1 |
#!/bin/bash |
#!/bin/bash |
2 |
# |
# |
3 |
# composed and tested by ce107 on ross/weddell |
# Composed and tested by ce107 on ross/weddell (Opteron system) |
4 |
# for more speed on Core2 processors replace -xW with -xT |
# Should work fine on EM64T and other AMD64 compatible Intel systems |
5 |
# for more speed on Pentium4 based EM64T processors replaces -xW with -xP |
# a) Processor specific flags: |
6 |
|
# 1) for more speed on Core2 processors replace -xW with -xT |
7 |
|
# 2) for more speed on Pentium4 based EM64T processors replaces -xW with -xP |
8 |
|
# b) For more speed, provided your data size doesn't exceed 2GB you can |
9 |
|
# remove -fPIC which carries a performance penalty of 2-6%. |
10 |
|
# c) Provided that the libraries you link to are compiled with -fPIC this |
11 |
|
# optfile should work. |
12 |
|
# d) You can replace -fPIC with -mcmodel=medium which may perform faster |
13 |
|
# then -fPIC and still support data sizes over 2GB per process but all |
14 |
|
# the libraries you link to myst be compiled with -fPIC or -mcmodel=medium |
15 |
|
# e) Changed from -O3 to -O2 to avoid buggy Intel v.10 compilers. Speed |
16 |
|
# impact appears to be minimal. |
17 |
FC=ifort |
FC=ifort |
18 |
F90C=ifort |
F90C=ifort |
19 |
CC=icc |
CC=icc |
20 |
LINK='mpif90 -i-static -no-ipo' |
LINK='ifort -i-dynamic -no-ipo' |
21 |
|
|
22 |
DEFINES='-DWORDLENGTH=4' |
DEFINES='-DWORDLENGTH=4' |
23 |
CPP='cpp -traditional -P' |
CPP='cpp -traditional -P' |
33 |
if test "x$DEVEL" != x ; then |
if test "x$DEVEL" != x ; then |
34 |
FFLAGS='-132 -r8 -i4 -w95 -W0 -WB -convert big_endian -assume byterecl -fPIC -O0 -g -noalign -fpstkchk -check all -fpe0 -traceback -ftrapuv -fpmodel except -warn all' |
FFLAGS='-132 -r8 -i4 -w95 -W0 -WB -convert big_endian -assume byterecl -fPIC -O0 -g -noalign -fpstkchk -check all -fpe0 -traceback -ftrapuv -fpmodel except -warn all' |
35 |
else |
else |
36 |
FFLAGS='-132 -r8 -i4 -w95 -W0 -WB -convert big_endian -assume byterecli -fPIC' |
FFLAGS='-132 -r8 -i4 -w95 -W0 -WB -convert big_endian -assume byterecl -fPIC' |
37 |
fi |
fi |
38 |
|
|
39 |
# Note that the -mp switch is for ieee "maintain precision" and is |
# Note that the -mp switch is for ieee "maintain precision" and is |
40 |
# roughly equivalent to -ieee |
# roughly equivalent to -ieee |
41 |
if test "x$IEEE" = x ; then |
if test "x$IEEE" = x ; then |
42 |
# No need for IEEE-754 |
# No need for IEEE-754 |
43 |
FOPTIM='-O3 -align -xW -ip' |
FOPTIM='-O2 -align -xW -ip' |
44 |
else |
else |
45 |
# Try to follow IEEE-754 |
# Try to follow IEEE-754 |
46 |
FOPTIM='-O2 -align -xW -ip -mp' |
FOPTIM='-O2 -align -xW -ip -mp' |
47 |
fi |
fi |
48 |
F90FLAGS=$FFLAGS |
F90FLAGS=$FFLAGS |
49 |
F90OPTIM=$FOPTIM |
F90OPTIM=$FOPTIM |
50 |
CFLAGS='-O3 -ip -fPIC' |
CFLAGS='-O2 -ip -fPIC' |
51 |
|
|
52 |
if [ "x$NETCDF_ROOT" != x ] ; then |
if [ "x$NETCDF_ROOT" != x ] ; then |
53 |
INCLUDEDIRS="${NETCDF_ROOT}/include" |
INCLUDEDIRS="${NETCDF_ROOT}/include" |