The following program will ICE when compiled with -O2 -fprefetch-loop-arrays on
a x86_64 box or on IA64 box (Linux and/or HP-UX in 64 bit mode).
scale_bbs_frequencies_int is called with nbbs=1, num=0, and den=0. Since we do
a division by den, the compiler aborts during the division. If we give X a
size of 2 instead of 1, the ICE does not happen.
$ cat x.f
SUBROUTINE TFHF(HINT,NVAR,NCOORD,LDI)
IMPLICIT DOUBLE PRECISION(A-H,O-Z)
DIMENSION HINT(LDI,NVAR)
COMMON /FMCOM / X(1)
LHI = 1 + LOADFM + (NCOORD*NVAR) + (NCOORD*NCOORD+NCOORD)/2
LOC = LHI-1
DO 160 I = 1,NVAR
DO 150 J = 1,I
LOC = LOC + 1
HINT(I,J) = X(LOC)
HINT(J,I) = X(LOC)
150 CONTINUE
160 CONTINUE
RETURN
END
$ gfortran -O2 -fprefetch-loop-arrays -c x.f
x.f: In function 'tfhf':
x.f:1: internal compiler error: Floating point exception
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
--
Summary: ICE with FORTRAN and -fprefetch-loop-arrays
Product: gcc
Version: 4.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: sje at cup dot hp dot com
GCC target triplet: x86_64-*-* ia64-*-*
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32087
