From the way benchmarks have been responding to recent PRs, it certainly seems like side exits could be made more efficient, especially in the happy path. This is especially true of _DYNAMIC_EXIT, which needs to do quite a bit of work for a successful stitch... perhaps the best option here is to just reduce the number of _DYNAMIC_EXITs that we run.
This could also include recompiling stitched traces together into the same executor to essentially remove stitching overhead, too.
From the way benchmarks have been responding to recent PRs, it certainly seems like side exits could be made more efficient, especially in the happy path. This is especially true of
_DYNAMIC_EXIT
, which needs to do quite a bit of work for a successful stitch... perhaps the best option here is to just reduce the number of_DYNAMIC_EXIT
s that we run.This could also include recompiling stitched traces together into the same executor to essentially remove stitching overhead, too.
Linked PRs