GNU Fortran

N8TM@aol.com N8TM@aol.com
Fri Jan 22 20:34:00 GMT 1999


In a message dated 1/22/99 12:19:22 PM Pacific Standard Time, mrs@wrs.com
writes:

<< For many problems, G77 is faster, but when
 >    doing heavy duty math G77 is horrible. Seems to be the math library
 >    (sqrt, sin, cos, log, exp). >>
Most of these may be speeded up by using in-line intrinsics provided by the
target architecture.  This is not difficult to do, although how to do it
cleanly and provide for adding support for all desired targets is not so
obvious.  I've shown some suggestions for the complex math of libf2c in my
files on ftp://members.aol.com/n8tm .

Use of f2c with gcc and math-inline options for critical subroutines is a way
to test out the possibility of performance improvement.



More information about the Gcc-bugs mailing list