AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Using semaphore to sync with all peer processes in finalization stage (#169)

Browse Source 
* Using semaphore to sync with all peer processes in finalization stage

[rocprofv3] Implement synchronization using POSIX semaphore in finalization

* clang format code

* clang 11 format code

* Add process sync option for rocprofv3

* Default value of process sync is false

* Update source/lib/rocprofiler-sdk-tool/tool.cpp

Apply suggestion by Copilot

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* update according to comments

* add new line to helper.hpp

---------

Co-authored-by: Huanran Wang <huanrwan@amd.com>
Co-authored-by: Huanran Wang <huanran.wang@amd.com>
Co-authored-by: Madsen, Jonathan <Jonathan.Madsen@amd.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
systems-assistant[bot]
2025-08-25 08:57:41 -05:00
committed by GitHub
parent 748c9b74d9
commit b645010655
5 changed files with 142 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler-sdk/source/bin/rocprofv3.py
+13
View File
@@ -705,6 +705,16 @@ For MPI applications (or other job launchers such as SLURM), place rocprofv3 ins
Note: glog still installs signal handlers which provide backtraces""",
)
add_parser_bool_argument(
advanced_options,
"--process-sync",
help="""Enables the process synchronization in the rocprofv3 tool finalization stage.
When --process-sync is set to true,
and rocprofv3 tool will force process to wait for its peer processes finishing write the trace data,
then they proceed.
Note: some workloads will teminate the process group when one of the process is finished""",
)
advanced_options.add_argument(
"--minimum-output-data",
help="""Output files are generated only if output data size > minimum output data".
@@ -1509,6 +1519,9 @@ def run(app_args, args, **kwargs):
if args.disable_signal_handlers is not None:
update_env("ROCPROF_SIGNAL_HANDLERS", not args.disable_signal_handlers)
if args.process_sync is not None:
update_env("ROCPROF_PROCESS_SYNC", args.process_sync)
if args.minimum_output_data:
update_env("ROCPROF_MINIMUM_OUTPUT_BYTES", args.minimum_output_data * 1024)