28 |
|
|
29 |
has_sse2=f |
has_sse2=f |
30 |
grep flags /proc/cpuinfo | grep sse2 > /dev/null 2>&1 && has_sse2=t |
grep flags /proc/cpuinfo | grep sse2 > /dev/null 2>&1 && has_sse2=t |
31 |
if test "x$has_sse2" = xt ; then |
|
32 |
FFLAGS="$FFLAGS -Wimplicit -Wunused -mfpmath=sse -msse -msse2" |
FFLAGS="$FFLAGS -Wimplicit -Wunused" |
|
else |
|
|
FFLAGS="$FFLAGS -Wimplicit -Wunused -ffloat-store" |
|
|
fi |
|
33 |
|
|
34 |
if test "x$IEEE" = x ; then #- with optimisation: |
if test "x$IEEE" = x ; then #- with optimisation: |
35 |
|
#- note(jmc): on aces, sse & sse2 are faster; -O3 is not always faster than -O2 |
36 |
|
if test "x$has_sse2" = xt ; then |
37 |
|
FFLAGS="$FFLAGS -mfpmath=sse -msse -msse2" |
38 |
|
fi |
39 |
#FOPTIM='-O3 -malign-double -funroll-loops' |
#FOPTIM='-O3 -malign-double -funroll-loops' |
40 |
#- note(jmc): |
# Problems with "-funroll-loops" for some set-up, e.g. tutorial_held_suarez_cs |
|
# -O3 is not always faster than -O2 ; sse & sse2 are faster ; |
|
|
# Problems with "-funroll-loops" for some set-up, e.g. tutorial_held_suarez_cs |
|
41 |
FOPTIM='-O3 -malign-double' |
FOPTIM='-O3 -malign-double' |
42 |
else #- no optimisation + IEEE : |
else #- no optimisation + IEEE : |
43 |
|
if test "x$has_sse2" = xt ; then |
44 |
|
FFLAGS="$FFLAGS -mfpmath=sse -msse -msse2" |
45 |
|
else |
46 |
|
FFLAGS="$FFLAGS -ffloat-store" |
47 |
|
fi |
48 |
FOPTIM='-O0 -malign-double' |
FOPTIM='-O0 -malign-double' |
49 |
fi |
fi |
50 |
|
|
77 |
if [ -n "$MPI_INC_DIR" -a "x$MPI" = xtrue ] ; then |
if [ -n "$MPI_INC_DIR" -a "x$MPI" = xtrue ] ; then |
78 |
INCLUDES="$INCLUDES -I$MPI_INC_DIR" |
INCLUDES="$INCLUDES -I$MPI_INC_DIR" |
79 |
INCLUDEDIRS="$INCLUDEDIRS $MPI_INC_DIR" |
INCLUDEDIRS="$INCLUDEDIRS $MPI_INC_DIR" |
80 |
# MPIINCLUDEDIR="$MPI_INC_DIR" |
#- used for parallel (MPI) DIVA |
81 |
# MPI_HEADER_FILES='mpif.h mpiof.h' |
MPIINCLUDEDIR="$MPI_INC_DIR" |
82 |
# MPI_HEADER_FILES_INC='./mpi_headers/mpif.h ./mpi_headers/mpiof.h' |
#MPI_HEADER_FILES='mpif.h mpiof.h' |
83 |
fi |
fi |