!871 rm compile macros

From: @chen_yemeng
Reviewed-by: 
Signed-off-by:

CMakeLists.txt

@@ -72,7 +72,7 @@ if (ENABLE_OPEN_SRC)
         endif()
         set(GE_LIB_PATH ${GE_LIB_PATH}/${GE_SYS_ARCH})
         set(STATIC_ACL_LIB ${GE_LIB_PATH})
-        find_module(slog libslog.so ${GE_LIB_PATH})
+        find_module(slog libalog.so ${GE_LIB_PATH})
         find_module(static_mmpa libmmpa.a ${GE_LIB_PATH})
         find_module(msprofiler_ext libmsprofiler.a ${GE_LIB_PATH})
         find_module(hccl libhccl.so ${GE_LIB_PATH})
@@ -88,7 +88,7 @@ if (ENABLE_OPEN_SRC)
     elseif(ENABLE_GE_COV OR ENABLE_GE_UT)
 	add_subdirectory(tests)
     else()
-        find_module(slog libslog.so ${ASCEND_ATC_DIR})
+        find_module(slog libalog.so ${ASCEND_ATC_DIR})
         find_module(static_mmpa libmmpa.a ${ASCEND_ATC_DIR})
 	find_module(error_manager liberror_manager.so ${ASCEND_ATC_DIR})
         if(PLATFORM STREQUAL "train")
@@ -154,7 +154,7 @@ elseif (ENABLE_D OR ENABLE_ACL)
     include(cmake/intf_pub_linux.cmake)
 
     # common libraries
-    find_module(slog libslog.so ${ASCEND_MS_DRIVER_PATH})
+    find_module(slog libalog.so ${ASCEND_MS_DRIVER_PATH})
     find_module(error_manager liberror_manager.so ${ASCEND_MS_RUNTIME_PATH} ${ATLAS_MS_RUNTIME_PATH})
     find_module(static_mmpa libmmpa.a ${ASCEND_MS_RUNTIME_PATH} ${ATLAS_MS_RUNTIME_PATH})
 
@@ -174,7 +174,7 @@ elseif(ENABLE_MS_TESTCASES)
     include(cmake/intf_pub_linux.cmake)
 
     # common libraries
-    find_module(slog libslog.so ${ASCEND_MS_DRIVER_PATH})
+    find_module(slog libalog.so ${ASCEND_MS_DRIVER_PATH})
     find_module(error_manager liberror_manager.so ${ASCEND_MS_RUNTIME_PATH} ${ATLAS_MS_RUNTIME_PATH})
     find_module(static_mmpa libmmpa.a ${ASCEND_MS_RUNTIME_PATH} ${ATLAS_MS_RUNTIME_PATH})
 

cmake/intf_pub_linux.cmake

@@ -16,6 +16,7 @@ target_compile_definitions(intf_pub INTERFACE
     $<$<CONFIG:Debug>:CFG_BUILD_DEBUG>   
     WIN64=1
     LINUX=0
+    LOG_CPP
 )
 target_link_options(intf_pub INTERFACE
     -Wl,-z,relro

ge/CMakeLists.txt

@@ -124,7 +124,7 @@ set(TRAIN_SRC_LIST
     "graph/manager/graph_var_manager.cc"
     "graph/manager/host_mem_manager.cc"
     "graph/manager/rdma_pool_allocator.cc"
-    $<$<NOT:$<STREQUAL:${ENABLE_OPEN_SRC},True>>:graph/manager/host_mem_allocator.cc>
+    "graph/manager/host_mem_allocator.cc"
     "graph/manager/memory_api.cc"
     "graph/manager/model_manager/event_manager.cc"
     "graph/manager/trans_var_data_utils.cc"
@@ -166,7 +166,7 @@ set(TRAIN_SRC_LIST
     "graph/passes/hccl_group_pass.cc"
     "graph/passes/enter_pass.cc"
     "graph/passes/assign_remove_pass.cc"
-    $<$<NOT:$<STREQUAL:${ENABLE_OPEN_SRC},True>>:graph/passes/inplace_support_check_pass.cc>
+    "graph/passes/inplace_support_check_pass.cc"
     "graph/passes/flow_ctrl_pass.cc"
     "graph/passes/global_step_insert_pass.cc"
     "host_kernels/transpose_kernel.cc"
@@ -409,7 +409,7 @@ set(INFER_SRC_LIST
     "graph/manager/graph_var_manager.cc"
     "graph/manager/host_mem_manager.cc"
     "graph/manager/rdma_pool_allocator.cc"
-    $<$<NOT:$<STREQUAL:${ENABLE_OPEN_SRC},True>>:graph/manager/host_mem_allocator.cc>
+    "graph/manager/host_mem_allocator.cc"
     "graph/manager/graph_mem_allocator.cc"
     "graph/manager/graph_caching_allocator.cc"
     "model/ge_model.cc"
@@ -531,7 +531,7 @@ set(INFER_SRC_LIST
     "graph/passes/for_pass.cc"
     "graph/passes/enter_pass.cc"
     "graph/passes/assign_remove_pass.cc"
-    $<$<NOT:$<STREQUAL:${ENABLE_OPEN_SRC},True>>:graph/passes/inplace_support_check_pass.cc>
+    "graph/passes/inplace_support_check_pass.cc"
     "graph/passes/addn_pass.cc"
     "graph/passes/common_subexpression_elimination_pass.cc"
     "graph/passes/remove_same_const_pass.cc"
@@ -654,7 +654,6 @@ target_compile_definitions(ge_runner PRIVATE
     FMK_SUPPORT_DUMP
     DAVINCI_CLOUD
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_options(ge_runner PRIVATE
@@ -718,7 +717,6 @@ target_compile_definitions(ge_compiler PRIVATE
     FMK_HOST_INFER
     COMPILE_OMG_PACKAGE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_options(ge_compiler PRIVATE
@@ -806,7 +804,6 @@ endif()
 
 target_compile_definitions(opensrc_ascendcl PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_options(opensrc_ascendcl PRIVATE

ge/client/ge_api.cc

@@ -32,9 +32,7 @@
 #include "graph/common/ge_call_wrapper.h"
 #include "register/op_registry.h"
 #include "common/ge/tbe_plugin_manager.h"
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "toolchain/plog.h"
-#endif
 
 using domi::OpRegistry;
 using std::map;
@@ -132,11 +130,9 @@ Status GEInitializeImpl(const std::map<string, string> &options) {
 
 // Initialize GE, prepare for execution, call GELib::Initialize
 Status GEInitialize(const std::map<string, string> &options) {
-#ifndef ONLY_COMPILE_OPEN_SRC
   if (DlogReportInitialize() != SUCCESS) {
     GELOGW("Dlog report device log initialize failed.");
   }
-#endif
   return GEInitializeImpl(options);
 }
 
@@ -151,11 +147,9 @@ Status GEInitialize(const std::map<AscendString, AscendString> &options) {
     std::string val = option.second.GetString();
     str_options[key] = val;
   }
-#ifndef ONLY_COMPILE_OPEN_SRC
   if (DlogReportInitialize() != SUCCESS) {
     GELOGW("Dlog report device log initialize failed.");
   }
-#endif
   return GEInitializeImpl(str_options);
 }
 
@@ -200,11 +194,9 @@ Status GEFinalize() {
   // to avoid memory fragment, use malloc_trim to back free stack to system
   malloc_trim(0);
 
-#ifndef ONLY_COMPILE_OPEN_SRC
   if (DlogReportFinalize() != SUCCESS) {
     GELOGW("Dlog report device log finalize failed.");
   }
-#endif
 
   GELOGT(TRACE_STOP, "GEFinalize finished");
   return ret;

ge/common/CMakeLists.txt

@@ -73,7 +73,6 @@ target_compile_definitions(ge_common PRIVATE
     FMK_SUPPORT_DUMP
     OS_CENTOS
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_options(ge_common PRIVATE
@@ -133,7 +132,6 @@ target_compile_definitions(ge_common_static PRIVATE
     $<IF:$<STREQUAL:${TARGET_SYSTEM_NAME},Windows>,OS_TYPE=WIN,OS_TYPE=0>
     $<$<STREQUAL:${TARGET_SYSTEM_NAME},Windows>:SECUREC_USING_STD_SECURE_LIB=0 NOMINMAX>
     LOG_CPP
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_options(ge_common_static PRIVATE
@@ -182,7 +180,6 @@ target_compile_definitions(ge_common PRIVATE
     FMK_SUPPORT_DUMP
     OS_CENTOS
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_options(ge_common PRIVATE

ge/executor/CMakeLists.txt

@@ -28,7 +28,7 @@ set(SRC_LIST
     "../graph/manager/trans_var_data_utils.cc"
     "../graph/manager/util/debug.cc"
     "../graph/manager/rdma_pool_allocator.cc"
-    $<$<NOT:$<STREQUAL:${ENABLE_OPEN_SRC},True>>:../graph/manager/host_mem_allocator.cc>
+    "../graph/manager/host_mem_allocator.cc"
     "../hybrid/node_executor/aicpu/aicpu_ext_info.cc"
     "../model/ge_model.cc"
     "../model/ge_root_model.cc"
@@ -175,7 +175,6 @@ target_compile_definitions(ge_executor PRIVATE
     $<IF:$<STREQUAL:${TARGET_SYSTEM_NAME},Windows>,OS_TYPE=WIN,OS_TYPE=0>
     $<$<STREQUAL:${TARGET_SYSTEM_NAME},Windows>:SECUREC_USING_STD_SECURE_LIB=0 NOMINMAX>
     LOG_CPP
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(ge_executor PRIVATE
@@ -218,7 +217,6 @@ target_compile_definitions(ge_executor_shared PRIVATE
     PROTOBUF_INLINE_NOT_IN_HEADERS=0
     DAVINCI_SUPPORT_PROFILING
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(ge_executor_shared PRIVATE

ge/ge_local_engine/CMakeLists.txt

@@ -31,7 +31,6 @@ target_compile_options(ge_local_engine PRIVATE
 
 target_compile_definitions(ge_local_engine PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(ge_local_engine PRIVATE
@@ -73,7 +72,6 @@ target_compile_options(atc_ge_local_engine PRIVATE
 target_compile_definitions(atc_ge_local_engine PRIVATE
     COMPILE_OMG_PACKAGE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(atc_ge_local_engine PRIVATE
@@ -119,7 +117,6 @@ target_compile_options(ge_local_opskernel_builder PRIVATE
 
 target_compile_definitions(ge_local_opskernel_builder PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(ge_local_opskernel_builder PRIVATE
@@ -161,7 +158,6 @@ target_compile_options(atc_ge_local_opskernel_builder PRIVATE
 
 target_compile_definitions(atc_ge_local_opskernel_builder PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(atc_ge_local_opskernel_builder PRIVATE
@@ -209,7 +205,6 @@ target_compile_options(ge_local_opskernel_builder_static PRIVATE
 target_compile_definitions(ge_local_opskernel_builder_static PRIVATE
     google=ascend_private
     LOG_CPP
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(ge_local_opskernel_builder_static PRIVATE

ge/ge_local_engine/engine/host_cpu_engine.cc

@@ -26,7 +26,6 @@
 #include "common/math/math_util.h"
 
