[Llvm-bgq-discuss] bgclang r209570-20140527 on vesta/mira

Hal Finkel hfinkel at anl.gov
Wed May 28 12:10:37 CDT 2014


Hi everyone,

As it turns out, the last two builds had a fairly severe code-generation bug that would cause miscompiles, and I've updated our bgclang build to fix the problem (the offending change has also been reverted upstream). If you've compiled anything with bgclang this week, I recommend you recompile it.

For those maintaining their own installs, the updated RPM (and SRPM) are available on the trac page: https://trac.alcf.anl.gov/projects/llvm-bgq.

I apologize for the inconvenience,
Hal

----- Original Message -----
> From: "Hal Finkel" <hfinkel at anl.gov>
> To: llvm-bgq-discuss at lists.alcf.anl.gov
> Sent: Tuesday, May 27, 2014 11:03:41 AM
> Subject: [Llvm-bgq-discuss]  bgclang r209570-20140527 on vesta/mira
> 
> Hello again,
> 
> I've updated our bgclang install on vesta and mira to
> r209570-20140527. This build is essentially the same as the last
> one, except for a fix to the remaining compiler crash reported by
> Phil Miller. However, in fixing this, I realized that the code
> quality for expressions like:
>   a_bool ? a_vector4double : another_vector4double;
> could be improved, and I implemented that improvement. As a result,
> the code generation changed for an entire class of expressions, not
> just the particular type of expression that caused the crash that
> Phil reported. To make it easier to isolate any regressions this
> might cause, I've created a separate build.
> 
> For those maintaining their own installs, RPMs (and SRPMs) are
> available on the trac page:
> https://trac.alcf.anl.gov/projects/llvm-bgq.
> 
> Thanks again,
> Hal
> 
> --
> Hal Finkel
> Assistant Computational Scientist
> Leadership Computing Facility
> Argonne National Laboratory
> 

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


More information about the llvm-bgq-discuss mailing list