Message294743
> Any high-level benchmarks (i.e. other than pybench and similar programs) impacted by this?
Sorry, I don't have any data. Currently I'm not able to run large benchmarks. I would be surprised if this increases the performance of a macrobenchmark even by 1%. But this change is a step in the direction of getting rid from the peephole optimizer.
> Will this negatively impact code coverage reporting or code tracing (by optimizing across basic blocks)?
I can't imagine the case. The compiler doesn't generate separate line numbers neither for JUMP instructions, nor for auxilary stack manipulating instructions in chained comparison, nor for UNARY_NOT. And these are the only things that are changed. Line number marks are left the same in optimized code, and they points to the same instructions. I think that coverage tools and debugger are not affected. |
|
Date |
User |
Action |
Args |
2017-05-30 07:25:48 | serhiy.storchaka | set | recipients:
+ serhiy.storchaka, brett.cannon, rhettinger, ncoghlan, pitrou, benjamin.peterson, yselivanov, emilyemorehouse |
2017-05-30 07:25:48 | serhiy.storchaka | set | messageid: <1496129148.7.0.365545380907.issue30501@psf.upfronthosting.co.za> |
2017-05-30 07:25:48 | serhiy.storchaka | link | issue30501 messages |
2017-05-30 07:25:48 | serhiy.storchaka | create | |
|