You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

aicpu_ext_info.cc 11 kB

5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
5 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245
  1. /**
  2. * Copyright 2019-2020 Huawei Technologies Co., Ltd
  3. *
  4. * Licensed under the Apache License, Version 2.0 (the "License");
  5. * you may not use this file except in compliance with the License.
  6. * You may obtain a copy of the License at
  7. *
  8. * http://www.apache.org/licenses/LICENSE-2.0
  9. *
  10. * Unless required by applicable law or agreed to in writing, software
  11. * distributed under the License is distributed on an "AS IS" BASIS,
  12. * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13. * See the License for the specific language governing permissions and
  14. * limitations under the License.
  15. */
  16. #include "hybrid/node_executor/aicpu/aicpu_ext_info.h"
  17. #include "framework/common/util.h"
  18. #include "framework/common/fmk_error_codes.h"
  19. #include "framework/common/debug/log.h"
  20. namespace ge {
  21. namespace hybrid {
  22. namespace {
  23. // if dim count is not reach kMaxShapeDims(8), use INT64_MIN to mark dim end.
  24. constexpr int64_t kDimEndFlag = INT64_MIN;
  25. }
  26. Status AicpuExtInfoHandler::Parse(const std::string &ext_info) {
  27. GELOGI("Node[%s] parse ext info start.", node_name_.c_str());
  28. if (ext_info.empty()) {
  29. GELOGE(ACL_ERROR_GE_PARAM_INVALID, "Node[%s] parse ext info failed as ext info is empty.",
  30. node_name_.c_str());
  31. return ACL_ERROR_GE_PARAM_INVALID;
  32. }
  33. ext_info_len_ = ext_info.size();
  34. ext_info_.reset(new(std::nothrow)uint8_t[ext_info_len_]);
  35. GE_CHECK_NOTNULL(ext_info_);
  36. if (memcpy_s(ext_info_.get(), ext_info_len_, ext_info.c_str(), ext_info.size()) != EOK) {
  37. GELOGE(ACL_ERROR_GE_MEMORY_OPERATE_FAILED, "[%s] Failed to coy ext info", node_name_.c_str());
  38. return ACL_ERROR_GE_MEMORY_OPERATE_FAILED;
  39. }
  40. input_shape_and_type_.clear();
  41. output_shape_and_type_.clear();
  42. auto ext_info_data = ext_info_.get();
  43. size_t offset = 0;
  44. while (offset + sizeof(AicpuExtInfo) <= ext_info_len_) {
  45. auto aicpu_ext_info = reinterpret_cast<AicpuExtInfo *>(ext_info_data + offset);
  46. GELOGD("Ext infoType=%d, infoLen=%u.", aicpu_ext_info->infoType, aicpu_ext_info->infoLen);
  47. switch (aicpu_ext_info->infoType) {
  48. case aicpu::FWKAdapter::FWK_ADPT_EXT_SHAPE_TYPE:
  49. GE_CHK_STATUS_RET(ParseExtShapeType(aicpu_ext_info), "Parse ext shape type failed.");
  50. break;
  51. case aicpu::FWKAdapter::FWK_ADPT_EXT_INPUT_SHAPE:
  52. GE_CHK_STATUS_RET(ParseExtInputShape(aicpu_ext_info), "Parse ext input shape failed.");
  53. break;
  54. case aicpu::FWKAdapter::FWK_ADPT_EXT_OUTPUT_SHAPE:
  55. GE_CHK_STATUS_RET(ParseExtOutputShape(aicpu_ext_info), "Parse ext output shape failed.");
  56. break;
  57. case aicpu::FWKAdapter::FWK_ADPT_EXT_SESSION_INFO:
  58. GE_CHK_STATUS_RET(ParseExtSessionInfo(aicpu_ext_info), "Parse ext session info failed.");
  59. break;
  60. default:
  61. GELOGD("Node[%s] ignore infoType=%d, infoLen=%u.",
  62. node_name_.c_str(), aicpu_ext_info->infoType, aicpu_ext_info->infoLen);
  63. break;
  64. }
  65. offset += sizeof(AicpuExtInfo);
  66. offset += aicpu_ext_info->infoLen;
  67. }
  68. GE_CHK_BOOL_RET_STATUS(offset == ext_info_len_, ACL_ERROR_GE_PARAM_INVALID,
  69. "Node[%s] ext_info format error, parse not reach end, offset=%zu, ext_info_len=%zu.",
  70. node_name_.c_str(), offset, ext_info_len_);
  71. GELOGI("Node[%s] parse ext info end.", node_name_.c_str());
  72. return SUCCESS;
  73. }
  74. Status AicpuExtInfoHandler::ParseExtShapeType(AicpuExtInfo *aicpu_ext_info) {