 namespace {
-#ifndef ONLY_COMPILE_OPEN_SRC
 #define CREATE_OUTPUT_CASE(DTYPE, TYPE)                                                                                \
   case (DTYPE): {                                                                                                      \
     GeTensorPtr ge_tensor = nullptr;                                                                                   \
@@ -50,43 +49,6 @@ namespace {
     named_outputs.emplace(tensor_name, tensor);                                                                        \
     break;                                                                                                             \
   }
-#else
-#define CREATE_OUTPUT_CASE(DTYPE, TYPE)                                                                                \
-  case (DTYPE): {                                                                                                      \
-    GeTensorPtr ge_tensor = nullptr;                                                                                   \
-    if (need_create_flag) {                                                                                            \
-      GELOGI("node:%s allocate output %zu start, size=%lld", op_desc->GetName().c_str(), i, data_num * sizeof(TYPE));  \
-      std::unique_ptr<TYPE[]> buf(new (std::nothrow) TYPE[data_num]());                                                \
-      if (buf == nullptr) {                                                                                            \
-        GELOGE(MEMALLOC_FAILED, "New sizeof(T) * data_num(%zu) memory failed",                                         \
-               static_cast<size_t>(sizeof(TYPE) * data_num));                                                          \
-        return MEMALLOC_FAILED;                                                                                        \
-      }                                                                                                                \
-      ge_tensor = MakeShared<GeTensor>(out_desc);                                                                      \
-      GE_CHECK_NOTNULL(ge_tensor);                                                                                     \
-      GELOGD("node:%s allocate output %zu success, size=%lld", op_desc->GetName().c_str(), i, data_num * sizeof(TYPE));\
-      if (ge_tensor->SetData(reinterpret_cast<uint8_t *>(buf.get()), data_num * sizeof(TYPE)) != GRAPH_SUCCESS) {      \
-        GELOGE(MEMALLOC_FAILED, "Set data for output %zu of node %s failed.", i, op_desc->GetName().c_str());          \
-        return MEMALLOC_FAILED;                                                                                        \
-      }                                                                                                                \
-      ge_tensor->MutableTensorDesc().SetDataType(out_desc.GetDataType());                                              \
-      ge_tensor->MutableTensorDesc().SetShape(out_desc.GetShape());                                                    \
-      outputs.emplace_back(ge_tensor);                                                                                 \
-    } else {                                                                                                           \
-      ge_tensor = outputs[i];                                                                                          \
-      GE_CHECK_NOTNULL(ge_tensor);                                                                                     \
-      GELOGD("node:%s existed output %zu", op_desc->GetName().c_str(), i);                                             \
-    }                                                                                                                  \
-    auto tensor = TensorAdapter::AsTensor(*ge_tensor);                                                                 \
-    auto tensor_name = op_desc->GetOutputNameByIndex(i);                                                               \
-    GE_RETURN_WITH_LOG_IF_TRUE(tensor_name.empty(), "Failed to get output name. node = %s, index = %zu",               \
-                               op_desc->GetName().c_str(), i);                                                         \
-    GELOGD("Successfully inserted output tensor. node = %s, index = %zu, output name = %s, addr = %p, size = %zu",     \
-           op_desc->GetName().c_str(), i, tensor_name.c_str(), tensor.GetData(), tensor.GetSize());                    \
-    named_outputs.emplace(tensor_name, tensor);                                                                        \
-    break;                                                                                                             \
-  }
-#endif
 }
 
 namespace ge {

ge/ge_runtime/CMakeLists.txt

@@ -27,7 +27,6 @@ target_compile_options(ge_runtime PRIVATE
 
 target_compile_definitions(ge_runtime PRIVATE 
     PROTOBUF_INLINE_NOT_IN_HEADERS=0
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(ge_runtime PRIVATE

ge/graph/manager/graph_manager.cc

@@ -38,10 +38,8 @@
 #include "graph/partition/stage_partition.h"
 #include "graph/passes/addn_pass.h"
 #include "graph/passes/bitcast_pass.h"
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "graph/passes/assign_remove_pass.h"
 #include "graph/passes/inplace_support_check_pass.h"
-#endif
 #include "graph/passes/atomic_addr_clean_pass.h"
 #include "graph/passes/attach_stream_label_pass.h"
 #include "graph/passes/cast_remove_pass.h"
@@ -2269,20 +2267,16 @@ Status GraphManager::OptimizeStage2(ge::ComputeGraphPtr &compute_graph) {
   ReshapeRemovePass reshape_remove_pass;
   CondRemovePass condition_remove_pass;
   BitcastPass bitcast_pass;
-#ifndef ONLY_COMPILE_OPEN_SRC
   AssignRemovePass assign_remove_pass;
   InplaceSupportCheckPass inplace_support_check_pass;
-#endif
   names_to_passes.emplace_back("ConstantFoldingPass", &constant_folding_pass);
   names_to_passes.emplace_back("ReshapeRemovePass", &reshape_remove_pass);
   names_to_passes.emplace_back("CondRemovePass", &condition_remove_pass);
   names_to_passes.emplace_back("BitcastPass", &bitcast_pass);
-#ifndef ONLY_COMPILE_OPEN_SRC
   if (GetContext().GetHostExecFlag()) {
     names_to_passes.emplace_back("AssignRemovePass", &assign_remove_pass);
     names_to_passes.emplace_back("InplaceSupportCheckPass", &inplace_support_check_pass);
   }
-#endif
   GE_TIMESTAMP_START(names_to_passes);
   ret = GEPass(compute_graph).Run(names_to_passes);
   GE_TIMESTAMP_END(names_to_passes, "OptimizeStage2::MergedGraphNameToPasses");

ge/graph/manager/graph_mem_allocator.cc

@@ -19,9 +19,7 @@
 #include <string>
 #include "graph/manager/graph_caching_allocator.h"
 #include "graph/manager/rdma_pool_allocator.h"
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "graph/manager/host_mem_allocator.h"
-#endif
 namespace ge {
 void MemoryAllocator::Initialize(uint32_t device_id) {
   GELOGI("MemoryAllocator::Initialize");
@@ -192,12 +190,10 @@ Status MemManager::Initialize(const std::vector<rtMemType_t> &memory_type) {
     GELOGE(ge::INTERNAL_ERROR, "Create RdmaAllocator failed.");
     return ge::INTERNAL_ERROR;
   }
-#ifndef ONLY_COMPILE_OPEN_SRC
   if (InitAllocator(memory_type, host_allocator_map_) != SUCCESS) {
     GELOGE(ge::INTERNAL_ERROR, "Create HostMemAllocator failed.");
     return ge::INTERNAL_ERROR;
   }
-#endif
   return SUCCESS;
 }
 
@@ -219,9 +215,7 @@ void MemManager::Finalize() noexcept {
   // caching and rdma allocator use memory allocator, so finalize them first
   FinalizeAllocatorMap(caching_allocator_map_);
   FinalizeAllocatorMap(rdma_allocator_map_);
-#ifndef ONLY_COMPILE_OPEN_SRC
   FinalizeAllocatorMap(host_allocator_map_);
-#endif
   FinalizeAllocatorMap(memory_allocator_map_);
 }
 
@@ -250,9 +244,7 @@ CachingAllocator &MemManager::CachingInstance(rtMemType_t memory_type) {
 RdmaPoolAllocator &MemManager::RdmaPoolInstance(rtMemType_t memory_type) {
   return Instance().GetAllocator(memory_type, rdma_allocator_map_);
 }
-#ifndef ONLY_COMPILE_OPEN_SRC
 HostMemAllocator &MemManager::HostMemInstance(rtMemType_t memory_type) {
   return Instance().GetAllocator(memory_type, host_allocator_map_);
 }
-#endif
 }  // namespace ge

ge/graph/manager/graph_mem_allocator.h

@@ -139,9 +139,7 @@ class MemoryAllocator {
 using MemoryAllocatorPtr = std::shared_ptr<MemoryAllocator>;
 class CachingAllocator;
 class RdmaPoolAllocator;
-#ifndef ONLY_COMPILE_OPEN_SRC
 class HostMemAllocator;
-#endif
 class MemManager {
  public:
   MemManager();
@@ -150,9 +148,7 @@ class MemManager {
   static MemoryAllocator *Instance(rtMemType_t memory_type);
   CachingAllocator &CachingInstance(rtMemType_t memory_type);
   RdmaPoolAllocator &RdmaPoolInstance(rtMemType_t memory_type);
-#ifndef ONLY_COMPILE_OPEN_SRC
   HostMemAllocator &HostMemInstance(rtMemType_t memory_type);
-#endif
   MemManager(const MemManager &) = delete;
   MemManager &operator=(const MemManager &) = delete;
   ///
@@ -240,9 +236,7 @@ class MemManager {
   std::map<rtMemType_t, MemoryAllocator *> memory_allocator_map_;
   std::map<rtMemType_t, CachingAllocator *> caching_allocator_map_;
   std::map<rtMemType_t, RdmaPoolAllocator *> rdma_allocator_map_;
-#ifndef ONLY_COMPILE_OPEN_SRC
   std::map<rtMemType_t, HostMemAllocator *> host_allocator_map_;
-#endif
   std::recursive_mutex allocator_mutex_;
 };
 }  // namespace ge

ge/graph/manager/host_mem_allocator.h

@@ -27,7 +27,7 @@
 namespace ge {
 class HostMemAllocator {
  public:
-  explicit HostMemAllocator(rtMemType_t)  {}
+  explicit HostMemAllocator(rtMemType_t) {}
   ~HostMemAllocator() = default;
 
   HostMemAllocator(const HostMemAllocator &) = delete;

ge/graph/manager/host_mem_manager.cc

@@ -43,29 +43,20 @@ Status SharedMemAllocator::Allocate(SharedMemInfo &mem_info) {
     return GE_GRAPH_MEMORY_ALLOC_FAILED;
   }
   mem_info.fd = output_para.fd;
-#ifndef ONLY_COMPILE_OPEN_SRC
   mem_info.host_aligned_ptr = AlignedPtr::BuildFromAllocFunc([&output_para](std::unique_ptr<uint8_t[], deleter> &ptr) {
                                                                ptr.reset(reinterpret_cast<uint8_t *>(output_para.ptr));
                                                              },
                                                              [](uint8_t *ptr) {
                                                                ptr = nullptr;
                                                              });
-#else
-  mem_info.host_address = reinterpret_cast<uint8_t *>(output_para.ptr);
-#endif
   mem_info.device_address = reinterpret_cast<uint8_t *>(output_para.devPtr);
   return SUCCESS;
 }
 
