Since we have the cargo bloat (#917), maybe we could do performance profiling in the same way with hotpath (which I just found out about!)
It profiles both latency and allocation, maybe this could be useful to pursue the perf goals here?
I'd set it up with a demo suite to be used to benchmark and cache the build as done for the cargo-bloat job.
The open question here is which pre-commit hooks to use as a 'standard' benchmark measure on...
- If you use a small number of built-in hooks then run-time will be dominated by setup
- If you use a larger number of external hooks then it'll be from running those
I'd pick 1 as it'd measure things directly in control of, and would likely be more sensitive (show small slowdowns as a bigger +% change overall), but open to others' feedback
Example
An example of it coming in handy from the author’s Twitter:
Since we have the cargo bloat (#917), maybe we could do performance profiling in the same way with hotpath (which I just found out about!)
It profiles both latency and allocation, maybe this could be useful to pursue the perf goals here?
I'd set it up with a demo suite to be used to benchmark and cache the build as done for the cargo-bloat job.
The open question here is which pre-commit hooks to use as a 'standard' benchmark measure on...
I'd pick 1 as it'd measure things directly in control of, and would likely be more sensitive (show small slowdowns as a bigger +% change overall), but open to others' feedback
Example
An example of it coming in handy from the author’s Twitter: