The following code ices with the given options:
$ cat icey.f
SUBROUTINE IRCGS2(NC,QA,QB,QC,RTRMS)
IMPLICIT DOUBLE PRECISION(A-H,O-Z)
DIMENSION QA(NC), QB(NC), QC(NC), QCM(6000), QBM(6000), QAM(6000),
* RTRMS(NC)
IF (NC .GT. 6000) NC = 6000
DO 30 I = 1,NC
RTRMSI = RTRMS(I)
QAM(I) = QA(I) / RTRMSI
QBM(I) = QB(I) / RTRMSI
QCM(I) = QC(I) / RTRMSI
30 CONTINUE
RETURN
END
$ gfc -c -march=pentium4 -O -ffast-math -ftree-vectorize icey.f
icey.f:12: internal compiler error: in output_constant_pool_2, at varasm.c:3339
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
Adding -fno-automatic and there is no ice.
$ gfc -c -march=pentium4 -O -ffast-math -ftree-vectorize -fno-automatic icey.f
$
Using -march=k8 instead of -march=pentium4 also ices.
Getting rid of the -march= and the ice goes away.
$ gfc -v Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../main/configure --prefix=/home/jerry/gcc/usr
--enable-languages=c,fortran --disable-libmudflap
Thread model: posix
gcc version 4.2.0 20060323 (experimental)
--
Summary: ice in varasm.c with certain flags
Product: gcc
Version: 4.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: jvdelisle at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26821
