Skip to content

[AIU Profiler] Fix flow event creation #1229

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mcalman wants to merge 1 commit into
base: main
Choose a base branch
from
+11 −7
Open

[AIU Profiler] Fix flow event creation #1229

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 9 additions & 5 deletions libkineto/src/plugin/aiupti/AiuptiActivityHandlers.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -172,8 +172,12 @@ void AiuptiActivityProfilerSession::handleRuntimeActivity(
runtime_activity->device = activity->process_id;
runtime_activity->resource = systemThreadId();
runtime_activity->threadId = threadId();
// TODO: verify the flow logic
runtime_activity->flow.id = activity->correlation_id;
// only enable outgoing flow for launch control block runtime activities
if (activity->cbid == AIUPTI_RUNTIME_TRACE_CBID_LAUNCH_CB_CMPT) {
runtime_activity->flow.id = activity->correlation_id;
} else {
runtime_activity->flow.id = 0;
}
runtime_activity->flow.type = libkineto::kLinkAsyncCpuGpu;
runtime_activity->flow.start = bool(
std::find(
Expand Down Expand Up @@ -345,7 +349,7 @@ void AiuptiActivityProfilerSession::handleMemcpyActivity(
memcpy_activity->device = activity->device_id;
memcpy_activity->resource = getResourceId(activity);
memcpy_activity->threadId = activity->stream_id;
memcpy_activity->flow.id = activity->correlation_id;
memcpy_activity->flow.id = 0;
memcpy_activity->flow.type = libkineto::kLinkAsyncCpuGpu;
memcpy_activity->flow.start = 0;
memcpy_activity->linked = linked;
Expand Down Expand Up @@ -417,7 +421,7 @@ void AiuptiActivityProfilerSession::handleMemoryActivity(
mem_activity->device = activity->device_id;
mem_activity->resource = getResourceId(activity);
mem_activity->threadId = activity->stream_id;
mem_activity->flow.id = activity->correlation_id;
mem_activity->flow.id = 0;
mem_activity->flow.type = libkineto::kLinkAsyncCpuGpu;
mem_activity->flow.start = 0;
mem_activity->linked = linked;
Expand Down Expand Up @@ -499,7 +503,7 @@ void AiuptiActivityProfilerSession::handleMemsetActivity(
// of order This prevents us from using getResourceId which handles overlap
memset_activity->resource = getBaseResourceId(activity);
memset_activity->threadId = activity->stream_id;
memset_activity->flow.id = activity->correlation_id;
memset_activity->flow.id = 0;
memset_activity->flow.type = libkineto::kLinkAsyncCpuGpu;
memset_activity->flow.start = 0;
memset_activity->linked = linked;
Expand Down
4 changes: 2 additions & 2 deletions libkineto/src/plugin/aiupti/AiuptiActivityProfiler.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,8 +9,8 @@ uint32_t AiuptiActivityProfilerSession::iterationCount_ = 0;
std::vector<std::array<unsigned char, 16>>
AiuptiActivityProfilerSession::deviceUUIDs_ = {};
std::vector<std::string> AiuptiActivityProfilerSession::correlateRuntimeOps_ = {
"launchCb",
"launchComputeStream"};
"aiuLaunchControlBlocks"
};

// =========== Session Constructor ============= //
AiuptiActivityProfilerSession::AiuptiActivityProfilerSession(
Expand Down