Add trace points in functions that depends on user input#9823
Conversation
|
Hi @stsoe, as discussed earlier added trace points only at places that depend on user input. |
5f6d084 to
3468a16
Compare
There was a problem hiding this comment.
Pull request overview
This PR adds additional runtime trace points around XRT object construction / configuration paths that are influenced by user input (e.g., ELF/module run creation and setting dtrace control file), to help diagnose performance bottlenecks via perf/DTrace-style tracing.
Changes:
- Added scoped trace points to AIE Gen2 / Gen2+ module run constructors.
- Added scoped trace points to AIE Gen2 / Gen2+ ELF constructors.
- Added a scoped trace point (and slightly reordered logic) in
run_impl::set_dtrace_control_file; also includes a few whitespace-only cleanups.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.
| File | Description |
|---|---|
| src/runtime_src/core/common/api/xrt_module.cpp | Adds scoped trace points in AIE module-run constructors. |
| src/runtime_src/core/common/api/xrt_kernel.cpp | Adds a scoped trace point in set_dtrace_control_file and refactors update ordering. |
| src/runtime_src/core/common/api/xrt_elf.cpp | Adds scoped trace points in AIE ELF constructors. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Signed-off-by: rahul <rbramand@amd.com>
3468a16 to
e74473b
Compare
|
clang-tidy review says "All clean, LGTM! 👍" |
|
clang-tidy review says "All clean, LGTM! 👍" |
Signed-off-by: rahul <rbramand@amd.com> (cherry picked from commit 307a964)
Problem solved by the commit
Added new trace points in xrt classes that depend on user input
During earlier debugs we identified bottlenecks while creating objects like xrt::run, xrt::elf etc, so added trace points at such places
Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered
How problem was solved, alternative solutions (if any) and why they were rejected
Added new trace points
Using tools like perf we can capture trace data and analyze bottlenecks
Risks (if any) associated the changes in the commit
Low
What has been tested and how, request additional testing if necessary
Tested trace data captured by running test on Telluride and things work as expected
Documentation impact (if any)
NA