 Status SharedMemAllocator::DeAllocate(SharedMemInfo &mem_info) {
   GELOGD("SharedMemAllocator::DeAllocate");
-#ifndef ONLY_COMPILE_OPEN_SRC
   rtFreeHostSharedMemoryIn free_para = {mem_info.shm_name.c_str(), mem_info.mem_size, mem_info.fd,
                                         mem_info.host_aligned_ptr->MutableGet(), mem_info.device_address};
-#else
-  rtFreeHostSharedMemoryIn free_para = {mem_info.shm_name.c_str(), mem_info.mem_size, mem_info.fd,
-                                        mem_info.host_address, mem_info.device_address};
-#endif
   rtError_t rt_ret = rtFreeHostSharedMemory(&free_para);
   if (rt_ret != RT_ERROR_NONE) {
     GELOGE(RT_FAILED, "Call rt api(rtFreeHostSharedMemory) failed, ret: 0x%X.", rt_ret);

ge/graph/manager/host_mem_manager.h

@@ -42,11 +42,7 @@ struct SharedMemInfo {
   uint64_t mem_size = 0;
   int fd = 0;
   uint8_t *device_address = nullptr;
-#ifndef ONLY_COMPILE_OPEN_SRC
   std::shared_ptr<AlignedPtr> host_aligned_ptr = nullptr;
-#else
-  uint8_t *host_address = nullptr;
-#endif
   SharedMemInfo() = default;
   SharedMemInfo(string name, uint64_t size) : op_name(std::move(name)), mem_size(size) {}
 };

ge/graph/optimize/graph_optimize.cc

@@ -127,6 +127,10 @@ Status GraphOptimize::OptimizeSubGraph(ComputeGraphPtr &compute_graph, const std
 }
 
 Status GraphOptimize::OptimizeOriginalGraph(ComputeGraphPtr &compute_graph) {
+  if (GetContext().GetHostExecFlag()) {
+    // graph exec on host, no need OptimizeOriginalGraph
+    return SUCCESS;
+  }
   if (compute_graph == nullptr) {
     GELOGE(GE_GRAPH_OPTIMIZE_COMPUTE_GRAPH_NULL, "[OptimizeOriginalGraph]: compute_graph is nullptr.");
     return GE_GRAPH_OPTIMIZE_COMPUTE_GRAPH_NULL;
@@ -162,7 +166,7 @@ Status GraphOptimize::OptimizeOriginalGraph(ComputeGraphPtr &compute_graph) {
 Status GraphOptimize::OptimizeOriginalGraphJudgeInsert(ComputeGraphPtr &compute_graph) {
   GELOGD("OptimizeOriginalGraphJudgeInsert in");
   if (GetContext().GetHostExecFlag()) {
-    // graph exec on host, no need OptimizeOriginalGraph
+    // graph exec on host, no need OptimizeOriginalGraphJudgeInsert
     return SUCCESS;
   }
 

ge/graph/passes/assign_remove_pass.cc

@@ -19,6 +19,7 @@
 #include "graph/utils/graph_utils.h"
 #include "graph/debug/ge_attr_define.h"
 
+namespace ge {
 namespace {
 constexpr uint32_t kValidInputNodeOutputNum = 1;
 constexpr int32_t kAssignRefInputIndex = 0;
@@ -28,8 +29,6 @@ static const std::set<std::string> kNoTaskNodeTypes = { ge::DATA, ge::ANN_DATA,
                                                         ge::VARIABLE, ge::VARIABLEV2 };
 }
 
-namespace ge {
-#ifndef ONLY_COMPILE_OPEN_SRC
 Status AssignRemovePass::Run(NodePtr &node) {
   GELOGD("AssignRemovePass running");
 
@@ -145,71 +144,7 @@ Status AssignRemovePass::TransformAttr(NodePtr &node) {
   }
   return SUCCESS;
 }
-#else
-Status AssignRemovePass::Run(NodePtr &node) {
-  GELOGD("AssignRemovePass running");
-  if (node->GetType() != ASSIGN) {
-    GELOGD("No need run AssignRemovePass on [%s, %s].", node->GetName().c_str(), node->GetType().c_str());
-    return SUCCESS;
-  }
-
-  const auto &ref_in_anchor = node->GetInDataAnchor(kAssignRefInputIndex);
-  const auto &value_in_anchor = node->GetInDataAnchor(kAssignValueInputIndex);
-  if ((ref_in_anchor == nullptr) || (value_in_anchor == nullptr)) {
-    GELOGE(FAILED, "In data anchor is null, node:%s", node->GetName().c_str());
-    return FAILED;
-  }
-  const auto &ref_peer_anchor = ref_in_anchor->GetPeerOutAnchor();
-  const auto &value_peer_anchor = value_in_anchor->GetPeerOutAnchor();
-  if ((ref_peer_anchor == nullptr) || (value_peer_anchor == nullptr)) {
-    GELOGE(FAILED, "Peer data anchor is null, node:%s", node->GetName().c_str());
-    return FAILED;
-  }
-
-  if (IsCondMatch(node, ref_peer_anchor, value_peer_anchor)) {
-    ///
-    ///    variable  not-const               not-const
-    ///         \     /                          |
-    ///          \   /                           |
-    ///         Assign           ---->        variable
-    ///           |                              |
-    ///           |                              |
-    ///         node                           node
-    ///
-    GELOGI("Optimization for assign_node %s start", node->GetName().c_str());
-    if (IsolateAndDeleteNode(node, {kAssignRefInputIndex}) != SUCCESS) {
-      GELOGE(FAILED, "Isolate and delete assign_node %s failed.", node->GetName().c_str());
-      return FAILED;
-    }
-    AddNodeDeleted(node);
 
-    const auto &ref_input = ref_peer_anchor->GetOwnerNode()->GetOpDesc();
-    const auto &value_input = value_peer_anchor->GetOwnerNode()->GetOpDesc();
-    if ((ref_input == nullptr) || (value_input == nullptr)) {
-      GELOGE(FAILED, "value input is null");
-      return FAILED;
-    }
-    if (!AttrUtils::SetStr(value_input->MutableOutputDesc(value_peer_anchor->GetIdx()), ASSIGN_VAR_NAME,
-                           ref_input->GetName())) {
-      GELOGE(FAILED, "Set attr ASSIGN_VAR_NAME failed.");
-      return FAILED;
-    }
-
-    // variable has and only has one input
-    if (ref_input->UpdateInputDesc(0, value_input->GetOutputDesc(value_peer_anchor->GetIdx())) != GRAPH_SUCCESS) {
-      GELOGE(FAILED, "Update input_desc for variable %s failed.", ref_input->GetName().c_str());
-      return FAILED;
-    }
-    if (GraphUtils::AddEdge(value_peer_anchor, ref_peer_anchor->GetOwnerNode()->GetInDataAnchor(0)) != GRAPH_SUCCESS) {
-      GELOGE(FAILED, "Add data edge %s->%s failed", value_input->GetName().c_str(), ref_input->GetName().c_str());
-      return FAILED;
-    }
-  }
-
-  GELOGD("AssignRemovePass success");
-  return SUCCESS;
-}
-#endif
 ///
 /// @brief Check if need optimize for assign_node
 /// @param [in] assign_node
@@ -218,7 +153,7 @@ Status AssignRemovePass::Run(NodePtr &node) {
 /// @return Status
 ///
 bool AssignRemovePass::IsCondMatch(const NodePtr &node, const OutDataAnchorPtr &ref_peer_anchor,
-                             const OutDataAnchorPtr &value_peer_anchor) {
+                                   const OutDataAnchorPtr &value_peer_anchor) {
   GELOGD("Check if assign_node %s match optimization condition, ref_input: %s, value_input: %s",
          node->GetName().c_str(), ref_peer_anchor->GetOwnerNode()->GetName().c_str(),
          value_peer_anchor->GetOwnerNode()->GetName().c_str());

ge/graph/passes/assign_remove_pass.h

@@ -25,7 +25,6 @@ class AssignRemovePass : public BaseNodePass {
   Status Run(NodePtr &node) override;
 
  private:
-#ifndef ONLY_COMPILE_OPEN_SRC
   ///
   /// @brief Optimize for assign_node
   /// @param [in] assign_node
@@ -39,7 +38,7 @@ class AssignRemovePass : public BaseNodePass {
   /// @return Status
   ///
   Status TransformAttr(NodePtr &node);
-#endif
+
   ///
   /// @brief Check if need optimize for assign_node
   /// @param [in] assign_node

ge/graph/passes/constant_fuse_same_pass.cc

@@ -115,21 +115,15 @@ void ConstantFuseSamePass::GetFuseConstNodes(ComputeGraphPtr &graph,
              TypeUtils::DataTypeToSerialString(data_type).c_str());
       continue;
     }
-#ifndef ONLY_COMPILE_OPEN_SRC
     if ((type_size != 0) && (weight->MutableData().GetAlignedPtr() == nullptr)) {
       GELOGW("aligned_ptr is null while size is not 0");
       continue;
     }
-#endif
     ++insert_const_nums;
 
     SameConstKey map_key;
     map_key.data_size = type_size;
-#ifndef ONLY_COMPILE_OPEN_SRC
     map_key.aligned_ptr = weight->MutableData().GetAlignedPtr();
-#else
-    map_key.data = weight->GetData().GetData();
-#endif
     map_key.data_type = data_type;
     map_key.format = output_tensor->GetFormat();
     map_key.shape = output_tensor->GetShape().GetDims();

ge/graph/passes/constant_fuse_same_pass.h

@@ -21,20 +21,14 @@
 #include <set>
 #include <utility>
 #include <vector>
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "graph/aligned_ptr.h"
-#endif
 #include "graph/types.h"
 #include "inc/graph_pass.h"
 
 namespace ge {
 struct SameConstKey {
   int data_size;
-#ifndef ONLY_COMPILE_OPEN_SRC
   std::shared_ptr<AlignedPtr> aligned_ptr;
-#else
-  const uint8_t *data;
-#endif
   DataType data_type;
   Format format;
   std::vector<int64_t> shape;
@@ -44,19 +38,12 @@ struct SameConstKey {
     if (data_size != key.data_size) {
       return data_size < key.data_size;
     }
-#ifndef ONLY_COMPILE_OPEN_SRC
     if (data_size != 0) {
       int ret = memcmp(aligned_ptr->Get(), key.aligned_ptr->Get(), data_size);
       if (ret != 0) {
         return ret < 0;
       }
     }
-#else
-    int ret = memcmp(data, key.data, data_size);
-    if (ret != 0) {
-      return ret < 0;
-    }
-#endif
     if (data_type != key.data_type) {
       return data_type < key.data_type;
     }

ge/graph/passes/inplace_support_check_pass.cc

@@ -19,6 +19,7 @@
 #include "graph/utils/graph_utils.h"
 #include "graph/debug/ge_attr_define.h"
 
