p47068915
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

save .pb graphs for GPU version 

fix Cppcheck error
This commit is contained in:
lichen_101010 2020-07-15 16:28:30 -04:00
parent 60927ef130
commit 515099d0a9
1 changed files with 13 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
mindspore/ccsrc/backend/session/gpu_session.cc
View File

@ -13,6 +13,7 @@
* See the License for the specific language governing permissions and * See the License for the specific language governing permissions and
* limitations under the License. * limitations under the License.
*/ */
#include "debug/anf_ir_utils.h"
#include "backend/session/gpu_session.h" #include "backend/session/gpu_session.h"
#include "runtime/device/gpu/kernel_info_setter.h" #include "runtime/device/gpu/kernel_info_setter.h"
#include "runtime/device/gpu/gpu_kernel_build.h" #include "runtime/device/gpu/gpu_kernel_build.h"
@ -163,6 +164,10 @@ GraphId GPUSession::CompileGraph(const AnfNodePtrList &lst, const AnfNodePtrList
auto graph_id = graph_sum_; auto graph_id = graph_sum_;
auto graph = ConstructKernelGraph(lst, outputs); auto graph = ConstructKernelGraph(lst, outputs);
MS_EXCEPTION_IF_NULL(graph); MS_EXCEPTION_IF_NULL(graph);
// Prepare ms context info for dump .pb graph
auto context_ptr = MsContext::GetInstance();
MS_EXCEPTION_IF_NULL(context_ptr);
bool save_graphs = context_ptr->save_graphs_flag();
// Optimize // Optimize
Optimize(graph); Optimize(graph);
// Select kernel build info // Select kernel build info
@ -171,8 +176,16 @@ GraphId GPUSession::CompileGraph(const AnfNodePtrList &lst, const AnfNodePtrList
predictmodel::StepConvertGraph(graph); predictmodel::StepConvertGraph(graph);
// Start gpu kernel runtime // Start gpu kernel runtime
StartKernelRT(); StartKernelRT();
// Dump .pb graph before hardware optimization
if (save_graphs) {
DumpIRProto(graph, "before_hwopt_" + std::to_string(graph_id));
}
// HardwareOptimize // HardwareOptimize
HardwareOptimize(graph); HardwareOptimize(graph);
// Dump .pb graph after hardware optimization
if (save_graphs) {
DumpIRProto(graph, "after_hwopt_" + std::to_string(graph_id));
}
// Assign CUDA streams // Assign CUDA streams
AssignStream(graph); AssignStream(graph);
// Hide NoOp from execution graph // Hide NoOp from execution graph