Line data Source code
1 : /**
2 : * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3 : * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4 : * CANN Open Software License Agreement Version 2.0 (the "License").
5 : * Please refer to the License for details. You may not use this file except in compliance with the License.
6 : * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7 : * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8 : * See LICENSE in the root of the software repository for the full text of the License.
9 : */
10 :
11 : /*!
12 : * \file relu_tiling_arch35.cpp
13 : * \brief
14 : */
15 : #include "relu_tiling_arch35.h"
16 : #include <iostream>
17 : #include <graph/utils/type_utils.h>
18 : #include "tiling/platform/platform_ascendc.h"
19 : #include "register/op_impl_registry.h"
20 : #include "log/log.h"
21 : #include "register/tilingdata_base.h"
22 : #include "activation/relu/op_kernel/arch35/relu_dag.h"
23 : #include "atvoss/elewise/elewise_base_struct.h"
24 :
25 : namespace optiling {
26 : using namespace ge;
27 : using namespace Ops::Base;
28 :
29 : constexpr uint64_t SYS_WORKSPACE = 16777216; // 16M
30 : constexpr uint64_t RELU_TILING_KEY_ELEMENTWISE_FP16 = 101;
31 : constexpr uint64_t RELU_TILING_KEY_ELEMENTWISE_BF16 = 102;
32 : constexpr uint64_t RELU_TILING_KEY_ELEMENTWISE_FP32 = 103;
33 : constexpr uint64_t RELU_TILING_KEY_ELEMENTWISE_INT8 = 104;
34 : constexpr uint64_t RELU_TILING_KEY_ELEMENTWISE_INT32 = 105;
35 : constexpr uint64_t RELU_TILING_KEY_ELEMENTWISE_INT64 = 106;
36 :
37 : ge::graphStatus ReluTiling::CalcOutputDtype()
38 : {
39 : auto inputDesc = tilingContext->GetInputDesc(0);
40 : OP_CHECK_NULL_WITH_CONTEXT(tilingContext, inputDesc);
41 : ge::DataType inputDtype = inputDesc->GetDataType();
42 :
43 : auto outputDesc = tilingContext->GetOutputDesc(0);
44 : OP_CHECK_NULL_WITH_CONTEXT(tilingContext, outputDesc);
45 : this->outputDtype = outputDesc->GetDataType();
46 :
47 : OP_CHECK_IF(inputDtype != this->outputDtype,
48 : OP_LOGE_FOR_INVALID_DTYPES_WITH_REASON(tilingContext->GetNodeName(), "x, y",
49 : ge::TypeUtils::DataTypeToSerialString(inputDtype) + ", " +
50 : ge::TypeUtils::DataTypeToSerialString(this->outputDtype),
51 : "The dtypes of x and y must be the same"),
52 : return ge::GRAPH_FAILED);
53 : return ge::GRAPH_SUCCESS;
54 : }
55 :
56 : ge::graphStatus ReluTiling::RunTiling()
57 : {
58 : auto tiling = tilingContext->GetTilingData<Ops::Base::EleBaseTilingData16B>();
59 : ElewiseBaseTiling elewiseBaseTiling(tilingContext);
60 : OP_CHECK_IF(CalcOutputDtype() == ge::GRAPH_FAILED, OP_LOGE(tilingContext, "get output dtype failed"),
61 : return ge::GRAPH_FAILED);
62 : ge::graphStatus res = ge::GRAPH_FAILED;
63 : if (this->outputDtype == ge::DT_FLOAT16) {
64 : res = elewiseBaseTiling.DoTiling<ReluOp::GraphRelu<half, half>::OpDag>(*tiling);
65 : } else if (this->outputDtype == ge::DT_FLOAT) {
66 : res = elewiseBaseTiling.DoTiling<ReluOp::GraphRelu<float, float>::OpDag>(*tiling);
67 : } else if (this->outputDtype == ge::DT_BF16) {
68 : res = elewiseBaseTiling.DoTiling<ReluOp::GraphRelu<half, float>::OpDag>(*tiling);
69 : } else if (this->outputDtype == ge::DT_INT8) {
