[PATCH] D95117: [NewPM][opt] Run the "default" AA pipeline by default
Arthur Eubanks via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Jan 22 09:31:14 PST 2021
aeubanks added inline comments.
================
Comment at: llvm/tools/opt/NewPMDriver.cpp:351
AAManager AA;
- if (!AAPipeline.empty()) {
- assert(Passes.empty() &&
- "--aa-pipeline and -foo-pass should not both be specified");
+ if (Passes.empty()) {
if (auto Err = PB.parseAAPipeline(AA, AAPipeline)) {
----------------
markus wrote:
> Realizing I am a bit late to the party but this change here confuses me.
>
> Previously in https://reviews.llvm.org/D82488?id=273164#inline-758886 it was requested that "Could you assert AAPipeline and AA passes inside Passes don't co-exist?" but the assert added at that point was actually stronger making sure that `Passes` contains no passes at all.
>
> While I guess this code is equivalent to its predecessor in that sense I don't quite understand why this limitation is needed. Especially as it gets in the way for our use case.
>
> E.g. if I run
> ```
> opt -enable-new-pm -loop-vectorize input.ll
> ```
> I do not get any "default" AA at all.
>
> Could of course be our use case is unintended and dumb but still … :)
If you want to use the new PM, you should use the `-passes=loop-vectorize` flag, not `-loop-vectorize`.
`-enable-new-pm` is strictly for the new PM transition, not intended as a long term thing.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D95117/new/
https://reviews.llvm.org/D95117
More information about the llvm-commits
mailing list