This nights trunk fails (gcc of Aug 19th was still fine) with: gfortran -c -O3 -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form -D__GFORTRAN -D__FFTSG -g -v pw_spline_utils.F Using built-in specs. Target: x86_64-unknown-linux-gnu Configured with: /data/vondele/gcc_bench/gcc_trunk/gcc/configure --prefix=/data/vondele/gcc_bench/gcc_trunk/build --enable-languages=c,fortran Thread model: posix gcc version 4.4.0 20080823 (experimental) (GCC) COLLECT_GCC_OPTIONS='-c' '-O3' '-ffast-math' '-funroll-loops' '-ftree-vectorize' '-ffree-form' '-D__GFORTRAN' '-D__FFTSG' '-g' '-v' /data/vondele/gcc_bench/gcc_trunk/build/libexec/gcc/x86_64-unknown-linux-gnu/4.4.0/f951 pw_spline_utils.F -cpp /tmp/ccfcaaIN.f90 -quiet -v -D__GFORTRAN -D__FFTSG pw_spline_utils.F -march=core2 -mcx16 -msahf --param l1-cache-size=32 --param l1-cache-line-size=64 --param l2-cache-size=4096 -mtune=core2 -ffast-math -funroll-loops -ftree-vectorize -ffree-form -g -fworking-directory -O3 -march=core2 -mcx16 -msahf --param l1-cache-size=32 --param l1-cache-line-size=64 --param l2-cache-size=4096 -mtune=core2 -quiet -dumpbase pw_spline_utils.F -auxbase pw_spline_utils -g -O3 -version -ffast-math -funroll-loops -ftree-vectorize -ffree-form -fintrinsic-modules-path /data/vondele/gcc_bench/gcc_trunk/build/lib/gcc/x86_64-unknown-linux-gnu/4.4.0/finclude -o /tmp/cc2IAzfr.s ignoring nonexistent directory "/data/vondele/gcc_bench/gcc_trunk/build/lib/gcc/x86_64-unknown-linux-gnu/4.4.0/../../../../x86_64-unknown-linux-gnu/include" #include "..." search starts here: /data/vondele/gcc_bench/gcc_trunk/build/lib/gcc/x86_64-unknown-linux-gnu/4.4.0/finclude . #include <...> search starts here: /usr/local/include /data/vondele/gcc_bench/gcc_trunk/build/include /data/vondele/gcc_bench/gcc_trunk/build/lib/gcc/x86_64-unknown-linux-gnu/4.4.0/include /data/vondele/gcc_bench/gcc_trunk/build/lib/gcc/x86_64-unknown-linux-gnu/4.4.0/include-fixed /usr/include End of search list. GNU Fortran (GCC) version 4.4.0 20080823 (experimental) (x86_64-unknown-linux-gnu) compiled by GNU C version 4.4.0 20080823 (experimental), GMP version 4.2.2, MPFR version 2.3.1. GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096 pw_spline_utils.F: In function ‘T.974’: pw_spline_utils.F:1482: internal compiler error: in referenced_var_lookup, at tree-dfa.c:563 Please submit a full bug report, with preprocessed source if appropriate. See <http://gcc.gnu.org/bugs.html> for instructions. testcase to be attached
Created attachment 16137 [details] testcase untar and compile with gfortran -c -O3 -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-form -D__GFORTRAN -D__FFTSG -g -v pw_spline_utils.F
(In reply to comment #1) > Created an attachment (id=16137) [edit] > testcase > > untar and compile with > > gfortran -c -O3 -ffast-math -funroll-loops -ftree-vectorize -march=native > -ffree-form -D__GFORTRAN -D__FFTSG -g -v pw_spline_utils I can't reproduce this. What I get if I compile this using your command line is NU Fortran (GCC) version 4.4.0 20080823 (experimental) (x86_64-unknown-linux-gnu) compiled by GNU C version 4.4.0 20080823 (experimental), GMP version 4.2.1, MPFR version 2.3.1. GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096 Fatal Error: Reading module cell_types at line 35 column 30: Expected integer
(In reply to comment #2) > (In reply to comment #1) > > Created an attachment (id=16137) [edit] > > testcase > > > > untar and compile with > > > > gfortran -c -O3 -ffast-math -funroll-loops -ftree-vectorize -march=native > > -ffree-form -D__GFORTRAN -D__FFTSG -g -v pw_spline_utils > strange, I downloaded the tarbal, and it reproduced 'just fine'. Why would module files be non-compatible between different installation of gcc? The full CP2K sources are the ones used for the CP2K gfortran benchmark: ftp://ftp.berlios.de/pub/cp2k/gfortran/gcc_bench.tgz
Created attachment 16138 [details] Fortran only testcase compile as: gfortran -c -O3 -ffast-math -funroll-loops -ftree-vectorize -march=native -ffree-line-length-none PR37223.f90
I think this was fixed by reverting the patch which turned on IPA-CP by default. Also I think this was really fixed by http://gcc.gnu.org/ml/gcc-cvs/2008-08/msg01104.html.
today's trunk gets this right.