70 : res = elewiseBaseTiling.DoTiling<ReluOp::GraphRelu<int8_t, half>::OpDag>(*tiling);
71 : } else if (this->outputDtype == ge::DT_INT32) {
72 : res = elewiseBaseTiling.DoTiling<ReluOp::GraphRelu<int32_t, int32_t>::OpDag>(*tiling);
73 : } else if (this->outputDtype == ge::DT_INT64) {
74 : res = elewiseBaseTiling.DoTiling<ReluOp::GraphReluMax<int64_t>::OpDag>(*tiling);
75 : } else {
76 : OP_LOGE_FOR_INVALID_DTYPE(tilingContext->GetNodeName(), "y",
77 : ge::TypeUtils::DataTypeToSerialString(this->outputDtype),
78 : "DT_FLOAT16, DT_BF16, DT_FLOAT, DT_INT8, DT_INT32, DT_INT64");
79 : return ge::GRAPH_FAILED;
80 : }
81 :
82 10 : OP_CHECK_IF(res == ge::GRAPH_FAILED,
83 : OP_LOGE(tilingContext->GetNodeName(), "DoTiling failed, output dtype: %s.",
84 : ge::TypeUtils::DataTypeToSerialString(this->outputDtype).c_str()),
85 : return ge::GRAPH_FAILED);
86 :
87 : size_t* currentWorkspace = tilingContext->GetWorkspaceSizes(1);
88 : OP_CHECK_NULL_WITH_CONTEXT(tilingContext, currentWorkspace);
89 : currentWorkspace[0] = SYS_WORKSPACE;
90 : if (this->outputDtype == ge::DT_FLOAT16) {
91 : tilingContext->SetTilingKey(RELU_TILING_KEY_ELEMENTWISE_FP16);
92 : } else if (this->outputDtype == ge::DT_BF16) {
93 : tilingContext->SetTilingKey(RELU_TILING_KEY_ELEMENTWISE_BF16);
94 : } else if (this->outputDtype == ge::DT_FLOAT) {
95 : tilingContext->SetTilingKey(RELU_TILING_KEY_ELEMENTWISE_FP32);
96 : } else if (this->outputDtype == ge::DT_INT8) {
97 : tilingContext->SetTilingKey(RELU_TILING_KEY_ELEMENTWISE_INT8);
98 : } else if (this->outputDtype == ge::DT_INT32) {
99 : tilingContext->SetTilingKey(RELU_TILING_KEY_ELEMENTWISE_INT32);
100 : } else if (this->outputDtype == ge::DT_INT64) {
101 : tilingContext->SetTilingKey(RELU_TILING_KEY_ELEMENTWISE_INT64);
102 : }
103 :
104 : tilingContext->SetBlockDim(elewiseBaseTiling.GetBlockDim());
105 : return ge::GRAPH_SUCCESS;
106 : }
107 :
108 : static ge::graphStatus Tiling4Relu(gert::TilingContext* context)
109 : {
110 : OP_LOGD("ReluTiling", "Enter Tiling4Relu");
111 : if (context == nullptr) {
112 : OP_LOGE("ReluTiling", "Tiling context is null");
113 : return ge::GRAPH_FAILED;
114 : }
115 :
116 : auto compileInfo = context->GetCompileInfo<ReluCompileInfo>();
117 : OP_CHECK_NULL_WITH_CONTEXT(context, compileInfo);
118 : ReluTiling tiling(context);
119 : return tiling.RunTiling();
120 : }
121 :
122 : ge::graphStatus TilingPrepareForRelu(gert::TilingParseContext* context)
123 : {
124 : auto compileInfoPtr = context->GetCompiledInfo<ReluCompileInfo>();
125 : OP_CHECK_NULL_WITH_CONTEXT(context, compileInfoPtr);
126 : fe::PlatFormInfos* platformInfoPtr = context->GetPlatformInfo();
127 : OP_CHECK_NULL_WITH_CONTEXT(context, platformInfoPtr);
128 : auto ascendcPlatform = platform_ascendc::PlatformAscendC(platformInfoPtr);
129 : compileInfoPtr->coreNum = ascendcPlatform.GetCoreNumAiv();
130 : ascendcPlatform.GetCoreMemSize(platform_ascendc::CoreMemType::UB, compileInfoPtr->ubSize);
131 : return ge::GRAPH_SUCCESS;
132 : }
133 :
134 : IMPL_OP_OPTILING(Relu).Tiling(Tiling4Relu).TilingParse<ReluCompileInfo>(TilingPrepareForRelu);
135 : } // namespace optiling
|
