This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Testing optimisation options on P4


I've built a little perl script to build-and-run flops.c on my P4/1300,
under gcc-3.1-20020204 with each of the 324 permutations of the following
options:

-O1 | -O2 | -O3
-fomit-frame-pointer | nothing
-mfpmath=sse | -mfpmath=387 | -mfpmath=sse,387
-march= {pentium3, pentium4, pentiumpro, athlon, athlon-xp, nothing}
-funroll-loops | -funroll-all-loops | nothing

As you might imagine, this takes a little while to run [and, yes, some of
the combinations of -mfpmath and -march are redundant, but I'll refine that
only if it's likely I'll be running the script more than once]; I'll report
any obvious observations on Monday, or over the weekend if I happen to go
into college.

Tom



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]