  75. GE_CHK_BOOL_RET_STATUS(aicpu_ext_info->infoLen == sizeof(int32_t), ACL_ERROR_GE_PARAM_INVALID,
  76. "Node[%s] parse ext shape type failed as infoLen must be %zu but %u.",
  77. node_name_.c_str(), sizeof(int32_t), aicpu_ext_info->infoLen);
  78. auto type = reinterpret_cast<const int32_t *>(aicpu_ext_info->infoMsg);
  79. GE_CHK_BOOL_RET_STATUS(*type == unknown_type_, ACL_ERROR_GE_PARAM_INVALID,
  80. "Node[%s] parse ext shape type failed as need %d but %d.",
  81. node_name_.c_str(), unknown_type_, *type);
  82. GELOGI("Node[%s] parse ext shape type success infoLen=%u.", node_name_.c_str(), aicpu_ext_info->infoLen);
  83. return SUCCESS;
  84. }
  85. Status AicpuExtInfoHandler::ParseExtInputShape(AicpuExtInfo *aicpu_ext_info) {
  86. auto need_len = input_num_ * sizeof(AicpuShapeAndType);
  87. GE_CHK_BOOL_RET_STATUS(aicpu_ext_info->infoLen == need_len, ACL_ERROR_GE_PARAM_INVALID,
  88. "Node[%s] parse ext input shape failed as infoLen must be "
  89. "input_num[%u]*sizeof(ShapeAndType)[%zu] but %u.",
  90. node_name_.c_str(), input_num_, sizeof(AicpuShapeAndType), aicpu_ext_info->infoLen);
  91. auto input = reinterpret_cast<AicpuShapeAndType *>(aicpu_ext_info->infoMsg);
  92. for (uint32_t index = 0; index < input_num_; ++index) {
  93. input_shape_and_type_.emplace_back(&input[index]);
  94. }
  95. GELOGI("Node[%s] parse ext input shape success infoLen=%u.", node_name_.c_str(), aicpu_ext_info->infoLen);
  96. return SUCCESS;
  97. }
  98. Status AicpuExtInfoHandler::ParseExtOutputShape(AicpuExtInfo *aicpu_ext_info) {
  99. if (unknown_type_ == DEPEND_COMPUTE) {
  100. GELOGD("Node[%s] is depend compute type no need ext output shape, ignore it, infoLen=%u.",
  101. node_name_.c_str(), aicpu_ext_info->infoLen);
  102. return SUCCESS;
  103. }
  104. auto need_len = output_num_ * sizeof(AicpuShapeAndType);
  105. GE_CHK_BOOL_RET_STATUS(aicpu_ext_info->infoLen == need_len, ACL_ERROR_GE_PARAM_INVALID,
  106. "Node[%s] parse ext output shape failed as infoLen must be "
  107. "output_num[%u]*sizeof(ShapeAndType)[%zu] but %u.",
  108. node_name_.c_str(), output_num_, sizeof(AicpuShapeAndType), aicpu_ext_info->infoLen);
  109. auto output = reinterpret_cast<AicpuShapeAndType *>(aicpu_ext_info->infoMsg);
  110. for (uint32_t index = 0; index < output_num_; ++index) {
  111. output_shape_and_type_.emplace_back(&output[index]);
  112. }
  113. GELOGI("Node[%s] parse ext output shape success infoLen=%u.", node_name_.c_str(), aicpu_ext_info->infoLen);
  114. return SUCCESS;
  115. }
  116. Status AicpuExtInfoHandler::ParseExtSessionInfo(AicpuExtInfo *aicpu_ext_info) {
  117. GE_CHK_BOOL_RET_STATUS(aicpu_ext_info->infoLen == sizeof(AicpuSessionInfo), ACL_ERROR_GE_PARAM_INVALID,
  118. "Node[%s] parse ext session info failed as infoLen must be %zu but %u.",
  119. node_name_.c_str(), sizeof(SessionInfo), aicpu_ext_info->infoLen);
  120. session_info_ = reinterpret_cast<AicpuSessionInfo *>(aicpu_ext_info->infoMsg);
  121. GELOGI("Node[%s] parse session info success infoLen=%u.", node_name_.c_str(), aicpu_ext_info->infoLen);
  122. return SUCCESS;
  123. }
  124. Status AicpuExtInfoHandler::UpdateSessionInfo(uint64_t session_id, uint64_t kernel_id, bool sess_flag) {
  125. if (session_info_ == nullptr) {
  126. GELOGD("There is no session info in ext_info, no need update.");
  127. return SUCCESS;
  128. }
  129. session_info_->sessionId = session_id;
  130. session_info_->kernelId = kernel_id;
  131. session_info_->sessFlag = sess_flag;
  132. return SUCCESS;
  133. }
  134. Status AicpuExtInfoHandler::UpdateSessionInfoSessionId(uint64_t session_id) {
  135. if (session_info_ == nullptr) {
  136. GELOGD("There is no session info in ext_info, no need update.");
  137. return SUCCESS;
  138. }
  139. session_info_->sessionId = session_id;
  140. session_info_->sessFlag = true;
