[PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-10-27 Thread David Majnemer via cfe-commits
This revision was automatically updated to reflect the committed changes. Closed by commit rL285316: [CodeGen] Provide an appropriate alignment for dynamic allocas (authored by majnemer). Changed prior to commit: https://reviews.llvm.org/D24378?vs=70789&id=76060#toc Repository: rL LLVM http

[PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-10-19 Thread Eli Friedman via cfe-commits
efriedma added a comment. Ping https://reviews.llvm.org/D24378 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Richard Smith via cfe-commits
On Fri, Sep 9, 2016 at 10:45 AM, Friedman, Eli via cfe-commits < cfe-commits@lists.llvm.org> wrote: > It probably makes sense to say that alloca should have the same alignment > as the default stack alignment, simply because it's hard to do anything > useful with completely unaligned memory. That

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Eli Friedman via cfe-commits
efriedma added a comment. Huh... then I guess this is fine. https://reviews.llvm.org/D24378 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Reid Kleckner via cfe-commits
rnk added a comment. In https://reviews.llvm.org/D24378#538430, @efriedma wrote: > This is probably going to lead to someone complaining about clang realigning > the stack on 32-bit Windows; are your sure we're choosing the right alignment > there? I checked, and MSVC emits a call to __alloca

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Friedman, Eli via cfe-commits
It probably makes sense to say that alloca should have the same alignment as the default stack alignment, simply because it's hard to do anything useful with completely unaligned memory. That said, the default stack alignment is 4 bytes on 32-bit Windows, not 16. -Eli On 9/9/2016 10:29 AM, R

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Richard Smith via cfe-commits
There's an (unconvincing to me) explanation for why gcc does this here: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=19131 On 9 Sep 2016 10:14 am, "Eli Friedman" wrote: > efriedma added a comment. > > This is probably going to lead to someone complaining about clang > realigning the stack on 32-

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Eli Friedman via cfe-commits
efriedma added a comment. This is probably going to lead to someone complaining about clang realigning the stack on 32-bit Windows; are your sure we're choosing the right alignment there? https://reviews.llvm.org/D24378 ___ cfe-commits mailing lis

Re: [PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-09 Thread Reid Kleckner via cfe-commits
rnk accepted this revision. rnk added a comment. This revision is now accepted and ready to land. lgtm https://reviews.llvm.org/D24378 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D24378: [CodeGen] Provide an appropriate alignment for dynamic allocas

2016-09-08 Thread David Majnemer via cfe-commits
majnemer created this revision. majnemer added reviewers: rnk, rsmith, efriedma, chandlerc. majnemer added a subscriber: cfe-commits. GCC documents __builtin_alloca as aligning the storage to at least __BIGGEST_ALIGNMENT__. MSVC documents essentially the same for the x64 ABI: https://msdn.microso