You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2021/11/29 17:15:59 UTC
[GitHub] [tvm] areusch commented on a change in pull request #9553: [PROFILING] Add ability to profile a single function_profiling
areusch commented on a change in pull request #9553:
URL: https://github.com/apache/tvm/pull/9553#discussion_r758572457
##########
File path: src/runtime/profiling.cc
##########
@@ -677,6 +677,56 @@ TVM_REGISTER_GLOBAL("runtime.profiling.FromJSON").set_body_typed(Report::FromJSO
TVM_REGISTER_GLOBAL("runtime.profiling.DeviceWrapper").set_body_typed([](Device dev) {
return DeviceWrapper(dev);
});
+
+PackedFunc ProfileFunction(Module mod, std::string func_name, int device_type, int device_id,
+ Array<MetricCollector> collectors) {
+ // Module::GetFunction is not const, so this lambda has to be mutable
+ return PackedFunc([=](TVMArgs args, TVMRetValue* ret) mutable {
+ PackedFunc f = mod.GetFunction(func_name);
+ Device dev{static_cast<DLDeviceType>(device_type), device_id};
+
+ // warmup
+ for (size_t i = 0; i < 10; i++) {
+ f.CallPacked(args, ret);
+ }
+
+ for (auto& collector : collectors) {
+ collector->Init({DeviceWrapper(dev)});
+ }
+ std::vector<ObjectRef> collector_data;
+ for (auto& collector : collectors) {
+ collector_data.push_back(collector->Start(dev));
+ }
+ // TODO(tkonolige): repeated calls if the runtime is small?
+ f.CallPacked(args, ret);
+ std::unordered_map<String, ObjectRef> results;
+ for (size_t i = 0; i < collectors.size(); i++) {
+ auto r = collectors[i]->Stop(collector_data[i]);
+ // We might want to do this in a separate loop to avoid unnecessary time
Review comment:
should we just do that now? seems better...
##########
File path: src/runtime/profiling.cc
##########
@@ -677,6 +677,56 @@ TVM_REGISTER_GLOBAL("runtime.profiling.FromJSON").set_body_typed(Report::FromJSO
TVM_REGISTER_GLOBAL("runtime.profiling.DeviceWrapper").set_body_typed([](Device dev) {
return DeviceWrapper(dev);
});
+
+PackedFunc ProfileFunction(Module mod, std::string func_name, int device_type, int device_id,
+ Array<MetricCollector> collectors) {
+ // Module::GetFunction is not const, so this lambda has to be mutable
+ return PackedFunc([=](TVMArgs args, TVMRetValue* ret) mutable {
+ PackedFunc f = mod.GetFunction(func_name);
+ Device dev{static_cast<DLDeviceType>(device_type), device_id};
+
+ // warmup
+ for (size_t i = 0; i < 10; i++) {
Review comment:
seems like this one should be configurable?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org