  141. return SUCCESS;
  142. }
  143. Status AicpuExtInfoHandler::UpdateInputShapeAndType(uint32_t input_index, const GeTensorDesc &input_desc) {
  144. GE_CHECK_LE(input_index, input_num_);
  145. const auto &shape = input_desc.GetShape();
  146. GE_CHK_STATUS_RET(UpdateShapeAndType(shape, input_desc.GetDataType(), input_shape_and_type_[input_index]),
  147. "Node[%s] input[%u] update input shape and type failed.",
  148. node_name_.c_str(), input_index);
  149. return SUCCESS;
  150. }
  151. Status AicpuExtInfoHandler::UpdateOutputShapeAndType(uint32_t output_index, const GeTensorDesc &output_desc) {
  152. GE_CHK_BOOL_RET_STATUS((unknown_type_ != DEPEND_COMPUTE), ACL_ERROR_GE_INTERNAL_ERROR,
  153. "Node[%s] is depend compute is no need update output shape and type by ext.",
  154. node_name_.c_str());
  155. GE_CHECK_LE(output_index, output_num_);
  156. auto shape = output_desc.GetShape();
  157. // shape range need use range update shape
  158. if (unknown_type_ == DEPEND_SHAPE_RANGE) {
  159. std::vector<std::pair<int64_t, int64_t>> range;
  160. auto range_ret = output_desc.GetShapeRange(range);
  161. GE_CHK_BOOL_RET_STATUS(range_ret == GRAPH_SUCCESS, ACL_ERROR_GE_INTERNAL_ERROR,
  162. "Node[%s] is shape range type but get GetShapeRange failed, ret=%u.",
  163. node_name_.c_str(), range_ret);
  164. for (size_t k = 0; k < range.size(); ++k) {
  165. if (shape.GetDim(k) < 0 && k < range.size()) {
  166. GELOGD("Node[%s] output[%u] update dim[%zu] from %ld to range max %ld.",
  167. node_name_.c_str(), output_index, k, shape.GetDim(k), range[k].second);
  168. shape.SetDim(k, range[k].second);
  169. }
  170. }
  171. }
  172. return UpdateShapeAndType(shape, output_desc.GetDataType(), output_shape_and_type_[output_index]);
  173. }
  174. Status AicpuExtInfoHandler::GetOutputShapeAndType(uint32_t output_index, GeShape &shape, DataType &data_type) {
  175. GE_CHK_BOOL_RET_STATUS((unknown_type_ != DEPEND_COMPUTE), INTERNAL_ERROR,
  176. "Node[%s] is depend compute type can not get output shape and type by ext.",
  177. node_name_.c_str());
  178. GetShapeAndType(output_shape_and_type_[output_index], shape, data_type);
  179. return SUCCESS;
  180. }
  181. Status AicpuExtInfoHandler::UpdateShapeAndType(const GeShape &shape, DataType data_type,
  182. AicpuShapeAndType *shape_and_type) {
  183. auto dim_num = shape.GetDimNum();
  184. if (dim_num > aicpu::FWKAdapter::kMaxShapeDims) {
  185. GELOGE(ACL_ERROR_GE_PARAM_INVALID, "Update shape and type failed, as dim_num %zu is over max shape dims %u.",
  186. dim_num, aicpu::FWKAdapter::kMaxShapeDims);
  187. return ACL_ERROR_GE_PARAM_INVALID;
  188. }
  189. size_t index = 0;
  190. for (; index < dim_num; ++index) {
  191. shape_and_type->dims[index] = shape.GetDim(index);
  192. }
  193. if (index < aicpu::FWKAdapter::kMaxShapeDims) {
  194. shape_and_type->dims[index] = kDimEndFlag;
  195. }
  196. // now only support update shape, type is not support
  197. return SUCCESS;
  198. }
  199. void AicpuExtInfoHandler::GetShapeAndType(const AicpuShapeAndType *shape_and_type,
  200. GeShape &shape,
  201. DataType &data_type) {
  202. std::vector<int64_t> dims;
  203. for (uint32_t index = 0; index < aicpu::FWKAdapter::kMaxShapeDims; ++index) {
  204. auto tmpDim = shape_and_type->dims[index];
  205. if (tmpDim == kDimEndFlag) {
  206. break;
  207. }
  208. dims.emplace_back(tmpDim);
  209. }
  210. data_type = static_cast<DataType>(shape_and_type->type);
  211. shape = GeShape(dims);
  212. }
  213. } // namespace hybrid
  214. } // namespace ge

图引擎模块(GE)是MindSpore的一个子模块,其代码由C++实现,位于前端模块ME和底层硬件之间,起到承接作用。图引擎模块以ME下发的图作为输入,然后进行一系列的深度图优化操作,最后输出一张可以在底层硬件上高效运行的图。GE针对昇腾AI处理器的硬件结构特点,做了特定的优化工作,以此来充分发挥出昇腾AI处理器的强大算力。在进行模型训练/推理时,GE会被自动调用而用户并不感知。GE主要由GE API和GE Core两部分组成,详细的架构图如下所示