff --git a/arm_compute/runtime/NEON/functions/NESlice.h b/arm_compute/runtime/NEON/functions/NESlice.h index 214ffa512c..550bfd2188 100644 --- a/arm_compute/runtime/NEON/functions/NESlice.h +++ b/arm_compute/runtime/NEON/functions/NESlice.h @@ -32,25 +32,44 @@ namespace arm_compute // Forward Declarations class ITensor; -namespace experimental -{ /** Basic function to perform tensor slicing */ -class NESlice : public INEOperator +class NESlice : public IFunction { public: + /** Default Constructor */ + NESlice(); + /** Default Destructor */ + ~NESlice(); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + NESlice(const NESlice &) = delete; + /** Default move constructor */ + NESlice(NESlice &&); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + NESlice &operator=(const NESlice &) = delete; + /** Default move assignment operator */ + NESlice &operator=(NESlice &&); + /** Configure kernel * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:------|:------| + * |All |All | + * * @note Supported tensor rank: up to 4 * @note Start indices must be non-negative. 0 <= starts[i] * @note End coordinates can be negative, which represents the number of elements before the end of that dimension. * @note End indices are not inclusive unless negative. * - * @param[in] input Source tensor info. Data type supported: All - * @param[out] output Destination tensor info. Data type supported: Same as @p input + * @param[in] input Source tensor. Data type supported: All + * @param[out] output Destination tensor. Data type supported: Same as @p input * @param[in] starts The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input). * @param[in] ends The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input). */ - void configure(const ITensorInfo *input, ITensorInfo *output, const Coordinates &starts, const Coordinates &ends); + void configure(const ITensor *input, ITensor *output, const Coordinates &starts, const Coordinates &ends); /** Static function to check if given info will lead to a valid configuration of @ref NESlice * @@ -67,26 +86,21 @@ public: * @return A status */ static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Coordinates &starts, const Coordinates &ends); + + // Inherited methods overridden: + void run() override; + +private: + struct Impl; + std::unique_ptr<Impl> _impl; }; -} // namespace experimental +namespace experimental +{ /** Basic function to perform tensor slicing */ -class NESlice : public IFunction +class NESlice : public INEOperator { public: - /** Default Constructor */ - NESlice(); - /** Default Destructor */ - ~NESlice(); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - NESlice(const NESlice &) = delete; - /** Default move constructor */ - NESlice(NESlice &&); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - NESlice &operator=(const NESlice &) = delete; - /** Default move assignment operator */ - NESlice &operator=(NESlice &&); - /** Configure kernel * * Valid data layouts: @@ -102,12 +116,12 @@ public: * @note End coordinates can be negative, which represents the number of elements before the end of that dimension. * @note End indices are not inclusive unless negative. * - * @param[in] input Source tensor. Data type supported: All - * @param[out] output Destination tensor. Data type supported: Same as @p input + * @param[in] input Source tensor info. Data type supported: All + * @param[out] output Destination tensor info. Data type supported: Same as @p input * @param[in] starts The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input). * @param[in] ends The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input). */ - void configure(const ITensor *input, ITensor *output, const Coordinates &starts, const Coordinates &ends); + void configure(const ITensorInfo *input, ITensorInfo *output, const Coordinates &starts, const Coordinates &ends); /** Static function to check if given info will lead to a valid configuration of @ref NESlice * @@ -124,13 +138,7 @@ public: * @return A status */ static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Coordinates &starts, const Coordinates &ends); - - // Inherited methods overridden: - void run() override; - -private: - struct Impl; - std::unique_ptr<Impl> _impl; }; +} // namespace experimental } // namespace arm_compute #endif /* ARM_COMPUTE_NE_SLICE_H */ diff --git a/arm_compute/runtime/NEON/functions/NEStridedSlice.h b/arm_compute/runtime/NEON/functions/NEStridedSlice.h index 7ba6a52a58..0b4c2a63a1 100644 --- a/arm_compute/runtime/NEON/functions/NEStridedSlice.h +++ b/arm_compute/runtime/NEON/functions/NEStridedSlice.h @@ -32,18 +32,37 @@ namespace arm_compute // Forward Declarations class ITensor; -namespace experimental -{ /** Basic function to run @ref NEStridedSliceKernel */ -class NEStridedSlice : public INEOperator +class NEStridedSlice : public IFunction { public: + /** Default Constructor */ + NEStridedSlice(); + /** Default Destructor */ + ~NEStridedSlice(); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + NEStridedSlice(const NEStridedSlice &) = delete; + /** Default move constructor */ + NEStridedSlice(NEStridedSlice &&); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + NEStridedSlice &operator=(const NEStridedSlice &) = delete; + /** Default move assignment operator */ + NEStridedSlice &operator=(NEStridedSlice &&); + /** Configure kernel * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:------|:------| + * |All |All | + * * @note Supported tensor rank: up to 4 * - * @param[in] input Source tensor info. Data type supported: All - * @param[out] output Destination tensor info. Data type supported: Same as @p input + * @param[in] input Source tensor. Data type supported: All + * @param[out] output Destination tensor. Data type supported: Same as @p input * @param[in] starts The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input). * @param[in] ends The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input). * @param[in] strides The strides of the dimensions of the input tensor to be sliced. The length must be of rank(input). @@ -52,7 +71,7 @@ public: * @param[in] shrink_axis_mask (Optional) If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1. * A slice of size 1 starting from starts[i] in the dimension must be preserved. */ - void configure(const ITensorInfo *input, ITensorInfo *output, + void configure(const ITensor *input, ITensor *output, const Coordinates &starts, const Coordinates &ends, const BiStrides &strides, int32_t begin_mask = 0, int32_t end_mask = 0, int32_t shrink_axis_mask = 0); @@ -73,26 +92,21 @@ public: static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Coordinates &starts, const Coordinates &ends, const BiStrides &strides, int32_t begin_mask = 0, int32_t end_mask = 0, int32_t shrink_axis_mask = 0); + + // Inherited methods overridden: + void run() override; + +private: + struct Impl; + std::unique_ptr<Impl> _impl; }; -} // namespace experimental +namespace experimental +{ /** Basic function to run @ref NEStridedSliceKernel */ -class NEStridedSlice : public IFunction +class NEStridedSlice : public INEOperator { public: - /** Default Constructor */ - NEStridedSlice(); - /** Default Destructor */ - ~NEStridedSlice(); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - NEStridedSlice(const NEStridedSlice &) = delete; - /** Default move constructor */ - NEStridedSlice(NEStridedSlice &&); - /** Prevent instances of this class from being copied (As this class contains pointers) */ - NEStridedSlice &operator=(const NEStridedSlice &) = delete; - /** Default move assignment operator */ - NEStridedSlice &operator=(NEStridedSlice &&); - /** Configure kernel * * Valid data layouts: @@ -105,8 +119,8 @@ public: * * @note Supported tensor rank: up to 4 * - * @param[in] input Source tensor. Data type supported: All - * @param[out] output Destination tensor. Data type supported: Same as @p input + * @param[in] input Source tensor info. Data type supported: All + * @param[out] output Destination tensor info. Data type supported: Same as @p input * @param[in] starts The starts of the dimensions of the input tensor to be sliced. The length must be of rank(input). * @param[in] ends The ends of the dimensions of the input tensor to be sliced. The length must be of rank(input). * @param[in] strides The strides of the dimensions of the input tensor to be sliced. The length must be of rank(input). @@ -115,7 +129,7 @@ public: * @param[in] shrink_axis_mask (Optional) If the ith bit of shrink_axis_mask is set, it implies that the ith specification shrinks the dimensionality by 1. * A slice of size 1 starting from starts[i] in the dimension must be preserved. */ - void configure(const ITensor *input, ITensor *output, + void configure(const ITensorInfo *input, ITensorInfo *output, const Coordinates &starts, const Coordinates &ends, const BiStrides &strides, int32_t begin_mask = 0, int32_t end_mask = 0, int32_t shrink_axis_mask = 0); @@ -136,13 +150,7 @@ public: static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Coordinates &starts, const Coordinates &ends, const BiStrides &strides, int32_t begin_mask = 0, int32_t end_mask = 0, int32_t shrink_axis_mask = 0); - - // Inherited methods overridden: - void run() override; - -private: - struct Impl; - std::unique_ptr<Impl> _impl; }; +} // namespace experimental } // namespace arm_compute #endif /* ARM_COMPUTE_NE_STRIDED_SLICE_H */ |