From: Wu Y. <ad...@ne...> - 2003-04-07 02:00:13
|
Hi, Earnie, Notice my wording "One of my test programs are really showing now i586-targeted code is faster than i686-targeted code with gcc 3.2.2 on my Celeron 550MHz Thinkpad." Yes, in this case, "-mcpu=i686 -march=i686" is slower. It is a program searching for prime numbers. If no one had similar results as mine (after testing), I would not object to changing the tuning. But I doubt that (until someone gives me the benchmarking results). I do recommend setting conservatively and letting the performance-hungry people test for their most suitable settings. Best regards, Wu Yongwei Earnie Boyd wrote: > What about -mcpu=i686 -march=i386? That is what Danny is talking > about. > > Earnie. > > Wu Yongwei wrote: >> Any real benchmarking results? My past experience with gcc 2.95 >> showed "-mcpu=i686 -march=i686" did not necessarily produce faster >> executable. One of my test programs are really showing now >> i586-targeted code is faster than i686-targeted code with gcc 3.2.2 >> on my Celeron 550MHz Thinkpad. >> >> Of course, others may have different results so benchmarking >> results are welcome. >> >> Best regard, >> >> Wu Yongwei |