[LLVMdev] Lots of regtest failures on PPC64/Linux

[Hal Finkel]

----- Original Message -----
> From: "İsmail Dönmez" <ismail at donmez.ws>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: "LLVM Developers Mailing List" <llvmdev at cs.uiuc.edu>
> Sent: Wednesday, March 26, 2014 8:45:11 AM
> Subject: Re: Lots of regtest failures on PPC64/Linux
> 
> 
> Hi,
> 
> 
> 
> On Wed, Mar 26, 2014 at 3:26 PM, Hal Finkel < hfinkel at anl.gov >
> wrote:
> 
> 
> 
> ----- Original Message -----
> > From: "İsmail Dönmez" < ismail at donmez.ws >
> 
> 
> > To: "Hal Finkel" < hfinkel at anl.gov >
> > Cc: "LLVM Developers Mailing List" < llvmdev at cs.uiuc.edu >
> > Sent: Wednesday, March 26, 2014 8:20:31 AM
> > Subject: Re: Lots of regtest failures on PPC64/Linux
> > 
> > 
> > Hi Hal,
> > 
> > 
> > 
> > On Wed, Mar 26, 2014 at 3:17 PM, Hal Finkel < hfinkel at anl.gov >
> > wrote:
> > 
> > 
> > İsmail,
> > 
> > These are self-hosted builds? It seems like a lot of crashes in
> > llvm::sys::AtomicIncrement.
> > 
> > 
> > Yes. stage1 clang is used in stage2.
> 
> Okay; I don't know particularly, but do you know about when things
> started failing? These are Release builds? with -mcpu=ppc64 or
> -mcpu=native?
> 
> 
> 
> Actually no -mcpu flag is used hence its clang's whatever default.
> And the regression introduced in the last ~7 days or so.

Okay, interesting. I also did a build with -mcpu=ppc64 and did not observe these issues. What standard C++ library are you building against?

 -Hal

> 
> 
> Thanks.
> 
> 
> 

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory