This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: patch for merging graphite branch (before tuplification)


Albert,
    The ppl developer says that ppl cvs is pretty well frozen except
for documentation fixes for the next ppl release in a couple weeks.
However I discovered that ppl cvs doesn't build with Apple's gcc 4.0.1
compilers. The author said that this was expected and has now set
gcc 4.0.3 as the minimum release for building the next ppl release.
Also he said that even gcc 4.0.2 had code generation bugs with ppl
so 4.0.3 is earliest usable compiler. The graphite developers might
want to consider this compiler restriction. On Darwin will we have
to use the Apple gcc-4.2 compiler to build the next ppl release
although they would imply that gcc 4.4.0 would only build on
darwin9. Otherwise we would have to resort to building ppl with FSF gcc 4.2
or 4.3 on darwin8. However I worry about that since the resulting
ppl would be linked to a different build of libgcc than that of the gcc
4.4.0 compiler it will be linked into.
                      Jack

On Mon, Aug 11, 2008 at 12:40:50PM +0200, Albert Cohen wrote:
> Sebastian is having holidays right now, but I am also 99.9% sure that
> there is no determinism issue in Graphite.
> 
> Albert
> 
> Mark Mitchell wrote:
> > Paolo Bonzini wrote:
> >>
> >>> Sebastian, will the behavior of Graphite be deterministic, in that
> >>> for a given compiler and a given input program, we can be assured of
> >>> the same output?  (I would assume so, but I would like to check.)
> >>
> >> Roberto said that the parts of PPL that are used in Graphite *do not*
> >> require floating point.  In fact, off list he even told me that some
> >> parts of PPL were written specially because other available packages
> >> used floating point.
> > 
> > OK, great!  I'd 99.9% sure it's not an issue, but I'd still like a
> > direct answer from Sebastian on the determinism issue.
> > 
> > Thanks,
> > 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]