+namespace ge {
 namespace {
 constexpr uint32_t kInplaceSupportOutputIndex = 0;
 constexpr uint32_t kInplaceSupportOutputNum = 1;
@@ -26,8 +27,6 @@ static const std::set<std::string> kSrcNodeTypes = { ge::DATA, ge::ANN_DATA, ge:
                                                      ge::CONSTANT, ge::CONSTANTOP,
                                                      ge::VARIABLE, ge::VARIABLEV2 };
 }
-
-namespace ge {
 Status InplaceSupportCheckPass::Run(NodePtr &node) {
   GELOGD("InplaceSupportCheckPass running");
   if (node->GetAllOutDataAnchorsSize() != kInplaceSupportOutputNum) {

ge/graph/preprocess/graph_preprocess.cc

@@ -38,9 +38,6 @@
 #include "graph/passes/aicpu_constant_folding_pass.h"
 #include "graph/passes/assert_pass.h"
 #include "ge/ge_api_types.h"
-#ifdef ONLY_COMPILE_OPEN_SRC
-#include "graph/passes/assign_remove_pass.h"
-#endif
 #include "graph/passes/common_subexpression_elimination_pass.h"
 #include "graph/passes/cond_pass.h"
 #include "graph/passes/cond_remove_pass.h"
@@ -1865,9 +1862,6 @@ Status GraphPrepare::PrepareOptimize() {
   VarIsInitializedOpPass var_is_initialized_pass;
   ParallelConcatStartOpPass parallel_concat_start_op_pass;
   IdentityPass identity_pass(false);
-#ifdef ONLY_COMPILE_OPEN_SRC
-  AssignRemovePass assign_remove_pass;
-#endif
   SnapshotPass snapshot_pass;
   if (!options_.train_graph_flag) {
     names_to_passes.emplace_back("DropOutPass", &dropout_pass);
@@ -1882,11 +1876,6 @@ Status GraphPrepare::PrepareOptimize() {
   names_to_passes.emplace_back("VarIsInitializedOpPass", &var_is_initialized_pass);
   names_to_passes.emplace_back("ParallelConcatStartOpPass", &parallel_concat_start_op_pass);
   names_to_passes.emplace_back("IdentityPass", &identity_pass);
-#ifdef ONLY_COMPILE_OPEN_SRC
-  if (GetContext().GetHostExecFlag()) {
-    names_to_passes.emplace_back("AssignRemovePass", &assign_remove_pass);
-  }
-#endif
   GE_TIMESTAMP_START(names_to_passes);
   ret = ge_passes.Run(names_to_passes);
   GE_TIMESTAMP_END(names_to_passes, "GraphPrepare::NamesToPasses");

ge/host_cpu_engine/CMakeLists.txt

@@ -25,7 +25,6 @@ target_compile_options(host_cpu_engine PRIVATE
 
 target_compile_definitions(host_cpu_engine PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(host_cpu_engine PRIVATE
@@ -66,7 +65,6 @@ target_compile_options(atc_host_cpu_engine PRIVATE
 target_compile_definitions(atc_host_cpu_engine PRIVATE
     COMPILE_OMG_PACKAGE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(atc_host_cpu_engine PRIVATE
@@ -111,7 +109,6 @@ target_compile_options(host_cpu_opskernel_builder PRIVATE
 
 target_compile_definitions(host_cpu_opskernel_builder PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(host_cpu_opskernel_builder PRIVATE
@@ -152,7 +149,6 @@ target_compile_options(atc_host_cpu_opskernel_builder PRIVATE
 
 target_compile_definitions(atc_host_cpu_opskernel_builder PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(atc_host_cpu_opskernel_builder PRIVATE
@@ -199,7 +195,6 @@ target_compile_options(host_cpu_opskernel_builder_static PRIVATE
 target_compile_definitions(host_cpu_opskernel_builder_static PRIVATE
     google=ascend_private
     LOG_CPP
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(host_cpu_opskernel_builder_static PRIVATE

ge/hybrid/common/npu_memory_allocator.cc

@@ -20,9 +20,7 @@
 #include "graph/manager/graph_caching_allocator.h"
 #include "graph/manager/graph_mem_allocator.h"
 #include "graph/manager/rdma_pool_allocator.h"
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "graph/manager/host_mem_allocator.h"
-#endif
 
 namespace ge {
 namespace hybrid {
@@ -67,11 +65,7 @@ void *NpuMemoryAllocator::Allocate(std::size_t size, AllocationAttr *attr) {
   if (mem_type == RDMA_HBM) {
     buffer = MemManager::Instance().RdmaPoolInstance(RT_MEMORY_HBM).Malloc(allocate_size, device_id_);
   } else if (mem_type == HOST_DDR) {
-#ifndef ONLY_COMPILE_OPEN_SRC
     buffer = MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).Malloc(allocate_size);
-#else
-    buffer = malloc(allocate_size);
-#endif
   } else {
     if (allocate_size > kMaxHbmMemorySize) {
       GELOGE(PARAM_INVALID, "Invalid HBM memory size: %zu", allocate_size);
@@ -108,11 +102,7 @@ void NpuMemoryAllocator::Deallocate(void *data, MemStorageType mem_type) {
     if (mem_type == RDMA_HBM) {
       MemManager::Instance().RdmaPoolInstance(RT_MEMORY_HBM).Free(reinterpret_cast<uint8_t *>(data), device_id_);
     } else if (mem_type == HOST_DDR) {
-#ifndef ONLY_COMPILE_OPEN_SRC
       MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).Free(data);
-#else
-      free(data);
-#endif
     } else {
       MemManager::Instance().CachingInstance(RT_MEMORY_HBM).Free(reinterpret_cast<uint8_t *>(data), device_id_);
     }

ge/hybrid/model/hybrid_model_builder.cc

@@ -25,10 +25,8 @@
 #include "graph/manager/graph_var_manager.h"
 #include "graph/manager/host_mem_manager.h"
 #include "graph/manager/trans_var_data_utils.h"
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "graph/manager/graph_mem_allocator.h"
 #include "graph/manager/host_mem_allocator.h"
-#endif
 #include "graph/utils/graph_utils.h"
 #include "hybrid/common/npu_memory_allocator.h"
 #include "hybrid/node_executor/node_executor.h"
@@ -865,7 +863,6 @@ Status HybridModelBuilder::InitConstantOps() {
 
     std::unique_ptr<TensorValue> var_tensor;
     if (GetContext().GetHostExecFlag()) {
-#ifndef ONLY_COMPILE_OPEN_SRC
       GE_CHECK_NOTNULL(ge_tensor);
       // Address for eigen kernel should be aligned with 16 bytes
       // Tensors return by api GetWeights share data with proto, whose addr is not confirmed to be aligned
@@ -878,11 +875,6 @@ Status HybridModelBuilder::InitConstantOps() {
       }
       var_tensor.reset(new(std::nothrow)TensorValue(aligned_tensor.MutableData().data(),
                                                     aligned_tensor.GetData().size()));
-#else
-      auto buffer = ge_tensor->MutableData();
-      GELOGD("Init tensor with host constant. size = %zu", buffer.GetSize());
-      var_tensor.reset(new(std::nothrow)TensorValue(buffer.GetData(), buffer.GetSize()));
-#endif
     } else {
       GE_CHK_STATUS_RET_NOLOG(VarNodeToTensor(var_node, var_tensor));
       GELOGD("Init const op tensor. name = %s, size = %ld", var_name.c_str(), var_tensor->GetSize());
@@ -937,7 +929,6 @@ Status HybridModelBuilder::InitVariableTensors() {
       GELOGE(GE_GRAPH_MALLOC_FAILED, "Host variable [%s] malloc failed.", it.first.c_str());
       return GE_GRAPH_MALLOC_FAILED;
     }
-#ifndef ONLY_COMPILE_OPEN_SRC
     if (MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).Malloc(mem_info.host_aligned_ptr,
                                                                      tensor_size) == nullptr) {
       GELOGE(MEMALLOC_FAILED, "Malloc host memory for an existed GeTensor failed.");
@@ -947,11 +938,6 @@ Status HybridModelBuilder::InitVariableTensors() {
 
     std::unique_ptr<TensorValue> tensor(new (std::nothrow) TensorValue(mem_info.host_aligned_ptr->MutableGet(),
                                                                        tensor_size));
-#else
-    GELOGD("Host variable [%s] malloc success.", it.first.c_str());
-
-    std::unique_ptr<TensorValue> tensor(new (std::nothrow) TensorValue(mem_info.host_address, tensor_size));
-#endif
     GE_CHECK_NOTNULL(tensor);
     hybrid_model_.variable_tensors_.emplace(it.first, std::move(tensor));
   }

ge/hybrid/node_executor/host_cpu/host_cpu_node_executor.cc

@@ -18,10 +18,8 @@
 #include "hybrid/node_executor/host_cpu/kernel_factory.h"
 #include "graph/passes/folding_pass.h"
 #include "hybrid/model/hybrid_model.h"
-#ifndef ONLY_COMPILE_OPEN_SRC
 #include "graph/manager/graph_mem_allocator.h"
 #include "graph/manager/host_mem_allocator.h"
-#endif
 #include "ge_local_engine/engine/host_cpu_engine.h"
 
 namespace ge {
@@ -54,18 +52,11 @@ Status CpuKernelNodeTask::Execute(TaskContext &context) {
     auto input_desc_ptr = context.GetInputDesc(i);
     GE_CHECK_NOTNULL(input_desc_ptr);
     const auto &input_desc = *input_desc_ptr;
-#ifndef ONLY_COMPILE_OPEN_SRC
     auto tensor = context.GetInput(i);
     GE_CHECK_NOTNULL(tensor);
     auto item = MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).GetAlignedPtr(tensor->GetData());
     GE_CHECK_NOTNULL(item.second);
     auto in_tensor = MakeShared<GeTensor>(input_desc, item.second, item.first);
-#else
-    GE_CHECK_NOTNULL(context.GetInput(i));
-    auto in_tensor = MakeShared<GeTensor>(input_desc,
-                                          reinterpret_cast<const uint8_t *>(context.GetInput(i)->GetData()),
-                                          context.GetInput(i)->GetSize());
-#endif
     GE_CHECK_NOTNULL(in_tensor);
     in_tensor->MutableTensorDesc().SetDataType(input_desc.GetDataType());
     in_tensor->MutableTensorDesc().SetShape(input_desc.GetShape());
@@ -84,15 +75,9 @@ Status CpuKernelNodeTask::Execute(TaskContext &context) {
     }
     auto tensor = context.GetOutput(i);
     GE_CHECK_NOTNULL(tensor);
-#ifndef ONLY_COMPILE_OPEN_SRC
     auto item = MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).GetAlignedPtr(tensor->GetData());
     GE_CHECK_NOTNULL(item.second);
     auto out_tensor = MakeShared<GeTensor>(output_desc, item.second, item.first);
-#else
-    auto out_tensor = MakeShared<GeTensor>(output_desc,
-                                           reinterpret_cast<const uint8_t *>(tensor->GetData()),
-                                           tensor->GetSize());
-#endif
     GE_CHECK_NOTNULL(out_tensor);
     out_tensor->MutableTensorDesc().SetDataType(output_desc.GetDataType());
     out_tensor->MutableTensorDesc().SetShape(output_desc.GetShape());

ge/offline/CMakeLists.txt

@@ -30,7 +30,6 @@ target_compile_definitions(atc PRIVATE
     COMPILE_OMG_PACKAGE
     google=ascend_private
     LOG_CPP
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(atc PRIVATE
@@ -93,7 +92,6 @@ target_compile_definitions(atc_atc.bin PRIVATE
     COMPILE_OMG_PACKAGE
     google=ascend_private
     LOG_CPP
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(atc_atc.bin PRIVATE
@@ -154,7 +152,6 @@ target_compile_options(fwk_atc.bin PRIVATE
     -O2
     -Wno-deprecated-declarations
     -fno-common
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_compile_definitions(fwk_atc.bin PRIVATE

ge/plugin/engine/CMakeLists.txt

@@ -14,7 +14,6 @@ target_compile_options(engine PRIVATE
 target_compile_definitions(engine PRIVATE
     REUSE_MEMORY=1
     PROTOBUF_INLINE_NOT_IN_HEADERS=0
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_include_directories(engine PRIVATE

inc/framework/common/ge_types.h

@@ -37,10 +37,7 @@ enum FrameworkType {
   MINDSPORE = 1,
   TENSORFLOW = 3,
   ANDROID_NN,
-#ifndef ONLY_COMPILE_OPEN_SRC
   ONNX,
-#endif
-  FRAMEWORK_RESERVED,
 };
 
 enum OpEngineType {

inc/framework/omg/parser/model_parser.h

@@ -65,7 +65,6 @@ class ModelParser {
    */
   virtual Status ParseFromMemory(const char *data, uint32_t size, ge::ComputeGraphPtr &graph) = 0;
 
