Janis Johnson wrote:
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.
Ok, here is what helped me on darwin-ppc and linux-ppc, both 32-bit
targets:
Index: gcc/testsuite/lib/gfortran.exp
===================================================================
--- gcc/testsuite/lib/gfortran.exp (revision 119856)
+++ gcc/testsuite/lib/gfortran.exp (working copy)
@@ -25,6 +25,7 @@
load_lib prune.exp
load_lib gcc-defs.exp
load_lib target-libpath.exp
+load_lib target-supports.exp
#
# GFORTRAN_UNDER_TEST is the compiler under test.
Index: gcc/testsuite/lib/gfortran-dg.exp
===================================================================
--- gcc/testsuite/lib/gfortran-dg.exp (revision 119856)
+++ gcc/testsuite/lib/gfortran-dg.exp (working copy)
@@ -14,8 +14,8 @@
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
02110-1301, USA.
+load_lib gcc-dg.exp
load_lib fortran-torture.exp
-load_lib gcc-dg.exp
# Define gfortran callbacks for dg.exp.
i?86-apple-darwin* is also happy with.
I try to find out why libgomp fails.