Has anyone taken the time to look at the glibc memcpy implementation? At least for my platforms it has made a significant improvement over the asmarm version. (At least for me).
Craig On Monday 23 March 2009 6:11:03 am vince wrote: > Sven, > > I agree with you, I think this patch will help DirectFB, as it will make > it independent of the underline libc. This version of memcpy will only > for for little endian, but that could be changed in the future. In the > mean time, a simple config directive could be used to filter out > unsupported targets. I will also try to get a patch working for uClibc > (when I get some free time). > > Regards, > > Vince > > On Mon, 2009-03-23 at 10:16 +0100, Sven Neumann wrote: > > Hi, > > > > On Mon, 2009-03-23 at 09:54 +0100, Marc Titinger wrote: > > > on imx31 I got similar times with Vince's routine and a glibc-small > > > memcpy(), with a 4.1.2 g++ from codesourcery, (doing raw memcopies in > > > a test binary). > > > > So then it looks like the uClibC implementation of memcpy has some > > potential for improvement while the glibc implementation seems to > > provide performance comparable to Vince's implementation. > > > > I think we should still get this patch into DirectFB. Vince, you said > > that there might be some more tests needed as your code doesn't deal > > with all possible ARM platforms? > > > > > > Sven > > > > > > _______________________________________________ > > directfb-dev mailing list > > directfb-dev@directfb.org > > http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
_______________________________________________ directfb-dev mailing list directfb-dev@directfb.org http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