-#ifndef ONLY_COMPILE_OPEN_SRC
   /**
    * @ingroup domi_omg
    * @brief Parse relevant data from memory and save it to graph
@@ -77,7 +76,6 @@ class ModelParser {
    * @author
    */
   virtual Status ParseFromMemory(const char *data, uint32_t size, ge::Graph &graph) = 0;
-#endif
 
   /**
    * @ingroup domi_omg

metadef

@@ -1 +1 @@
-Subproject commit fe37bc343ea52c76d35e9e9ec83cea0151bfa900
+Subproject commit f08320a6d699f5b537bf66da572bf225b9cd330e

parser

@@ -1 +1 @@
-Subproject commit 336cd3107253d3fe41cfb9fec2db62b5f3d8a33b
+Subproject commit b2df31dc5810283e2e483df5ba9517e2ece132a0

tests/depends/cce/CMakeLists.txt

@@ -46,6 +46,7 @@ set(SRCS
     "${GE_CODE_DIR}/metadef/graph/anchor.cc"
     "${GE_CODE_DIR}/metadef/graph/ge_attr_value.cc"
     "${GE_CODE_DIR}/metadef/graph/buffer.cc"
+    "${GE_CODE_DIR}/metadef/graph/aligned_ptr.cc"
     "${GE_CODE_DIR}/metadef/graph/compute_graph.cc"
     "${GE_CODE_DIR}/metadef/graph/graph.cc"
     "${GE_CODE_DIR}/metadef/graph/model.cc"

tests/ut/common/graph/CMakeLists.txt

@@ -67,6 +67,7 @@ set(SRC_FILES
     "${GE_CODE_DIR}/metadef/graph/ge_attr_value.cc"
     "${GE_CODE_DIR}/metadef/graph/attr_value.cc"
     "${GE_CODE_DIR}/metadef/graph/buffer.cc"
+    "${GE_CODE_DIR}/metadef/graph/aligned_ptr.cc"
     "${GE_CODE_DIR}/metadef/graph/compute_graph.cc"
     "${GE_CODE_DIR}/metadef/graph/ge_attr_define.cc"
     "${GE_CODE_DIR}/metadef/graph/graph.cc"
@@ -110,7 +111,6 @@ target_compile_options(ut_libgraph PRIVATE
 
 target_compile_definitions(ut_libgraph PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ut_libgraph 

tests/ut/common/graph/testcase/ge_graph/ge_tensor_unittest.cc

@@ -230,7 +230,7 @@ TEST_F(UtestGeTensor, test_tensor_invalid_null) {
   GeTensor tensor(msg_owner, nullptr);
   EXPECT_EQ(tensor.GetData().size(), 0);
   EXPECT_EQ(tensor.MutableData().size(), 0);
-  EXPECT_EQ(tensor.SetData(Buffer(100)), ge::GRAPH_PARAM_INVALID);
+  EXPECT_EQ(tensor.SetData(Buffer(100)), GRAPH_SUCCESS);
 
   TensorUtils::SetWeightSize(tensor.MutableTensorDesc(), 100);
   EXPECT_EQ(TensorUtils::GetWeightSize(tensor), 0);

tests/ut/ge/CMakeLists.txt

@@ -89,6 +89,7 @@ set(COMMON_SRC_FILES
     "${GE_CODE_DIR}/metadef/graph/ge_attr_value.cc"
     "${GE_CODE_DIR}/metadef/graph/attr_value.cc"
     "${GE_CODE_DIR}/metadef/graph/buffer.cc"
+    "${GE_CODE_DIR}/metadef/graph/aligned_ptr.cc"
     "${GE_CODE_DIR}/metadef/graph/compute_graph.cc"
     "${GE_CODE_DIR}/metadef/graph/graph.cc"
     "${GE_CODE_DIR}/metadef/graph/gnode.cc"
@@ -227,6 +228,7 @@ set(COMMON_SRC_FILES
     "${GE_CODE_DIR}/ge/graph/passes/for_pass.cc"
     "${GE_CODE_DIR}/ge/graph/passes/enter_pass.cc"
     "${GE_CODE_DIR}/ge/graph/passes/assign_remove_pass.cc"
+    "${GE_CODE_DIR}/ge/graph/passes/inplace_support_check_pass.cc"
     "${GE_CODE_DIR}/ge/graph/passes/addn_pass.cc"
     "${GE_CODE_DIR}/ge/graph/passes/common_subexpression_elimination_pass.cc"
     "${GE_CODE_DIR}/ge/graph/passes/transop_symmetry_elimination_pass.cc"
@@ -303,6 +305,7 @@ set(COMMON_SRC_FILES
     "${GE_CODE_DIR}/ge/graph/common/local_context.cc"
     "${GE_CODE_DIR}/ge/graph/manager/graph_caching_allocator.cc"
     "${GE_CODE_DIR}/ge/graph/manager/rdma_pool_allocator.cc"
+    "${GE_CODE_DIR}/ge/graph/manager/host_mem_allocator.cc"
     "${GE_CODE_DIR}/ge/common/dump/dump_op.cc"
     "${GE_CODE_DIR}/ge/common/model_saver.cc"
     "${GE_CODE_DIR}/ge/hybrid/node_executor/aicpu/aicpu_ext_info.cc"
@@ -370,6 +373,7 @@ set(GRAPH_LOAD_COMMON_SRC_FILES
     "${GE_CODE_DIR}/ge/graph/manager/trans_var_data_utils.cc"
     "${GE_CODE_DIR}/ge/graph/manager/graph_caching_allocator.cc"
     "${GE_CODE_DIR}/ge/graph/manager/rdma_pool_allocator.cc"
+    "${GE_CODE_DIR}/ge/graph/manager/host_mem_allocator.cc"
     "${GE_CODE_DIR}/ge/common/thread_pool.cc"
 )
 
@@ -723,7 +727,6 @@ add_library(ge_ut_common STATIC ${COMMON_SRC_FILES} ${PROTO_SRCS} ${PROTO_HDRS})
 
 target_compile_definitions(ge_ut_common PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ge_ut_common PRIVATE
@@ -738,7 +741,6 @@ add_library(ge_ut_common_format STATIC ${COMMON_SRC_FILES} ${COMMON_FORMAT_SRC_F
 
 target_compile_definitions(ge_ut_common_format PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ge_ut_common_format PRIVATE
@@ -795,7 +797,6 @@ add_library(ge_load_common STATIC ${GRAPH_LOAD_COMMON_SRC_FILES} ${PROTO_SRCS} $
 
 target_compile_definitions(ge_load_common PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ge_load_common PRIVATE
@@ -810,7 +811,6 @@ add_library(ge_execute_common STATIC ${GRAPH_EXECUTE_COMMON_SRC_FILES} ${PROTO_S
 
 target_compile_definitions(ge_execute_common PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ge_execute_common PRIVATE
@@ -825,7 +825,6 @@ add_library(ge_build_common STATIC ${GRAPH_BUILD_COMMON_SRC_FILES} ${PROTO_SRCS}
 
 target_compile_definitions(ge_build_common PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ge_build_common PRIVATE
@@ -898,10 +897,6 @@ target_compile_options(ut_libge_others_utest PRIVATE
     -g --coverage -fprofile-arcs -ftest-coverage
 )
 
-target_compile_definitions(ut_libge_others_utest PRIVATE
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
-)
-
 target_link_libraries(ut_libge_others_utest
     $<BUILD_INTERFACE:intf_pub>
     ge_load_common ge_execute_common ge_ut_common gtest gtest_main ascend_protobuf ${COMMON_SHARED_LIBRARIES} json -lrt -ldl -lgcov
@@ -919,10 +914,6 @@ target_compile_options(ut_libge_kernel_utest PRIVATE
     -g --coverage -fprofile-arcs -ftest-coverage
 )
 
-target_compile_definitions(ut_libge_kernel_utest PRIVATE
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
-)
-
 target_link_libraries(ut_libge_kernel_utest
     $<BUILD_INTERFACE:intf_pub>
     ge_load_common ge_ut_common gtest gtest_main ascend_protobuf ${COMMON_SHARED_LIBRARIES} json -lrt -ldl -lgcov
@@ -943,7 +934,6 @@ target_compile_options(ut_libge_distinct_load_utest PRIVATE
 
 target_compile_definitions(ut_libge_distinct_load_utest PRIVATE
     google=ascend_private
-    $<$<STREQUAL:${ENABLE_OPEN_SRC},True>:ONLY_COMPILE_OPEN_SRC>
 )
 
 target_link_libraries(ut_libge_distinct_load_utest

third_party/fwkacllib/inc/aicpu/aicpu_schedule/aicpu_op_type_list.h

@@ -1,60 +1,60 @@
-/**
- * Copyright 2019-2020 Huawei Technologies Co., Ltd
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- * http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#ifndef AICPU_OP_TYPE_LIST_H_
-#define AICPU_OP_TYPE_LIST_H_
-
-enum OpKernelType {
-    TF_KERNEL,
-    CPU_KERNEL
-};
-
-enum ReturnCode {
-    OP_TYPE_NOT_SUPPORT,
-    FORMAT_NOT_SUPPORT,
-    DTYPE_NOT_SUPPORT
-};
-
-#pragma pack(push, 1)
-//One byte alignment
-struct SysOpInfo {
-    uint64_t opLen;
-    uint64_t opType;
-    OpKernelType kernelsType;
-};
-
-struct OpParamInfo {
-    uint64_t num;
-    uint64_t dtypeList;
-    uint64_t formatList;
-};
-
-struct SysOpCheckInfo {
-    uint64_t opListNum;
-    uint64_t offSetLen;
-    uint64_t sysOpInfoList;
-    uint64_t opParamInfoList;
-};
-
-struct SysOpCheckResp {
-    uint64_t opListNum;
-    bool isWithoutJson;
-    uint64_t returnCodeList;
-    uint64_t sysOpInfoList;
-    uint64_t opParamInfoList;
-};
-#pragma pack(pop)
-#endif  // AICPU_OP_TYPE_LIST_H_
+/**
+ * Copyright 2019-2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef AICPU_OP_TYPE_LIST_H_
+#define AICPU_OP_TYPE_LIST_H_
+
+enum OpKernelType {
+  TF_KERNEL,
+  CPU_KERNEL
+};
+
+enum ReturnCode {
+  OP_TYPE_NOT_SUPPORT,
+  FORMAT_NOT_SUPPORT,
+  DTYPE_NOT_SUPPORT
+};
+
+#pragma pack(push, 1)
+//One byte alignment
+struct SysOpInfo {
+  uint64_t opLen;
+  uint64_t opType;
+  OpKernelType kernelsType;
+};
+
+struct OpParamInfo {
+  uint64_t num;
+  uint64_t dtypeList;
+  uint64_t formatList;
+};
+
+struct SysOpCheckInfo {
+  uint64_t opListNum;
+  uint64_t offSetLen;
+  uint64_t sysOpInfoList;
+  uint64_t opParamInfoList;
+};
+
+struct SysOpCheckResp {
+  uint64_t opListNum;
+  bool isWithoutJson;
+  uint64_t returnCodeList;
+  uint64_t sysOpInfoList;
+  uint64_t opParamInfoList;
+};
+#pragma pack(pop)
+#endif  // AICPU_OP_TYPE_LIST_H_

third_party/fwkacllib/inc/hccl/hcom.h

@@ -33,15 +33,6 @@ extern "C" {
 
 
 
