That option (-mfpmath=sse,387) was deemed unstable on the gcc page that Bruno posted on this thread. Quoting: "Use this option with care, as it is still experimental, because the GCC register allocator does not model separate functional units well resulting in instable performance. "
I think it is better to set sse on a USE flag so if a program is made to use them (i.e. X.org), it can, but it doesn't affect the other packages that weren't made with them in mind. 2005/6/13, Colin <[EMAIL PROTECTED]>: > Mark Shields wrote: > > >Oh, Raphael, concerning what you said: "By the way, look for something about > >using sse instruction for math. It speeds things up a bit.". I think you're > >speaking of "-mfpmath=sse". > > > You might want to specify -mfpmath=sse,387 instead. SSE is capable of > faster, more precise math than the 387 coprocessor. But not all > programs like/use SSE, so always specify a fallback. > > -- > Colin > > -- > gentoo-user@gentoo.org mailing list > > -- gentoo-user@gentoo.org mailing list
