Skip to content

XProf GPU - Expose # of Chips to Profile #33182

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

Merged
merged 1 commit into from
Nov 10, 2025
+7 −1
Merged

XProf GPU - Expose # of Chips to Profile #33182

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
8 changes: 7 additions & 1 deletion docs/profiling.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -362,7 +362,13 @@ The following options are available for GPU profiling:
[NVIDIA's CUPTI documentation](https://docs.nvidia.com/cupti/main/main.html#metrics-table).
* `gpu_pm_sample_interval_us`: Sets the sampling interval in microseconds
for CUPTI PM sampling. Defaults to `500`.
* `gpu_pm_sample_buffer_size_per_gpu_mb`: Sets the system memory buffer size per device in MB for CUPTI PM sampling. Defaults to 64MB. The maximum supported value is 4GB.
* `gpu_pm_sample_buffer_size_per_gpu_mb`: Sets the system memory buffer size
per device in MB for CUPTI PM sampling. Defaults to 64MB. The maximum
supported value is 4GB.
* `gpu_num_chips_to_profile_per_task`: Specifies the number of GPU devices to
profile per task. If not specified, set to 0, or set to an invalid value,
all available GPUs will be profiled. This can be used to decrease the trace
collection size.
* `gpu_dump_graph_node_mapping`: If enabled, dumps CUDA graph node
mapping information into the trace. Defaults to `False`.

Expand Down
Loading