diff options
Diffstat (limited to 'src/runtime/gpu/cl/operators')
-rw-r--r-- | src/runtime/gpu/cl/operators/ClScale.cpp | 69 | ||||
-rw-r--r-- | src/runtime/gpu/cl/operators/ClScale.h | 74 |
2 files changed, 143 insertions, 0 deletions
diff --git a/src/runtime/gpu/cl/operators/ClScale.cpp b/src/runtime/gpu/cl/operators/ClScale.cpp new file mode 100644 index 0000000000..4730c8a16e --- /dev/null +++ b/src/runtime/gpu/cl/operators/ClScale.cpp @@ -0,0 +1,69 @@ +/* + * Copyright (c) 2021 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#include "src/runtime/gpu/cl/operators/ClScale.h" + +#include "arm_compute/core/Error.h" +#include "arm_compute/runtime/CL/CLScheduler.h" +#include "src/core/CL/kernels/CLFillBorderKernel.h" +#include "src/core/gpu/cl/ClCompileContext.h" +#include "src/core/gpu/cl/kernels/ClScaleKernel.h" + +namespace arm_compute +{ +namespace opencl +{ +void ClScale::configure(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, const ScaleKernelInfo &info) +{ + ARM_COMPUTE_ERROR_ON_NULLPTR(src); + // Configure Scale kernel + auto k = std::make_unique<kernels::ClScaleKernel>(); + k->set_target(CLScheduler::get().target()); + k->configure(compile_context, src, dst, info); + _kernel = std::move(k); + if(!_kernel->border_size().empty()) + { + auto b = std::make_unique<CLFillBorderKernel>(); + b->configure(compile_context, src, _kernel->border_size(), info.border_mode, info.constant_border_value); + _border_handler = std::move(b); + } + // Tune kernel + CLScheduler::get().tune_kernel_static(*_kernel); +} + +Status ClScale::validate(const ITensorInfo *src, const ITensorInfo *dst, const ScaleKernelInfo &info) +{ + return kernels::ClScaleKernel::validate(src, dst, info); +} + +void ClScale::run(ITensorPack &tensors) +{ + ARM_COMPUTE_ERROR_ON_MSG(tensors.empty(), "No inputs provided"); + if(!_kernel->border_size().empty()) + { + CLScheduler::get().enqueue_op(*_border_handler.get(), tensors, false); + } + CLScheduler::get().enqueue_op(*_kernel.get(), tensors); +} +} // namespace opencl +} // namespace arm_compute
\ No newline at end of file diff --git a/src/runtime/gpu/cl/operators/ClScale.h b/src/runtime/gpu/cl/operators/ClScale.h new file mode 100644 index 0000000000..6eccb59be8 --- /dev/null +++ b/src/runtime/gpu/cl/operators/ClScale.h @@ -0,0 +1,74 @@ +/* + * Copyright (c) 2021 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#ifndef ARM_COMPUTE_CL_SCALE_H +#define ARM_COMPUTE_CL_SCALE_H + +#include "arm_compute/core/KernelDescriptors.h" +#include "src/core/gpu/cl/ClCompileContext.h" +#include "src/runtime/gpu/cl/IClOperator.h" + +namespace arm_compute +{ +namespace opencl +{ +/** Basic function to simulate a scale layer. This function calls the following OpenCL kernels: + * + * -# @ref CLFillBorderKernel (executed if padding size is different from zero) + * -# @ref kernels::ClScaleKernel + */ +class ClScale : public IClOperator +{ +public: + /** Constructor */ + ClScale() = default; + /** Initialize the function's source, destination, interpolation type and border_mode. + * + * @param[in] compile_context The compile context to be used. + * @param[in,out] src Source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/F16/F32. (Written to only for @p border_mode != UNDEFINED) + * @param[out] dst Destination tensor info. Data types supported: Same as @p src + * All but the lowest two dimensions must be the same size as in the input tensor, i.e. scaling is only performed within the XY-plane. + * @param[in] info @ref ScaleKernelInfo descriptor to be used to configure + */ + void configure(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, const ScaleKernelInfo &info); + + /** Static function to check if given info will lead to a valid configuration of @ref ClScale + * + * @param[in] src Source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/F16/F32. + * @param[in] dst Output tensor info. Data type supported: Same as @p src + * All but the lowest two dimensions must be the same size as in the input tensor, i.e. scaling is only performed within the XY-plane. + * @param[in] info @ref ScaleKernelInfo descriptor to be used to validate + * + * @return a status + */ + static Status validate(const ITensorInfo *src, const ITensorInfo *dst, const ScaleKernelInfo &info); + + // Inherited method overridden + void run(ITensorPack &tensors) override; + +protected: + std::unique_ptr<ICLKernel> _border_handler{ nullptr }; +}; +} // namespace opencl +} // namespace arm_compute +#endif /*ARM_COMPUTE_CLSCALE_H */ |