已合并
feat: 逻辑流分配后dump图 #4508
夏国正创建于 16 天前
feat: 逻辑流分配后dump图 #4508
已合并
共 3 个文件变更+35-32
| @@ -1160,6 +1160,7 @@ Status ModelBuilder::BuildModelForGetTask(ge::Model &model) { | |||
| 1160 | GE_ASSERT_SUCCESS(stream_allocator_.AssignLogicalStreams(stream_max_parallel_num_, hcom_parallel_), | 1160 | GE_ASSERT_SUCCESS(stream_allocator_.AssignLogicalStreams(stream_max_parallel_num_, hcom_parallel_), |
| 1161 | "[Assign][LogicalStreams] failed. graph:%s", compute_graph_->GetName().c_str()); | 1161 | "[Assign][LogicalStreams] failed. graph:%s", compute_graph_->GetName().c_str()); |
| 1162 | GE_COMPILE_TRACE_TIMESTAMP_END(AssignLogicalStreams, "GraphBuilder::AssignLogicalStreams"); | 1162 | GE_COMPILE_TRACE_TIMESTAMP_END(AssignLogicalStreams, "GraphBuilder::AssignLogicalStreams"); |
| 1163 | + GE_DUMP(compute_graph_, "AfterAssignLogicalStreams"); | ||
| 1163 | 1164 | ||
| 1164 | // Assign functional op labels. | 1165 | // Assign functional op labels. |
| 1165 | auto root_graph = GraphUtils::FindRootGraph(compute_graph_); | 1166 | auto root_graph = GraphUtils::FindRootGraph(compute_graph_); |
| @@ -130,19 +130,20 @@ export DUMP_GRAPH_LEVEL=1 #控制dump图的个数 | |||
| 130 | | 106 | ge_proto_xxxx_AfterUnfoldSubgraphs.txt | GE | 动态shape图展开之后的图 | - | | 130 | | 106 | ge_proto_xxxx_AfterUnfoldSubgraphs.txt | GE | 动态shape图展开之后的图 | - | |
| 131 | | 107 | ge_proto_xxxx_RunCustomPass_BeforeAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之前的图 | 白名单 | | 131 | | 107 | ge_proto_xxxx_RunCustomPass_BeforeAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之前的图 | 白名单 | |
| 132 | | 108 | ge_proto_xxxx_RunCustomPass_AfterAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之后的图 | 白名单 | | 132 | | 108 | ge_proto_xxxx_RunCustomPass_AfterAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之后的图 | 白名单 | |
| 133 | -| 109 | ge_proto_xxxx_AfterAssignResource.txt | GE | 流分配、内存分配、GenTask之后的图 | - | | 133 | +| 109 | ge_proto_xxxx_AfterAssignLogicalStreams.txt | GE | 完成逻辑流分配后的图 | - | |
| 134 | -| 110 | ge_proto_xxxx_Build.txt | GE | 模型编译出口图 | 白名单 | | 134 | +| 110 | ge_proto_xxxx_AfterAssignResource.txt | GE | 流分配、内存分配、GenTask之后的图 | - | |
| 135 | -| 111 | ge_proto_xxxx_PreRunAfterBuild.txt | GE | 与上图相同 | - | | 135 | +| 111 | ge_proto_xxxx_Build.txt | GE | 模型编译出口图 | 白名单 | |
| 136 | -| 112 | ge_proto_xxxx_BeforeAttrsCompress.txt | GE | 离线模型压缩前的图 | - | | 136 | +| 112 | ge_proto_xxxx_PreRunAfterBuild.txt | GE | 与上图相同 | - | |
| 137 | -| 113 | ge_proto_xxxx_AfterAttrsCompress.txt | GE | 离线模型压缩后的图 | - | | 137 | +| 113 | ge_proto_xxxx_BeforeAttrsCompress.txt | GE | 离线模型压缩前的图 | - | |
| 138 | -| 114 | ge_proto_xxxx_BeforeAttrsDecompress.txt | GE | 离线模型解压前的图 | - | | 138 | +| 114 | ge_proto_xxxx_AfterAttrsCompress.txt | GE | 离线模型压缩后的图 | - | |
