The heavy importance of the compiler in number crunching application is the
major cause for migration of the compilers' preprocessor, a critical
technological component of the supercomputer microcosm of the eighties, to
the mainstream of the RISC chip world, in the first half of the nineties.
Thanks to these preprocessors, and to other beautifiers and F90 translators,
the original source evolves into many derived versions available
here:
- the SPAG version comes from
's fortran source reorganizer (06/1988);
- the STRUCT version structured by FOR_STRUCT, v2.0, (s=v sp=ddegl v) (02/1992);
- the
- FOREST and
- FOREST90 versions come from ForeSys' fortran restructurer by ConneXité -
(08/1993);
- the
versions comes from
- the VAST versions comes from
- IBM xlf 3.1 (-Pv! -Wp,-l,-ea789), (translated by IBM AIX XL FORTRAN Preprocessor/6000, Version 02.01.0000.0000 3.03G17a (1991) or from
- Pacific-Sierra Research VAST-2, Version 03.01.0000.0000 4.03H24 (1994));
Yet, ATT's f2c among other less-able softwares, allows one more version of the original source: this translation, from f77 to C, although crude but very solid and incredibly reliable, starts new comparisons with C compilers in numerical processing and shows that GNU's g77 might be a real alternative for fortran programming without a commercial fortran compiler.
- the C version comes from a translation by ATT's f2c, ( "version of 27 June 1990 8:41:01" obtained from netlib@research.att.com) .
Next: The report's organization
Up: Fortran Execution Time
Previous: The benchmark's correctness