-/**
- * @brief Get the rank number in the group.
- *
- * @param group A string identifying the group name.
- * @param rankSize A pointer identifying the rank number.
- * @return HcclResult 
- */
-HcclResult hcom_get_rank_size(const char *group, u32 *rankSize);
-
 /**
  * @brief Get the rank number in the group.
  *
@@ -51,15 +42,6 @@ HcclResult hcom_get_rank_size(const char *group, u32 *rankSize);
  */
 HcclResult HcomGetRankSize(const char *group, u32 *rankSize);
 
-/**
- * @brief Get the rank number of this rank's server within the group.
- *
- * @param group A string identifying the group name.
- * @param localRankSize A pointer identifying the rank number.
- * @return HcclResult 
- */
-HcclResult hcom_get_local_rank_size(const char *group, u32 *localRankSize);
-
 /**
  * @brief Get the rank number of this rank's server within the group.
  *
@@ -69,15 +51,6 @@ HcclResult hcom_get_local_rank_size(const char *group, u32 *localRankSize);
  */
 HcclResult HcomGetLocalRankSize(const char *group, u32 *localRankSize);
 
-/**
- * @brief Get the rank id of this rank.
- *
- * @param group A string identifying the group name.
- * @param rankId A pointer identifying the rank id.
- * @return HcclResult 
- */
-HcclResult hcom_get_rank_id(const char *group, u32 *rankId);
-
 /**
  * @brief Get the rank id of this rank.
  *
@@ -87,15 +60,6 @@ HcclResult hcom_get_rank_id(const char *group, u32 *rankId);
  */
 HcclResult HcomGetRankId(const char *group, u32 *rankId);
 
-/**
- * @brief Get the local rank id of this rank's server within the group.
- *
- * @param group A string identifying the group name.
- * @param localRankId A pointer identifying the local rank id.
- * @return HcclResult 
- */
-HcclResult hcom_get_local_rank_id(const char *group, u32 *localRankId);
-
 /**
  * @brief Get the local rank id of this rank's server within the group.
  *
@@ -105,16 +69,6 @@ HcclResult hcom_get_local_rank_id(const char *group, u32 *localRankId);
  */
 HcclResult HcomGetLocalRankId(const char *group, u32 *localRankId);
 
-/**
- * @brief Get the world rank id according to the group rank id.
- *
- * @param group A string identifying the group name.
- * @param groupRank An integer(u32) identifying the group rank id.
- * @param worldRank A pointer identifying the world rank id.
- * @return HcclResult 
- */
-HcclResult hcom_get_world_rank_from_group_rank(const char *group, u32 groupRank, u32 *worldRank);
-
 /**
  * @brief Get the world rank id according to the group rank id.
  *
@@ -125,16 +79,6 @@ HcclResult hcom_get_world_rank_from_group_rank(const char *group, u32 groupRank,
  */
 HcclResult HcomGetWorldRankFromGroupRank(const char *group, u32 groupRank, u32 *worldRank);
 
-/**
- * @brief Get the group rank id according to the world rank id.
- *
- * @param worldRank An integer(u32) identifying the world rank id.
- * @param group A string identifying the group name.
- * @param groupRank A pointer identifying the group rank id.
- * @return HcclResult 
- */
-HcclResult hcom_get_group_rank_from_world_rank(u32 worldRank, const char *group, u32 *groupRank);
-
 /**
  * @brief Get the group rank id according to the world rank id.
  *
@@ -145,16 +89,6 @@ HcclResult hcom_get_group_rank_from_world_rank(u32 worldRank, const char *group,
  */
 HcclResult HcomGetGroupRankFromWorldRank(u32 worldRank, const char *group, u32 *groupRank);
 
-/**
- * @brief Create group.
- *
- * @param group A string identifying the group name.
- * @param rankNum An integer(u32) identifying the number of ranks in the group.
- * @param rankIds A list identifying the ranks in the group.
- * @return HcclResult 
- */
-HcclResult hcom_create_group(const char *group, u32 rankNum, u32 *rankIds);
-
 /**
  * @brief Create group.
  *
@@ -165,14 +99,6 @@ HcclResult hcom_create_group(const char *group, u32 rankNum, u32 *rankIds);
  */
 HcclResult HcomCreateGroup(const char *group, u32 rankNum, u32 *rankIds);
 
-/**
- * @brief Destroy group
- *
- * @param group A string identifying the group name.
- * @return HcclResult 
- */
-HcclResult hcom_destroy_group(const char *group);
-
 /**
  * @brief Destroy group
  *
@@ -189,46 +115,54 @@ HcclResult HcomDestroyGroup(const char *group);
  * @param IdxList A list identifying the index of end gradient in each segment.
  * @return HcclResult
  */
-extern HcclResult hcom_set_split_strategy_by_index(const char *group, u32 segmentNum, const u32 *IdxList);
+extern HcclResult HcomSetGradFusionByIndex(const char *group, u32 segmentNum, const u32 *IdxList);
 
 /**
- * @brief Set the gradient split strategy with in the group, according to gradient index.
+ * @brief Set the gradient split strategy with in the group, according to gradient data size.
  *
  * @param group A string identifying the group name.
  * @param segmentNum An integer(u32) identifying the segments number of gradients.
- * @param IdxList A list identifying the index of end gradient in each segment.
+ * @param sizeList A list identifying the percent of each segment.
  * @return HcclResult
  */
-extern HcclResult HcomSetGradFusionByIndex(const char *group, u32 segmentNum, const u32 *IdxList);
+extern HcclResult HcomSetGradFusionBySize(const char *group, u32 segmentNum, const float *sizeList);
 
 /**
- * @brief Set the gradient split strategy with in the group, according to gradient data size.
+ * @brief Initialize hcom executor.
  *
- * @param group A string identifying the group name.
- * @param segmentNum An integer(u32) identifying the segments number of gradients.
- * @param sizeList A list identifying the percent of each segment.
+ * @param void
  * @return HcclResult
  */
-extern HcclResult hcom_set_split_strategy_by_size(const char *group, u32 segmentNum, const float *sizeList);
+HcclResult HcomExecInitialize();
 
 /**
- * @brief Set the gradient split strategy with in the group, according to gradient data size.
+ * @brief Finalize hcom executor.
  *
- * @param group A string identifying the group name.
- * @param segmentNum An integer(u32) identifying the segments number of gradients.
- * @param sizeList A list identifying the percent of each segment.
+ * @param void
  * @return HcclResult
  */
-extern HcclResult HcomSetGradFusionBySize(const char *group, u32 segmentNum, const float *sizeList);
+HcclResult HcomExecFinalize();
 
 /**
- * @brief Register memories and init resources for remote access.
+ * @brief Put collective communication operation into hcom executor.
  *
- * @param addrList memory addresses for remote access.
- * @param count number of remote memory addresses.
+ * @param opInfo information about collective communication operation.
+ * @param callback callback after collective communication operation.
  * @return HcclResult
  */
-extern HcclResult hcom_remote_access_mem_register(const MemRegisterAddr* addrList, u32 count);
+HcclResult HcomExecEnqueueOperation(HcomOperation opInfo, std::function<void(HcclResult status)> callback);
+
+/**
+ * @brief Put remote access operation into hcom executor.
+ *
+ * @param remoteAccessType operation type (read or write).
+ * @param addrInfos address information about collective communication operation.
+ * @param callback callback after collective communication operation.
+ * @return HcclResult
+ */
+HcclResult HcomExecEnqueueRemoteAccess(const std::string& remoteAccessType,
+                                       const std::vector<HcomRemoteAccessAddrInfo>& addrInfos,
+                                       std::function<void(HcclResult status)> callback);
 
 /**
  * @brief Register memories and init resources for remote access.
@@ -239,16 +173,6 @@ extern HcclResult hcom_remote_access_mem_register(const MemRegisterAddr* addrLis
  */
 extern HcclResult HcomRegRemoteAccessMem(const MemRegisterAddr* addrList, u32 count);
 
-HcclResult HcomExecInitialize();
-
-HcclResult HcomExecFinalize();
-
-HcclResult HcomExecEnqueueOperation(HcomOperation opInfo, std::function<void(HcclResult status)> callback);
-
-HcclResult HcomExecEnqueueRemoteAccess(const std::string& remoteAccessType,
-                               const std::vector<HcomRemoteAccessAddrInfo>& addrInfos,
-                               std::function<void(HcclResult status)> callback);
-
 #ifdef __cplusplus
 }
 #endif // __cplusplus

third_party/fwkacllib/inc/mmpa/sub_inc/mmpa_linux.h

@@ -279,8 +279,9 @@ typedef struct {
 #define M_NAME_MAX MAX_FNAME
 
 #define M_F_OK F_OK
-#define M_R_OK R_OK
+#define M_X_OK X_OK
 #define M_W_OK W_OK
+#define M_R_OK R_OK
 
