Sheri Zhang | f9ab9f9 | 2021-03-16 12:09:15 +0000 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright (c) 2021 Arm Limited. |
| 3 | * |
| 4 | * SPDX-License-Identifier: MIT |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to |
| 8 | * deal in the Software without restriction, including without limitation the |
| 9 | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in all |
| 14 | * copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | * SOFTWARE. |
| 23 | */ |
| 24 | #include "src/runtime/gpu/cl/operators/ClPixelWiseMultiplication.h" |
| 25 | |
| 26 | #include "arm_compute/runtime/CL/CLScheduler.h" |
| 27 | #include "src/core/CL/kernels/CLFillBorderKernel.h" |
| 28 | #include "src/core/gpu/cl/ClCompileContext.h" |
| 29 | #include "src/core/gpu/cl/kernels/ClPixelWiseMultiplicationKernel.h" |
| 30 | |
| 31 | namespace arm_compute |
| 32 | { |
| 33 | namespace opencl |
| 34 | { |
| 35 | namespace |
| 36 | { |
| 37 | ITensorPack select_border_input(ITensorPack &tensors) |
| 38 | { |
| 39 | ITensorPack pack; |
| 40 | if(tensors.get_tensor(TensorType::ACL_DST)->info()->dimension(0) > 1) |
| 41 | { |
| 42 | if(tensors.get_const_tensor(TensorType::ACL_SRC_1)->info()->dimension(0) == 1) |
| 43 | { |
| 44 | pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_1)); |
| 45 | } |
| 46 | else |
| 47 | { |
| 48 | pack.add_tensor(TensorType::ACL_SRC, tensors.get_const_tensor(TensorType::ACL_SRC_0)); |
| 49 | } |
| 50 | } |
| 51 | return pack; |
| 52 | } |
| 53 | } // namespace |
| 54 | |
| 55 | void ClPixelWiseMultiplication::configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, float scale, |
| 56 | ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info) |
| 57 | { |
| 58 | auto k = std::make_unique<kernels::ClPixelWiseMultiplicationKernel>(); |
| 59 | k->configure(compile_context, src1, src2, dst, scale, overflow_policy, rounding_policy, act_info); |
| 60 | _kernel = std::move(k); |
| 61 | |
| 62 | if(dst->dimension(0) > 1) |
| 63 | { |
| 64 | ITensorInfo *broadcasted_info = (src1->dimension(0) == 1) ? src1 : src2; |
| 65 | |
| 66 | if(broadcasted_info->dimension(0) == 1) |
| 67 | { |
| 68 | auto b = std::make_unique<CLFillBorderKernel>(); |
| 69 | b->configure(compile_context, broadcasted_info, _kernel->border_size(), BorderMode::REPLICATE); |
| 70 | _border_handler = std::move(b); |
| 71 | } |
| 72 | } |
| 73 | } |
| 74 | |
| 75 | Status ClPixelWiseMultiplication::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, float scale, |
| 76 | ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info) |
| 77 | { |
| 78 | return kernels::ClPixelWiseMultiplicationKernel::validate(src1, src2, dst, scale, overflow_policy, rounding_policy, act_info); |
| 79 | } |
| 80 | |
| 81 | void ClPixelWiseMultiplication::run(ITensorPack &tensors) |
| 82 | { |
| 83 | if(_border_handler) |
| 84 | { |
| 85 | auto border_pack = select_border_input(tensors); |
| 86 | CLScheduler::get().enqueue_op(*_border_handler, border_pack); |
| 87 | } |
| 88 | ICLOperator::run(tensors); |
| 89 | } |
| 90 | |
| 91 | void ClComplexPixelWiseMultiplication::configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info) |
| 92 | { |
| 93 | auto k = std::make_unique<kernels::ClComplexPixelWiseMultiplicationKernel>(); |
| 94 | k->configure(compile_context, src1, src2, dst, act_info); |
| 95 | _kernel = std::move(k); |
| 96 | |
| 97 | if(dst->dimension(0) > 1) |
| 98 | { |
| 99 | ITensorInfo *broadcasted_info = (src1->dimension(0) == 1) ? src1 : src2; |
| 100 | |
| 101 | if(broadcasted_info->dimension(0) == 1) |
| 102 | { |
| 103 | auto b = std::make_unique<CLFillBorderKernel>(); |
| 104 | b->configure(compile_context, broadcasted_info, _kernel->border_size(), BorderMode::REPLICATE); |
| 105 | _border_handler = std::move(b); |
| 106 | } |
| 107 | } |
| 108 | } |
| 109 | |
| 110 | Status ClComplexPixelWiseMultiplication::validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info) |
| 111 | { |
| 112 | return kernels::ClComplexPixelWiseMultiplicationKernel::validate(src1, src2, dst, act_info); |
| 113 | } |
| 114 | |
| 115 | void ClComplexPixelWiseMultiplication::run(ITensorPack &tensors) |
| 116 | { |
| 117 | if(_border_handler) |
| 118 | { |
| 119 | auto border_pack = select_border_input(tensors); |
| 120 | CLScheduler::get().enqueue_op(*_border_handler, border_pack); |
| 121 | } |
| 122 | ICLOperator::run(tensors); |
| 123 | } |
| 124 | } // namespace opencl |
| 125 | } // namespace arm_compute |