[Llvm-bgq-discuss] bgclang r185769-20130706 on vesta/mira

[Hal Finkel]

Erik,

Embarrassingly enough, you're right again! This has now been fixed.

A (-v2) archive is on the https://trac.alcf.anl.gov/projects/llvm-bgq/ page.

Thanks again,
Hal

----- Original Message -----
> Hal
> 
> I believe this new version still exhibits an error with the []
> operator.
> 
> My source code is in /home/eschnett/src/vecmathlib-qpx3. The source
> file is
> called test.cc. I built it with
> 
> /home/projects/llvm/bin/bgclang++
> -I/soft/compilers/ibmcmp-may2013/xlmass/bg/7.3/include -o test
> test.cc
> -L/soft/compilers/ibmcmp-may2013/xlmass/bg/7.3/bglib64 -lmass_simd
> 
> The output contains the lines
> 
> n=1
>   i=0 bits=1
>   i=1 bits=0
>   i=2 bits=0
>   i=3 bits=0
>   x=[1,0,1,0]
> 0x3ff0000000000000bff0000000000000bff0000000000000bff0000000000000
>   i=0
>   i=1
>   i=2
> Error in []:
>    x=[0,0,0,0]
> [0x3ff0000000000000bff0000000000000bff0000000000000bff0000000000000]
>    fstd(x)=0 [0x00]
>    fvml(x)=1 [0x01]
>    isbad(x)=1
>   i=3
> 
> which indicate that there is an error: for n=1, the value stored in x
> should be [-1.0, +1.0, +1.0 +1.0], corresponding to the bit pattern
> found
> in the lowermost 4 bits of n. The hex output indicates that this is
> the
> case, the bool output [1,0,1,0] indicates that something is wrong
> with the
> [] operator accessing these values.
> 
> The line "Error in []" below "i=2" indicates that bit 2 is found to
> be
> incorrect -- it is expected to be 0 (fstd), but found to be 1 (fvml).
> 
> -erik
> 
> 
> 
> 
> On Sun, Jul 7, 2013 at 12:43 AM, Hal Finkel <hfinkel at anl.gov> wrote:
> 
> > Hello,
> >
> > I've installed a new bgclang build (r185769-20130706) on vesta and
> > mira.
> > This build incorporates some recent upstream enhancements and bug
> > fixes.
> >
> > As usual, a new patch archive is available on:
> > http://trac.alcf.anl.gov/projects/llvm-bgq/
> >
> >  -Hal
> >
> > --
> > Hal Finkel
> > Assistant Computational Scientist
> > Leadership Computing Facility
> > Argonne National Laboratory
> > _______________________________________________
> > llvm-bgq-discuss mailing list
> > llvm-bgq-discuss at lists.alcf.anl.gov
> > https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> >
> 
> 
> 
> --
> Erik Schnetter <schnetter at cct.lsu.edu>
> http://www.perimeterinstitute.ca/personal/eschnetter/
> 

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