Skip to content

🔨 Add XPU runtime launch events #305

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

Draft
wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

🔨 Add XPU runtime launch events #305

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
58 changes: 57 additions & 1 deletion hta/common/trace_symbol_table.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -317,6 +317,60 @@ def get_operator_or_cuda_runtime_mask(self, df: pd.DataFrame) -> pd.Series:
| (df["cat"] == cuda_driver_id)
)

def _get_xpu_runtime_launch_events_mask(self, df: pd.DataFrame) -> pd.Series:
"""Returns a boolean mask you can use with pandas dataframes
to filter events that are XPU runtime kernel and memory operations."""

urEnqueueUSMFill_id = self.sym_index.get("urEnqueueUSMFill", self.NULL)
urEnqueueUSMFill2D_id = self.sym_index.get("urEnqueueUSMFill2D", self.NULL)
urEnqueueUSMMemcpy_id = self.sym_index.get("urEnqueueUSMMemcpy", self.NULL)
urEnqueueUSMMemcpy2D_id = self.sym_index.get("urEnqueueUSMMemcpy2D", self.NULL)

urEnqueueKernelLaunch_id = self.sym_index.get(
"urEnqueueKernelLaunch", self.NULL
)
urEnqueueKernelLaunchCustomExp_id = self.sym_index.get(
"urEnqueueKernelLaunchCustomExp", self.NULL
)
urEnqueueCooperativeKernelLaunchExp_id = self.sym_index.get(
"urEnqueueCooperativeKernelLaunchExp", self.NULL
)

urEnqueueMemBufferFill_id = self.sym_index.get(
"urEnqueueMemBufferFill", self.NULL
)
urEnqueueMemBufferRead_id = self.sym_index.get(
"urEnqueueMemBufferRead", self.NULL
)
urEnqueueMemBufferWrite_id = self.sym_index.get(
"urEnqueueMemBufferWrite", self.NULL
)
urEnqueueMemBufferCopy_id = self.sym_index.get(
"urEnqueueMemBufferCopy", self.NULL
)
urUSMHostAlloc_id = self.sym_index.get("urUSMHostAlloc", self.NULL)
urUSMSharedAlloc_id = self.sym_index.get("urUSMSharedAlloc", self.NULL)
urUSMDeviceAlloc_id = self.sym_index.get("urUSMDeviceAlloc", self.NULL)

name_mask = (
(df["name"] == urEnqueueUSMFill_id)
| (df["name"] == urEnqueueUSMFill2D_id)
| (df["name"] == urEnqueueUSMMemcpy_id)
| (df["name"] == urEnqueueUSMMemcpy2D_id)
| (df["name"] == urEnqueueKernelLaunch_id)
| (df["name"] == urEnqueueKernelLaunchCustomExp_id)
| (df["name"] == urEnqueueCooperativeKernelLaunchExp_id)
| (df["name"] == urEnqueueMemBufferFill_id)
| (df["name"] == urEnqueueMemBufferRead_id)
| (df["name"] == urEnqueueMemBufferWrite_id)
| (df["name"] == urEnqueueMemBufferCopy_id)
| (df["name"] == urUSMHostAlloc_id)
| (df["name"] == urUSMSharedAlloc_id)
| (df["name"] == urUSMDeviceAlloc_id)
)

return name_mask

def get_runtime_launch_events_mask(self, df: pd.DataFrame) -> pd.Series:
"""Returns a boolean mask you can use with pandas dataframes
to filter events that are CUDA runtime kernel and memcpy launches."""
Expand Down Expand Up @@ -353,8 +407,10 @@ def get_runtime_launch_events_mask(self, df: pd.DataFrame) -> pd.Series:
| (df["name"] == cuLaunchKernelEx_id)
)

xpu_name_mask = self._get_xpu_runtime_launch_events_mask(df)

# Add the index_correlation > 0 condition
return name_mask & (df["index_correlation"] > 0)
return (name_mask | xpu_name_mask) & (df["index_correlation"] > 0)

def get_events_mask(self, df: pd.DataFrame, events: list[str] | None) -> pd.Series:
"""Returns a boolean mask you can use with pandas dataframes
Expand Down