| 139 | -| 115 | ge_proto_xxxx_AfterAttrsDecompress.txt | GE | 离线模型解压后的图 | - | | 139 | +| 115 | ge_proto_xxxx_BeforeAttrsDecompress.txt | GE | 离线模型解压前的图 | - | |
| 140 | -| 116 | ge_proto_xxxx_ComputeGraphBeforeLowering.txt | GE | lowering前的计算图 | 白名单 | | 140 | +| 116 | ge_proto_xxxx_AfterAttrsDecompress.txt | GE | 离线模型解压后的图 | - | |
| 141 | -| 117 | ge_proto_xxxx_Before_MultiStream_LoweringFirstLastEventSync.txt | GE | 多流插入Event之前的执行图 | - | | 141 | +| 117 | ge_proto_xxxx_ComputeGraphBeforeLowering.txt | GE | lowering前的计算图 | 白名单 | |
| 142 | -| 118 | ge_proto_xxxx_ExeGraphBeforeOptimize.txt | GE | 执行图优化前的执行图 | 白名单 | | 142 | +| 118 | ge_proto_xxxx_Before_MultiStream_LoweringFirstLastEventSync.txt | GE | 多流插入Event之前的执行图 | - | |
| 143 | -| 119 | ge_proto_xxxx_AfterZeroCopy.txt | GE | 零拷贝优化之后的执行图 | - | | 143 | +| 119 | ge_proto_xxxx_ExeGraphBeforeOptimize.txt | GE | 执行图优化前的执行图 | 白名单 | |
| 144 | -| 120 | ge_proto_xxxx_AfterCEM.txt | GE | CEM优化之后的执行图 | - | | 144 | +| 120 | ge_proto_xxxx_AfterZeroCopy.txt | GE | 零拷贝优化之后的执行图 | - | |
| 145 | -| 121 | ge_proto_xxxx_AfterCopyFlowLaunch.txt | GE | 随路拷贝优化后的执行图 | - | | 145 | +| 121 | ge_proto_xxxx_AfterCEM.txt | GE | CEM优化之后的执行图 | - | |
| 146 | -| 122 | ge_proto_xxxx_TrustOutTensorAfter.txt | GE | TrustOutTensor优化之后的执行图 | - | | 146 | +| 122 | ge_proto_xxxx_AfterCopyFlowLaunch.txt | GE | 随路拷贝优化后的执行图 | - | |
| 147 | -| 123 | ge_proto_xxxx_AfterAicpuFuseHostInputs.txt | GE | AicpuFuseHostInputs优化之后的图 | - | | 147 | +| 123 | ge_proto_xxxx_TrustOutTensorAfter.txt | GE | TrustOutTensor优化之后的执行图 | - | |
| 148 | -| 124 | ge_proto_xxxx_ExecuteGraphAfterSplit.txt | GE | 动态shape最终的执行图 | 白名单 | | 148 | +| 124 | ge_proto_xxxx_AfterAicpuFuseHostInputs.txt | GE | AicpuFuseHostInputs优化之后的图 | - | |
| 149 | +| 125 | ge_proto_xxxx_ExecuteGraphAfterSplit.txt | GE | 动态shape最终的执行图 | 白名单 | | ||
| @@ -150,22 +150,23 @@ DUMP\_GRAPH\_LEVEL环境变量只有在[DUMP\_GE\_GRAPH](DUMP_GE_GRAPH.md)开启 | |||
| 150 | | 106 | ge_proto_xxxx_AfterUnfoldSubgraphs.txt | GE | 动态shape图展开之后的图 | - | | 150 | | 106 | ge_proto_xxxx_AfterUnfoldSubgraphs.txt | GE | 动态shape图展开之后的图 | - | |
| 151 | | 107 | ge_proto_xxxx_RunCustomPass_BeforeAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之前的图 | 白名单 | | 151 | | 107 | ge_proto_xxxx_RunCustomPass_BeforeAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之前的图 | 白名单 | |
| 152 | | 108 | ge_proto_xxxx_RunCustomPass_AfterAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之后的图 | 白名单 | | 152 | | 108 | ge_proto_xxxx_RunCustomPass_AfterAssignLogicStream{pass_name}.txt | GE | 用户自定义流分配Pass处理之后的图 | 白名单 | |
| 153 | -| 109 | ge_proto_xxxx_AfterAssignResource.txt | GE | 流分配、内存分配、GenTask之后的图 | - | | 153 | +| 109 | ge_proto_xxxx_AfterAssignLogicalStreams.txt | GE | 完成逻辑流分配后的图 | - | |
