IMHO, most optimization in ISA level (say sse) are from compiler and glibc, including some compiler options setting. But should not be fitting to both archs. Gfx benefits from this obviously, but many app also. Like you are guessing, both compiler and glibc saw the diff in arch and do thing differently. You can find detail from codes/patches.
-----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Glen Gray Sent: Tuesday, April 06, 2010 7:48 PM To: meego-dev Subject: Re: [MeeGo-dev] CPU Architectures (was Re: Build for a standard PC) Further to this topic on CPU Architectures, I've a few questions I'd like to ask for personal edification. How does the performance balance out across the architectures if this is a generic OS ? If ATOM benefits so heavily from SSSE3 instructions, how does ARM keep up. The same software can't be designed to utilize the speed of one arch and sacrifice performance on the other arch ? I know there are some SIMD instructions for ARM also (NEON ?) but I don't know how that compares nor what versions of ARM those are available on (not looked into it). How does the ssse3 optimization affect the core OS binaries ? It seems to me that the kernel and graphics libs are the potential big winners for any heavy usage fp math, but the rest of the system looses out. Unless something fundamental happens with gcc when it sees ssse3, e.g it implements optimized core functions like memcpy or something. How does a generic 686+sse2 build of meego compare to a core2+SSE3 optimized build ? Are there performance metrics that we can look at that influenced this decision ? Kind Regards, -- Glen Gray <[email protected]> _______________________________________________ MeeGo-dev mailing list [email protected] http://lists.meego.com/listinfo/meego-dev _______________________________________________ MeeGo-dev mailing list [email protected] http://lists.meego.com/listinfo/meego-dev
