[llvm-commits] [llvm] r47758 - in /llvm/trunk: lib/CodeGen/RegAllocLocal.cpp test/CodeGen/ARM/2008-02-29-RegAllocLocal.ll

[Bill Wendling]

On Fri, Feb 29, 2008 at 3:00 PM, Chris Lattner <clattner at apple.com> wrote:
>
>  On Feb 29, 2008, at 1:34 PM, Bill Wendling wrote:
>
>  > On Fri, Feb 29, 2008 at 1:22 PM, Chris Lattner <clattner at apple.com>
>  > wrote:
>  >>
>  >> On Feb 29, 2008, at 11:19 AM, Bill Wendling wrote:
>  >>
>  >>> On Fri, Feb 29, 2008 at 11:15 AM, Anton Korobeynikov
>  >>> <asl at math.spbu.ru> wrote:
>  >>>> Hello, Bill
>  >>>>
>  >>>>
>  >>>>> + at _ZL20__gthrw_pthread_oncePiPFvvE = alias weak i32 (i32*, void
>  >>>>> ()*)* @pthread_once           ; <i32 (i32*, void ()*)*> [#uses=0]
>  >>>>> + at _ZL27__gthrw_pthread_getspecificj = alias weak i8* (i32)*
>  >>>>> @pthread_getspecific              ; <i8* (i32)*> [#uses=0]
>  >>>> What's about DCE'ing this testcase little bit? :)
>  >>>>
>  >>> Hi Anton,
>  >>
>  >>>
>  >>> This is the output from bugpoint. ;-) We should probably modify
>  >>> bugpoint to do this for us. Anyway, sure. I'll give it a shot.
>  >>
>  >> If it is a single function, llvm-extract is good at doing this, just
>  >> use llvm-extract -func whatever foo.bc
>  >>
>  > Good point. Anton also suggested running opt with -dce,
>  > -strip-dead-prototypes, -deadtypeelim, etc.
>
>  llvmextract runs strip-dead-prototypes, -deadtypeelim.
>
>  Usually dce is bad because it removes instructions that are important
>  to the failure.  Did you verify that you didn't break the testcase by
>  cleaning it up?
>
Actually, I cleaned it up by hand. But, yes, it still crashed without the fix.

-bw