| 154 | -| 110 | ge_proto_xxxx_Build.txt | GE | 模型编译出口图 | 白名单 | | 154 | +| 110 | ge_proto_xxxx_AfterAssignResource.txt | GE | 流分配、内存分配、GenTask之后的图 | - | |
| 155 | -| 111 | ge_proto_xxxx_PreRunAfterBuild.txt | GE | 与上图相同 | - | | 155 | +| 111 | ge_proto_xxxx_Build.txt | GE | 模型编译出口图 | 白名单 | |
| 156 | -| 112 | ge_proto_xxxx_BeforeAttrsCompress.txt | GE | 离线模型压缩前的图 | - | | 156 | +| 112 | ge_proto_xxxx_PreRunAfterBuild.txt | GE | 与上图相同 | - | |
| 157 | -| 113 | ge_proto_xxxx_AfterAttrsCompress.txt | GE | 离线模型压缩后的图 | - | | 157 | +| 113 | ge_proto_xxxx_BeforeAttrsCompress.txt | GE | 离线模型压缩前的图 | - | |
| 158 | -| 114 | ge_proto_xxxx_BeforeAttrsDecompress.txt | GE | 离线模型解压前的图 | - | | 158 | +| 114 | ge_proto_xxxx_AfterAttrsCompress.txt | GE | 离线模型压缩后的图 | - | |
| 159 | -| 115 | ge_proto_xxxx_AfterAttrsDecompress.txt | GE | 离线模型解压后的图 | - | | 159 | +| 115 | ge_proto_xxxx_BeforeAttrsDecompress.txt | GE | 离线模型解压前的图 | - | |
| 160 | -| 116 | ge_proto_xxxx_ComputeGraphBeforeLowering.txt | GE | lowering前的计算图 | 白名单 | | 160 | +| 116 | ge_proto_xxxx_AfterAttrsDecompress.txt | GE | 离线模型解压后的图 | - | |
| 161 | -| 117 | ge_proto_xxxx_Before_MultiStream_LoweringFirstLastEventSync.txt | GE | 多流插入Event之前的执行图 | - | | 161 | +| 117 | ge_proto_xxxx_ComputeGraphBeforeLowering.txt | GE | lowering前的计算图 | 白名单 | |
| 162 | -| 118 | ge_proto_xxxx_ExeGraphBeforeOptimize.txt | GE | 执行图优化前的执行图 | 白名单 | | 162 | +| 118 | ge_proto_xxxx_Before_MultiStream_LoweringFirstLastEventSync.txt | GE | 多流插入Event之前的执行图 | - | |
| 163 | -| 119 | ge_proto_xxxx_AfterZeroCopy.txt | GE | 零拷贝优化之后的执行图 | - | | 163 | +| 119 | ge_proto_xxxx_ExeGraphBeforeOptimize.txt | GE | 执行图优化前的执行图 | 白名单 | |
| 164 | -| 120 | ge_proto_xxxx_AfterCEM.txt | GE | CEM优化之后的执行图 | - | | 164 | +| 120 | ge_proto_xxxx_AfterZeroCopy.txt | GE | 零拷贝优化之后的执行图 | - | |
| 165 | -| 121 | ge_proto_xxxx_AfterCopyFlowLaunch.txt | GE | 随路拷贝优化后的执行图 | - | | 165 | +| 121 | ge_proto_xxxx_AfterCEM.txt | GE | CEM优化之后的执行图 | - | |
| 166 | -| 122 | ge_proto_xxxx_TrustOutTensorAfter.txt | GE | TrustOutTensor优化之后的执行图 | - | | 166 | +| 122 | ge_proto_xxxx_AfterCopyFlowLaunch.txt | GE | 随路拷贝优化后的执行图 | - | |
| 167 | -| 123 | ge_proto_xxxx_AfterAicpuFuseHostInputs.txt | GE | AicpuFuseHostInputs优化之后的图 | - | | 167 | +| 123 | ge_proto_xxxx_TrustOutTensorAfter.txt | GE | TrustOutTensor优化之后的执行图 | - | |
| 168 | -| 124 | ge_proto_xxxx_ExecuteGraphAfterSplit.txt | GE | 动态shape最终的执行图 | 白名单 | | 168 | +| 124 | ge_proto_xxxx_AfterAicpuFuseHostInputs.txt | GE | AicpuFuseHostInputs优化之后的图 | - | |
| 169 | +| 125 | ge_proto_xxxx_ExecuteGraphAfterSplit.txt | GE | 动态shape最终的执行图 | 白名单 | | ||
| 169 | 170 | ||
| 170 | ## 产品支持情况 | 171 | ## 产品支持情况 |
| 171 | 172 | ||