 #define MM_DT_DIR DT_DIR
 #define MM_DT_REG DT_REG

third_party/fwkacllib/inc/mmpa/sub_inc/mmpa_win.h

@@ -322,6 +322,7 @@ typedef VOID (*mmPf)(VOID);
 #define M_NAME_MAX  _MAX_FNAME
 
 #define M_F_OK 0
+#define M_X_OK 1
 #define M_W_OK 2
 #define M_R_OK 4
 

third_party/fwkacllib/inc/register/op_kernel_registry.h

@@ -0,0 +1,49 @@
+/**
+ * Copyright 2019-2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef INC_REGISTER_OP_KERNEL_REGISTRY_H_
+#define INC_REGISTER_OP_KERNEL_REGISTRY_H_
+#include <memory>
+#include <string>
+#include "register/register_types.h"
+#include "register.h"
+
+namespace ge {
+class FMK_FUNC_HOST_VISIBILITY FMK_FUNC_DEV_VISIBILITY OpKernelRegistry {
+ public:
+  using CreateFn = HostCpuOp* (*)();
+  ~OpKernelRegistry();
+
+  static OpKernelRegistry& GetInstance() {
+    static OpKernelRegistry instance;
+    return instance;
+  }
+
+  bool IsRegistered(const std::string &op_type);
+
+  void RegisterHostCpuOp(const std::string &op_type, CreateFn create_fn);
+
+  std::unique_ptr<HostCpuOp> CreateHostCpuOp(const std::string &op_type);
+
+ private:
+  OpKernelRegistry();
+  class OpKernelRegistryImpl;
+  /*lint -e148*/
+  std::unique_ptr<OpKernelRegistryImpl> impl_;
+};
+} // namespace ge
+
+#endif // INC_REGISTER_OP_KERNEL_REGISTRY_H_

third_party/fwkacllib/inc/register/op_registry.h

@@ -0,0 +1,96 @@
+/**
+ * Copyright 2019-2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef INC_REGISTER_OP_REGISTRY_H_
+#define INC_REGISTER_OP_REGISTRY_H_
+
+#include <limits.h>
+#include <set>
+#include <string>
+#include <unordered_map>
+#include <vector>
+
+#include "register/register.h"
+
+namespace domi {
+enum RemoveInputType {
+  OMG_MOVE_TYPE_DTYPE = 0,
+  OMG_MOVE_TYPE_VALUE,
+  OMG_MOVE_TYPE_SHAPE,
+  OMG_MOVE_TYPE_FORMAT,
+  OMG_MOVE_TYPE_AXIS,
+  OMG_MOVE_TYPE_SCALAR_VALUE,
+  OMG_REMOVE_TYPE_WITH_COND = 1000,
+  OMG_REMOVE_INPUT_WITH_ORIGINAL_TYPE,
+  OMG_INPUT_REORDER,
+};
+
+struct RemoveInputConfigure {
+  int inputIdx = INT_MAX;
+  std::string attrName;
+  RemoveInputType moveType;
+  bool attrValue = false;
+  std::string originalType;
+  std::vector<int> input_order;
+};
+
+class FMK_FUNC_HOST_VISIBILITY FMK_FUNC_DEV_VISIBILITY OpRegistry {
+ public:
+  static OpRegistry *Instance();
+
+  std::vector<OpRegistrationData> registrationDatas;
+
+  bool Register(const OpRegistrationData &reg_data);
+
+  domi::ImplyType GetImplyType(const std::string &op_type);
+
+  void GetOpTypeByImplyType(std::vector<std::string> &vec_op_type, const domi::ImplyType &imply_type);
+
+  domi::ParseParamFunc GetParseParamFunc(const std::string &op_type, const std::string &ori_type);
+
+  domi::ParseParamByOpFunc GetParseParamByOperatorFunc(const std::string &ori_type);
+
+  domi::FusionParseParamFunc GetFusionParseParamFunc(const std::string &op_type, const std::string &ori_type);
+
+  domi::FusionParseParamByOpFunc GetFusionParseParamByOpFunc(const std::string &op_type,
+                                                             const std::string &ori_type);
+
+  domi::ParseSubgraphFunc GetParseSubgraphPostFunc(const std::string &op_type);
+
+  Status GetParseSubgraphPostFunc(const std::string &op_type, domi::ParseSubgraphFuncV2 &parse_subgraph_func);
+
+  domi::ImplyType GetImplyTypeByOriOpType(const std::string &ori_optype);
+
+  const std::vector<RemoveInputConfigure> &GetRemoveInputConfigure(const std::string &ori_optype) const;
+
+  bool GetOmTypeByOriOpType(const std::string &ori_optype, std::string &om_type);
+
+  ParseOpToGraphFunc GetParseOpToGraphFunc(const std::string &op_type, const std::string &ori_type);
+
+ private:
+  std::unordered_map<std::string, domi::ImplyType> op_run_mode_map_;
+  std::unordered_map<std::string, ParseParamFunc> op_parse_params_fn_map_;
+  std::unordered_map<std::string, ParseParamByOpFunc> parse_params_by_op_func_map_;
+  std::unordered_map<std::string, FusionParseParamFunc> fusion_op_parse_params_fn_map_;
+  std::unordered_map<std::string, FusionParseParamByOpFunc> fusion_parse_params_by_op_fn_map_;
+  std::unordered_map<std::string, ParseSubgraphFunc> op_types_to_parse_subgraph_post_func_;
+  std::unordered_map<std::string, std::vector<RemoveInputConfigure>> remove_input_configure_map_;
+  std::unordered_map<std::string, std::string> origin_type_to_om_type_;
+  std::unordered_map<std::string, ParseOpToGraphFunc> parse_op_to_graph_fn_map_;
+  std::unordered_map<std::string, ParseSubgraphFuncV2> op_types_to_parse_subgraph_post_func_v2_;
+};
+}  // namespace domi
+#endif  // INC_REGISTER_OP_REGISTRY_H_

third_party/fwkacllib/inc/runtime/base.h

@@ -81,26 +81,17 @@ typedef enum tagRtLimitType {
 } rtLimitType_t;
 
 typedef struct rtExceptionInfo {
-    uint32_t taskid;
-    uint32_t streamid;
-    uint32_t tid;
-    uint32_t deviceid;
+  uint32_t taskid;
+  uint32_t streamid;
+  uint32_t tid;
+  uint32_t deviceid;
+  uint32_t retcode;
 } rtExceptionInfo;
 
-typedef struct rtTaskFailInfo {
-    uint32_t taskid;
-    uint32_t streamid;
-    uint32_t tid;
-    uint32_t deviceid;
-    uint32_t retcode;
-} rtTaskFailInfo;
-
 typedef void (*rtErrorCallback)(rtExceptionType);
 
 typedef void (*rtTaskFailCallback)(rtExceptionInfo *exceptionInfo);
 
-typedef void (*rtTaskFailCallbackByModule)(rtTaskFailInfo *exceptionInfo);
-
 typedef void (*rtDeviceStateCallback)(uint32_t devId, bool isOpen);
 
 /**
@@ -143,13 +134,13 @@ RTS_API rtError_t rtProfilerConfig(uint16_t type);
  * @ingroup profiling_base
  * @brief start rts profiler.
  */
-RTS_API rtError_t rtProfilerStart(uint64_t profConfig, int32_t numsDev, uint32_t* deviceList);
+RTS_API rtError_t rtProfilerStart(uint64_t profConfig, int32_t numsDev, uint32_t *deviceList);
 
 /**
  * @ingroup profiling_base
  * @brief stop rts profiler.
  */
-RTS_API rtError_t rtProfilerStop(uint64_t profConfig, int32_t numsDev, uint32_t* deviceList);
+RTS_API rtError_t rtProfilerStop(uint64_t profConfig, int32_t numsDev, uint32_t *deviceList);
 
 /**
  * @ingroup profiling_base
@@ -209,7 +200,7 @@ RTS_API rtError_t rtRegDeviceStateCallback(const char *regName, rtDeviceStateCal
  * @param [out] NA
  * @return RT_ERROR_NONE for ok
  */
-RTS_API rtError_t rtRegTaskFailCallbackByModule(const char *moduleName, rtTaskFailCallbackByModule callback);
+RTS_API rtError_t rtRegTaskFailCallbackByModule(const char *moduleName, rtTaskFailCallback callback);
 
 /**
  * @ingroup dvrt_base

third_party/fwkacllib/inc/runtime/config.h

@@ -42,6 +42,7 @@ typedef enum tagRtChipType {
   CHIP_MDC,
   CHIP_LHISI,
   CHIP_DC,
+  CHIP_CLOUD_V2,
   CHIP_END,
 } rtChipType_t;
 
@@ -62,6 +63,7 @@ typedef enum tagRtPlatformType {
   PLATFORM_LHISI_ES,
   PLATFORM_LHISI_CS,
   PLATFORM_DC,
+  PLATFORM_CLOUD_V2,
   PLATFORM_END,
 } rtPlatformType_t;
 
@@ -119,7 +121,9 @@ typedef struct tagRtMemoryConfig {
   uint32_t compilerSize;
 } rtMemoryConfig_t;
 
-typedef struct tagRtPlatformConfig { uint32_t platformConfig; } rtPlatformConfig_t;
+typedef struct tagRtPlatformConfig {
+  uint32_t platformConfig;
+} rtPlatformConfig_t;
 
 /**
  * @ingroup

third_party/fwkacllib/inc/runtime/context.h

@@ -47,7 +47,7 @@ typedef struct tagRtGroupInfo {
   uint32_t aivectorNum;
   uint32_t sdmaNum;
   uint32_t activeStreamNum;
-  void*  extrPtr;
+  void *extrPtr;
 } rtGroupInfo_t;
 
 /**

third_party/fwkacllib/inc/runtime/dev.h

@@ -185,7 +185,7 @@ RTS_API rtError_t rtDisableP2P(uint32_t devIdDes, uint32_t phyIdSrc);
  * @return RT_ERROR_NONE for ok
  * @return RT_ERROR_INVALID_VALUE for error input
  */
-RTS_API rtError_t rtDeviceCanAccessPeer(int32_t* canAccessPeer, uint32_t device, uint32_t peerDevice);
+RTS_API rtError_t rtDeviceCanAccessPeer(int32_t *canAccessPeer, uint32_t device, uint32_t peerDevice);
 
 /**
  * @ingroup dvrt_dev

third_party/fwkacllib/inc/runtime/kernel.h

@@ -387,7 +387,7 @@ typedef void *rtModel_t;
  * @return RT_ERROR_NONE for ok
  * @return RT_ERROR_INVALID_VALUE for error input
  */
- RTS_API rtError_t rtDumpAddrSet(rtModel_t model, void *addr, uint32_t dumpSize, uint32_t flag);
+RTS_API rtError_t rtDumpAddrSet(rtModel_t model, void *addr, uint32_t dumpSize, uint32_t flag);
 
 /**
  * @ingroup rt_kernel

third_party/fwkacllib/inc/runtime/mem.h

@@ -159,11 +159,11 @@ typedef struct rtAiCoreMemorySize {
  * @ingroup dvrt_mem
  * @brief memory type
  */
-typedef enum tagRtMemoryType { 
-    RT_MEMORY_TYPE_HOST = 1, 
-    RT_MEMORY_TYPE_DEVICE = 2 , 
-    RT_MEMORY_TYPE_SVM = 3,
-    RT_MEMORY_TYPE_DVPP = 4
+typedef enum tagRtMemoryType {
+  RT_MEMORY_TYPE_HOST = 1,
+  RT_MEMORY_TYPE_DEVICE = 2,
+  RT_MEMORY_TYPE_SVM = 3,
+  RT_MEMORY_TYPE_DVPP = 4
 } rtMemoryType_t;
 
 /**
@@ -179,23 +179,23 @@ typedef struct tagRtPointerAttributes {
 
 
 typedef struct rtMallocHostSharedMemoryIn {
-    const char* name;
-    const uint64_t size;
-    uint32_t flag;
+  const char *name;
+  const uint64_t size;
+  uint32_t flag;
 } rtMallocHostSharedMemoryIn;
 
 typedef struct rtMallocHostSharedMemoryOut {
-    int fd;
-    void* ptr;
-    void* devPtr;
+  int fd;
+  void *ptr;
+  void *devPtr;
 } rtMallocHostSharedMemoryOut;
 
 typedef struct rtFreeHostSharedMemoryIn {
-    const char* name;
-    const uint64_t size;
-    int fd;
-    void* ptr;
-    void* devPtr;
+  const char *name;
+  const uint64_t size;
+  int fd;
+  void *ptr;
+  void *devPtr;
 } rtFreeHostSharedMemoryIn;
 
 
@@ -267,7 +267,7 @@ RTS_API rtError_t rtFreeHost(void *hostPtr);
  */
 
 RTS_API rtError_t rtMallocHostSharedMemory(rtMallocHostSharedMemoryIn *in,
-    rtMallocHostSharedMemoryOut *out);
+                                           rtMallocHostSharedMemoryOut *out);
 
