Ahmed Bougacha
2015-Feb-27 23:44 UTC
[LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
On Fri, Feb 27, 2015 at 3:13 PM, Quentin Colombet <qcolombet at apple.com> wrote:> To be precise, GlobalMerge is registered as a pre-ISel pass, but still it runs very early in the pipeline, because all its work in done during doInitialization… Pretty broken, I know.Oh god, I forgot about this... it actually runs pretty early, not sure when exactly.. -Ahmed> -Quentin
Eric Christopher
2015-Feb-27 23:51 UTC
[LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
On Fri, Feb 27, 2015 at 3:48 PM Ahmed Bougacha <ahmed.bougacha at gmail.com> wrote:> On Fri, Feb 27, 2015 at 3:13 PM, Quentin Colombet <qcolombet at apple.com> > wrote: > > To be precise, GlobalMerge is registered as a pre-ISel pass, but still > it runs very early in the pipeline, because all its work in done during > doInitialization… Pretty broken, I know. > > Oh god, I forgot about this... it actually runs pretty early, not > sure when exactly.. > >During pass setup basically. And yeah, the pass itself needs to be turned into a real Module level pass. I'm also OK killing if someone is rewriting it :) -eric> -Ahmed > > > -Quentin > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150227/390d8bf6/attachment.html>
Ahmed Bougacha
2015-Feb-28 00:06 UTC
[LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
On Fri, Feb 27, 2015 at 3:51 PM, Eric Christopher <echristo at gmail.com> wrote:> > > On Fri, Feb 27, 2015 at 3:48 PM Ahmed Bougacha <ahmed.bougacha at gmail.com> > wrote: >> >> On Fri, Feb 27, 2015 at 3:13 PM, Quentin Colombet <qcolombet at apple.com> >> wrote: >> > To be precise, GlobalMerge is registered as a pre-ISel pass, but still >> > it runs very early in the pipeline, because all its work in done during >> > doInitialization… Pretty broken, I know. >> >> Oh god, I forgot about this... it actually runs pretty early, not >> sure when exactly.. >> > > During pass setup basically.Yeah, wasn't sure where that was. Looking at LTO, we do run some optimizations before setting up the CodeGen pipeline, but there's much more after that..> And yeah, the pass itself needs to be turned into a real Module level pass. > I'm also OK killing if someone is rewriting it :)The debug info problems seem pretty scary, but I'll try to have a second look! -Ahmed> -eric > >> >> -Ahmed >> >> > -Quentin >> >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
Anton Korobeynikov
2015-Feb-28 13:08 UTC
[LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
> And yeah, the pass itself needs to be turned into a real Module level pass. > I'm also OK killing if someone is rewriting it :)The only reason for this iirc is that we did not have any possibility to run Module level pass during early codegen stage - only function-level ones... -- With best regards, Anton Korobeynikov Faculty of Mathematics and Mechanics, Saint Petersburg State University
Reasonably Related Threads
- [LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
- [LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
- [LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
- [LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?
- [LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?