1 /*
2 * Copyright (c) 2017-2021 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 "arm_compute/runtime/NEON/functions/NESoftmaxLayer.h"
25 #include "arm_compute/core/Validate.h"
26 #include "arm_compute/runtime/MemoryGroup.h"
27 #include "arm_compute/runtime/Tensor.h"
28 #include "src/core/helpers/MemoryHelpers.h"
29 #include "src/core/helpers/SoftmaxHelpers.h"
30 #include "src/cpu/kernels/CpuSoftmaxKernel.h"
31 #include "src/cpu/operators/CpuSoftmax.h"
32
33 namespace arm_compute
34 {
35 template <bool IS_LOG>
36 struct NESoftmaxLayerGeneric<IS_LOG>::Impl
37 {
38 const ITensor *src{ nullptr };
39 ITensor *dst{ nullptr };
40 Tensor max{ nullptr };
41 std::unique_ptr<cpu::CpuSoftmaxGeneric<IS_LOG>> op{ nullptr };
42 MemoryGroup memory_group{};
43 ITensorPack run_pack{};
44 WorkspaceData<Tensor> workspace_tensors{};
45 };
46
47 template <bool IS_LOG>
NESoftmaxLayerGeneric(std::shared_ptr<IMemoryManager> memory_manager)48 NESoftmaxLayerGeneric<IS_LOG>::NESoftmaxLayerGeneric(std::shared_ptr<IMemoryManager> memory_manager)
49 : _impl(std::make_unique<Impl>())
50 {
51 _impl->memory_group = MemoryGroup(std::move(memory_manager));
52 }
53
54 template <bool IS_LOG>
55 NESoftmaxLayerGeneric<IS_LOG>::NESoftmaxLayerGeneric(NESoftmaxLayerGeneric &&) = default;
56 template <bool IS_LOG>
57 NESoftmaxLayerGeneric<IS_LOG> &NESoftmaxLayerGeneric<IS_LOG>::operator=(NESoftmaxLayerGeneric &&) = default;
58 template <bool IS_LOG>
59 NESoftmaxLayerGeneric<IS_LOG>::~NESoftmaxLayerGeneric() = default;
60
61 template <bool IS_LOG>
configure(ITensor * input,ITensor * output,float beta,int32_t axis)62 void NESoftmaxLayerGeneric<IS_LOG>::configure(ITensor *input, ITensor *output, float beta, int32_t axis)
63 {
64 ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
65
66 _impl->src = input;
67 _impl->dst = output;
68 _impl->op = std::make_unique<cpu::CpuSoftmaxGeneric<IS_LOG>>();
69 _impl->op->configure(input->info(), output->info(), beta, axis);
70
71 _impl->run_pack = { { TensorType::ACL_SRC, _impl->src }, { TensorType::ACL_DST, _impl->dst } };
72 _impl->workspace_tensors = manage_workspace<Tensor>(_impl->op->workspace(), _impl->memory_group, _impl->run_pack);
73 }
74
75 template <bool IS_LOG>
validate(const ITensorInfo * input,const ITensorInfo * output,float beta,int32_t axis)76 Status NESoftmaxLayerGeneric<IS_LOG>::validate(const ITensorInfo *input, const ITensorInfo *output, float beta, int32_t axis)
77 {
78 ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
79 ARM_COMPUTE_RETURN_ON_ERROR(cpu::CpuSoftmaxGeneric<IS_LOG>::validate(input, output, beta, axis));
80 return Status{};
81 }
82
83 template <bool IS_LOG>
run()84 void NESoftmaxLayerGeneric<IS_LOG>::run()
85 {
86 // Acquire all the temporaries
87 MemoryGroupResourceScope scope_mg(_impl->memory_group);
88 ARM_COMPUTE_ERROR_ON_NULLPTR(_impl->src, _impl->dst);
89 _impl->op->run(_impl->run_pack);
90 }
91
92 template class NESoftmaxLayerGeneric<false>;
93 template class NESoftmaxLayerGeneric<true>;
94
95 } // namespace arm_compute
96