Chandler Carruth
2014-Jan-21 22:51 UTC
[LLVMdev] Loop unrolling opportunity in SPEC's libquantum with profile info
On Tue, Jan 21, 2014 at 2:44 PM, Arnold Schwaighofer < aschwaighofer at apple.com> wrote:> The LoopVectorizer depends on LCSSA and LoopSimplify. Both are loop > passes. We will have to make them also available as utility functions.Yuck. We still need to fix these at least, but that's much better than teaching *all* the loop passes to preserve BPI and BFI. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140121/a2943357/attachment.html>
Andrew Trick
2014-Jan-21 22:55 UTC
[LLVMdev] Loop unrolling opportunity in SPEC's libquantum with profile info
On Jan 21, 2014, at 2:51 PM, Chandler Carruth <chandlerc at google.com> wrote:> > On Tue, Jan 21, 2014 at 2:44 PM, Arnold Schwaighofer <aschwaighofer at apple.com> wrote: > The LoopVectorizer depends on LCSSA and LoopSimplify. Both are loop passes. We will have to make them also available as utility functions. > > Yuck. We still need to fix these at least, but that's much better than teaching *all* the loop passes to preserve BPI and BFI.Or have function-level variants of these passes? -Andy -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140121/b90ac375/attachment.html>
Chandler Carruth
2014-Jan-21 23:16 UTC
[LLVMdev] Loop unrolling opportunity in SPEC's libquantum with profile info
On Tue, Jan 21, 2014 at 2:55 PM, Andrew Trick <atrick at apple.com> wrote:> On Jan 21, 2014, at 2:51 PM, Chandler Carruth <chandlerc at google.com> > wrote: > > > On Tue, Jan 21, 2014 at 2:44 PM, Arnold Schwaighofer < > aschwaighofer at apple.com> wrote: > >> The LoopVectorizer depends on LCSSA and LoopSimplify. Both are loop >> passes. We will have to make them also available as utility functions. > > > Yuck. We still need to fix these at least, but that's much better than > teaching *all* the loop passes to preserve BPI and BFI. > > > Or have function-level variants of these passes? >Yea, that's the other option. LCSSA and LoopSimplify are already really weird in that they are "dependencies" but are actually transform passes not analyses. Anyways, this should probably be a different thread... -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140121/5bab5728/attachment.html>