For a few years now there've been some 'historical' compiler options on the Makefile. I've noticed 'unroll loops' may slow it down here, at least marginally, and someone else had said he noticed the same. The gcc manual reports it may or may not make it faster, I guess the 'not' got the better of it in some cases.
'fast math' is also historically there but the gcc manual reports it's excluded from -O3 due to reliability issues it may arise in certain cases. I know it had been tested for years but the gcc version had advanced a lot till then. [Can someone confirm it's fully reliable now?] Recently I went the 'safe' way and I was running builds with latest gcc version that had only -O3 and up to -march and -fpmath=sse. (IIRC omitting the frame pointer is done automatically on -O3 lately though that may need double checking) Anyway, what are your thoughts on it? Do you think other options are needed? What about the 'alignment' options? etc. _______________________________________________ ioquake3 mailing list [email protected] http://lists.ioquake.org/listinfo.cgi/ioquake3-ioquake.org By sending this message I agree to love ioquake3 and libsdl.