 /**
  * @ingroup dvrt_mem

third_party/fwkacllib/inc/runtime/stream.h

@@ -36,6 +36,7 @@ extern "C" {
 #define RT_STREAM_FORBIDDEN_DEFAULT (0x10)
 #define RT_STREAM_HEAD (0x20)
 #define RT_STREAM_PRIMARY_DEFAULT (0x40)
+#define RT_STREAM_PRIMARY_FIRST_DEFAULT (0x80)
 
 /**
  * @ingroup stream_type

third_party/fwkacllib/inc/soft_dp/ExternalSoftDp.h

@@ -0,0 +1,52 @@
+/**
+* @file ExternalSoftDp.h
+*
+* Copyright (c) Huawei Technologies Co., Ltd. 2012-2018. All rights reserved.
+*
+* This program is distributed in the hope that it will be useful,
+* but WITHOUT ANY WARRANTY; without even the implied warranty of
+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+*/
+
+#ifndef EXTERNALSOFTDP_H
+#define EXTERNALSOFTDP_H
+
+#include <stdint.h>
+
+extern "C" {
+struct SoftDpProcsessInfo {
+  uint8_t* inputBuffer;
+  uint32_t inputBufferSize;
+
+  uint8_t* outputBuffer;
+  uint32_t outputBufferSize;
+
+  uint32_t outputWidth;
+  uint32_t outputHeight;
+
+  uint32_t reserved;
+};
+
+struct DpCropInfo {
+  uint32_t left;
+  uint32_t right;
+  uint32_t up;
+  uint32_t down;
+};
+
+/*
+ * @brief decode and resize interface
+ * @param [in] SoftDpProcsessInfo& softDpProcsessInfo : soft dp struct
+ * @return success: return 0, fail: return error number
+ */
+uint32_t DecodeAndResizeJpeg(SoftDpProcsessInfo& softDpProcsessInfo);
+
+/*
+ * @brief decode crop and resize interface
+ * @param [in] SoftDpProcsessInfo& softDpProcsessInfo : soft dp struct
+ * @param [in] const DpCropInfo& cropInfo: crop struct
+ * @return success: return 0, fail: return error number
+ */
+uint32_t DecodeAndCropAndResizeJpeg(SoftDpProcsessInfo& softDpProcsessInfo, const DpCropInfo& cropInfo);
+}
+#endif // EXTERNALSOFTDP_H

third_party/fwkacllib/inc/toolchain/plog.h

@@ -0,0 +1,59 @@
+/**
+ * Copyright 2019-2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef _PLOG_H_
+#define _PLOG_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif // __cplusplus
+
+#ifndef LINUX
+#define LINUX 0
+#endif // LINUX
+
+#ifndef WIN
+#define WIN 1
+#endif
+
+#ifndef OS_TYPE
+#define OS_TYPE 0
+#endif // OS_TYPE
+
+#if (OS_TYPE == LINUX)
+#define DLL_EXPORT __attribute__((visibility("default")))
+#else
+#define DLL_EXPORT _declspec(dllexport)
+#endif
+
+/**
+ * @ingroup plog
+ * @brief DlogReportInitialize: init log in service process before all device setting.
+ * @return: 0: SUCCEED, others: FAILED
+ */
+DLL_EXPORT int DlogReportInitialize();
+
+/**
+ * @ingroup plog
+ * @brief DlogReportFinalize: release log resource in service process after all device reset.
+ * @return: 0: SUCCEED, others: FAILED
+ */
+DLL_EXPORT int DlogReportFinalize();
+
+#ifdef __cplusplus
+}
+#endif // __cplusplus
+#endif // D_PLOG_H_

third_party/fwkacllib/inc/toolchain/slog.h

@@ -18,7 +18,9 @@
 #define D_SYSLOG_H_
 
 #ifdef __cplusplus
+#ifndef LOG_CPP
 extern "C" {
+#endif
 #endif // __cplusplus
 
 #ifndef LINUX
@@ -105,6 +107,7 @@ extern "C" {
 #define SECURITY_LOG_MASK   (0x00100000)
 #define RUN_LOG_MASK        (0x01000000)
 #define OPERATION_LOG_MASK  (0x10000000)
+#define RESERVERD_LENGTH 52
 
 typedef struct tagDCODE {
   const char *cName;
@@ -116,6 +119,18 @@ typedef struct tagKV {
   char *value;
 } KeyValue;
 
+typedef enum {
+  APPLICATION = 0,
+  SYSTEM
+} ProcessType;
+
+typedef struct {
+  ProcessType type;
+  unsigned int pid;
+  unsigned int deviceId;
+  char reserved[RESERVERD_LENGTH];
+} LogAttr;
+
 /**
  * @ingroup slog
  *
@@ -228,6 +243,14 @@ DLL_EXPORT int dlog_setlevel(int moduleId, int level, int enableEvent);
  */
 DLL_EXPORT int CheckLogLevel(int moduleId, int logLevel);
 
+/**
+ * @ingroup slog
+ * @brief DlogSetAttr: set log attr, default pid is 0, default device id is 0, default process type is APPLICATION
+ * @param [in]logAttr: attr info, include pid(must be larger than 0), process type and device id(chip ID)
+ * @return: 0: SUCCEED, others: FAILED
+ */
+DLL_EXPORT int DlogSetAttr(LogAttr logAttr);
+
 /**
  * @ingroup slog
  * @brief dlog_error: print error log
@@ -367,6 +390,121 @@ void DlogInner(int moduleId, int level, const char *fmt, ...);
 void DlogWithKVInner(int moduleId, int level, KeyValue *pstKVArray, int kvNum, const char *fmt, ...);
 
 #ifdef __cplusplus
+#ifndef LOG_CPP
 }
+#endif // LOG_CPP
 #endif // __cplusplus
+
+#ifdef LOG_CPP
+#ifdef __cplusplus
+extern "C" {
+#endif
+/**
+ * @ingroup slog
+ * @brief DlogGetlevelForC: get module loglevel and enableEvent
+ *
+ * @param [in]moduleId: moudule id(see slog.h, eg: CCE), others: invalid
+ * @param [out]enableEvent: 1: enable; 0: disable
+ * @return: module level(0: debug, 1: info, 2: warning, 3: error, 4: null output)
+ */
+DLL_EXPORT int DlogGetlevelForC(int moduleId, int *enableEvent);
+
+/**
+ * @ingroup slog
+ * @brief DlogSetlevelForC: set module loglevel and enableEvent
+ *
+ * @param [in]moduleId: moudule id(see slog.h, eg: CCE), -1: all modules, others: invalid
+ * @param [in]level: log level(0: debug, 1: info, 2: warning, 3: error, 4: null output)
+ * @param [in]enableEvent: 1: enable; 0: disable, others:invalid
+ * @return: 0: SUCCEED, others: FAILED
+ */
+DLL_EXPORT int DlogSetlevelForC(int moduleId, int level, int enableEvent);
+
+/**
+ * @ingroup slog
+ * @brief CheckLogLevelForC: check module level enable or not
+ * users no need to call it because all dlog interface(include inner interface) has already called
+ *
+ * @param [in]moduleId: module id, eg: CCE
+ * @param [in]logLevel: eg: DLOG_EVENT/DLOG_ERROR/DLOG_WARN/DLOG_INFO/DLOG_DEBUG
+ * @return: 1:enable, 0:disable
+ */
+DLL_EXPORT int CheckLogLevelForC(int moduleId, int logLevel);
+
+/**
+ * @ingroup slog
+ * @brief DlogSetAttrForC: set log attr, default pid is 0, default device id is 0, default process type is APPLICATION
+ * @param [in]logAttr: attr info, include pid(must be larger than 0), process type and device id(chip ID)
+ * @return: 0: SUCCEED, others: FAILED
+ */
+DLL_EXPORT int DlogSetAttrForC(LogAttr logAttr);
+
+/**
+ * @ingroup slog
+ * @brief DlogForC: print log, need caller to specify level
+ * call CheckLogLevelForC in advance to optimize performance, call interface with fmt input take time
+ *
+ * @param [in]moduleId: module id, eg: CCE
+ * @param [in]level(0: debug, 1: info, 2: warning, 3: error, 5: trace, 6: oplog, 16: event)
+ * @param [in]fmt: log content
+ */
+#define DlogForC(moduleId, level, fmt, ...)                                                 \
+  do {                                                                                  \
+    if(CheckLogLevelForC(moduleId, level) == 1) {                                           \
+        DlogInnerForC(moduleId, level, "[%s:%d]" fmt, __FILE__, __LINE__, ##__VA_ARGS__);   \
+     }                                                                                  \
+  } while (0)
+
+/**
+ * @ingroup slog
+ * @brief DlogSubForC: print log, need caller to specify level and submodule
+ * call CheckLogLevelForC in advance to optimize performance, call interface with fmt input take time
+ *
+ * @param [in]moduleId: module id, eg: CCE
+ * @param [in]submodule: eg: engine
+ * @param [in]level(0: debug, 1: info, 2: warning, 3: error, 5: trace, 6: oplog, 16: event)
+ * @param [in]fmt: log content
+ */
+#define DlogSubForC(moduleId, submodule, level, fmt, ...)                                                   \
+  do {                                                                                                  \
+    if(CheckLogLevelForC(moduleId, level) == 1) {                                                           \
+        DlogInnerForC(moduleId, level, "[%s:%d][%s]" fmt, __FILE__, __LINE__, submodule, ##__VA_ARGS__);    \
+    }                                                                                                   \
+  } while (0)
+
+/**
+ * @ingroup slog
+ * @brief DlogWithKVForC: print log, need caller to specify level and other paramters
+ * call CheckLogLevelForC in advance to optimize performance, call interface with fmt input take time
+ *
+ * @param [in]moduleId: module id, eg: CCE
+ * @param [in]level(0: debug, 1: info, 2: warning, 3: error, 5: trace, 6: oplog, 16: event)
+ * @param [in]pstKVArray: key-value array
+ * @param [in]kvNum: key-value element num in array
+ * @param [in]fmt: log content
+ */
+#define DlogWithKVForC(moduleId, level, pstKVArray, kvNum, fmt, ...)                                                \
+  do {                                                                                                          \
+    if(CheckLogLevelForC(moduleId, level) == 1) {                                                                   \
+        DlogWithKVInnerForC(moduleId, level, pstKVArray, kvNum, "[%s:%d]" fmt, __FILE__, __LINE__, ##__VA_ARGS__);  \
+    }                                                                                                           \
+  } while (0)
+
+/**
+ * @ingroup slog
+ * @brief DlogFlushForC: flush log buffer to file
+ */
+DLL_EXPORT void DlogFlushForC(void);
+
+/**
+ * @ingroup slog
+ * @brief Internal log interface, other modules are not allowed to call this interface
+ */
+void DlogInnerForC(int moduleId, int level, const char *fmt, ...);
+void DlogWithKVInnerForC(int moduleId, int level, KeyValue *pstKVArray, int kvNum, const char *fmt, ...);
+
+#ifdef __cplusplus
+}
+#endif
+#endif // LOG_CPP
 #endif // D_SYSLOG_H_