We have both the GNU and Intel compilers - the Intel compilers generally give the best performance. We also have the highly optimized Intel MKL math libraries - you can see significant performance improvements by linking to these libraries.
The GNU compilers are accessed by simply "g77" for FORTRAN, or "gcc" for C.
The Intel compilers (version 8.1) are "ifort" and "icc". You do need to have /opt/intel/intel_fc_80/bin and /opt/intel/intel_cc_80/bin in your path - this is usually defined in your ".cshrc" file by default. From your .cshrc file you may also want to source the file /opt/intel/intel_fc_80/bin/ifortvars.csh and/or /opt/intel/intel_cc_80/bin/iccvars.csh which will set up some variables that are often helpful if you use the Intel compilers (e.g. LD_LIBRARY_PATH)
There are a great many flags available for the Intel compiler (use "man ifort" or the manual) however, some common flags for FORTRAN are:
Compiler Flags
| Compiler Flag | Description |
|---|---|
| O3 | General high performance optimizations |
| fast | Aggressive optimizations, may slow down compilation |
| Xn -tpp7 | Xeon (P4) specific optimizations |
| ip | Inlining functions, -ipo produces optimization across source codes |
| align all | Align variables, common blocks, etc. on 8 byte boundaries |
Summary information on C or FORTRAN can be obtained via the command line by typing "man icc" or "man ifort", or you can browse the full "ifort" and "icc" manuals online.
Note that some of these optimizations may have an affect on the rounding errors, etc. so you should test the more aggressive ones carefully before large scale runs.
Look at the special libraries section for information about the installed scientific libraries.