Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
disable irinterp for interpreters with
may_optimize(...)=false
(#53580
) As discussed at <b8a0a39#commitcomment-139076159>, currently external abstract interpreter that configures `may_optimize` to return `false` may end up with the internal error from irinterp since it fundamentally required optimized IR but it currently assumes that all sources from cached `CodeInstance`s are optimized. This commit addresses the issue by incorporating a `may_optimize` check in `concrete_eval_eligible`, which in turn automatically disables irinterp for such interpreters. Although there were earlier discussions suggesting the revival of `codeinfo.inferred::Bool`, this commit does not need it, and I think this approach maintains the current state more cleanly. This should fix the error of `"inference"` benchmarks from BaseBenchmarks.jl.
- Loading branch information
144f58b
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@nanosoldier
runbenchmarks(ALL, isdaily = true)
144f58b
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here.
144f58b
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Inference took another large hit again in performance and allocations. It might be this commit, but I didn’t think this commit was supposed to affect this test too
But rest looked pretty good. Collections and push seemed much faster, though not clear what commit could have altered that