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: Bounds-checking for Fortran testsuite was: Re: [gfortran, testsuite] Fix bounds checking for substrings, enable bounds chekcing in dg testsuite, assorted fixes


On Wed, Dec 13, 2006 at 06:45:19PM +0100, Tobias Schlüter wrote:
> Andrew Pinski <pinskia@gmail.com> wrote on Wed, 13 Dec 2006:
> >>My Tcl-fu is nil, but does this patch fix it?
> >
> >Most likely but this patch is not enough to keep some failures away if
> >the target GCC supports altivec but the machine does not.  Likewise for
> >SSE2.  FX's testresults show those failures.  Maybe adding the testing
> >of a trivial.f90 which contains a couple of vectorizable loops inside
> >fortran-torture.exp will fix that issue.
> 
> Hm, so why does this not happen with the vect testsuite?  I was
> precisely following its logic, which says "on i686, always _run_
> tests".  I could add something like check_vmx_hw_available, of course,
> I'm just wondering.

I think the initial error is a Tcl problem. I'm trying to improve my
Tcl-fu so I'll take a look at it.

There are two effective targets for AltiVec tests; powerpc_altivec_ok
says the test compiler can compile AltiVec code, and vmx_hw says that
the test system can run the test.  I've got PowerPC systems that do
and do not support VMX so I'll try these tests on both.

Janis


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