[Llvm-bgq-discuss] linker errors with llvm/current/bin/bgclang++

[Hal Finkel]

----- Geoffrey Irving <irving at naml.us> wrote:
> On Tue, Feb 5, 2013 at 9:17 AM, Hal Finkel <hfinkel at anl.gov> wrote:
> >
> > ----- Geoffrey Irving <irving at naml.us> wrote:
> >> On Mon, Feb 4, 2013 at 11:54 AM, Geoffrey Irving <irving at naml.us> wrote:
> >> > On Sun, Feb 3, 2013 at 9:07 AM, Hal Finkel <hfinkel at anl.gov> wrote:
> >> >> ----- Original Message -----
> >> >>> From: "Hal Finkel" <hfinkel at anl.gov>
> >> >>> To: "Geoffrey Irving" <irving at naml.us>
> >> >>> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> >> >>> Sent: Saturday, February 2, 2013 8:45:31 PM
> >> >>> Subject: Re: [Llvm-bgq-discuss] linker errors with llvm/current/bin/bgclang++
> >> >>>
> >> >>> ----- Original Message -----
> >> >>> > From: "Geoffrey Irving" <irving at naml.us>
> >> >>> > To: "Hal Finkel" <hfinkel at anl.gov>
> >> >>> > Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> >> >>> > Sent: Saturday, February 2, 2013 4:09:56 PM
> >> >>> > Subject: Re: [Llvm-bgq-discuss] linker errors with
> >> >>> > llvm/current/bin/bgclang++
> >> >>> >
> >> >>> > Alas, using the new clang I get a new compiler error. Output and
> >> >>> > preprocessed source here:
> >> >>>
> >> >>> Well, that's not good. I have some idea of where the problem lies,
> >> >>> and I'll fix it (or translate it into something that can be an
> >> >>> upstream bug report). Thanks for testing!
> >> >>
> >> >> As it turned out, this was not an upstream bug (meaning it was completely my fault). I think that particular bug is now fixed. Can you please try the build in /home/projects/llvm/r174287-20130202 to confirm.
> >> >
> >> > Cool. I'll give it a try once vesta comes back from maintenance.
> >> >
> >> > Geoffrey
> >>
> >> Confirmed: /home/projects/llvm/r174287-20130202/bin/clang++ works fine
> >> for me now, at least on the login nodes. Thanks!
> >>
> >
> > Great. You might also try r174287-20130203 which has fixes for some of the aforementioned performance regressions I've been investigating.
> 
> That also works. I've confirmed that it doesn't fix the __thread
> issue though; I'll isolate that for you tonight.
> 

Great, thanks! The __thread issue is just when compiling for the host? Or also when compiling for the compute nodes?

 -Hal

> Geoffrey

-- 
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory