[PATCH] D140271: [NFCI][llvm-exegesis] Benchmark: parallelize codegen (5x ... 8x less wallclock)
Roman Lebedev via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Wed Jan 11 08:37:11 PST 2023
lebedev.ri added a comment.
The reason is always the same - making iterative development cycles faster.
I would just like to point out that the analysis used to be 50x times slower originally.
It was borderline unusable. But sure, one rarely needs it, so perhaps that was fine.
I would like to note that it is still slow[er than needed]. And the linear algebra stuff,
which i'm hoping to revisit, will make it even slower.
This is exactly the same story here. We are spending massive amounts of time,
and not doing anything useful with it, when we could be doing something useful there,
like, dunno, making more measurements. Why do we want it to be slow?
For me that long runtime has historically been the pain point.
If that was a joke, it wasn't a funny one i'm afraid.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D140271/new/
https://reviews.llvm.org/D140271
More information about the llvm-commits
mailing list