Date: Wed, 27 Mar 2002 08:33:54 -0500 From: "Brian F. Feldman" <green@FreeBSD.ORG> To: Alexander Leidinger <Alexander@Leidinger.net> Cc: green@FreeBSD.ORG, sw@anthologeek.net, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: cvs commit: ports/lang Makefile ports/lang/icc Makefile distinfo pkg-comment pkg-descr pkg-plist ports/lang/icc/files patch-include Message-ID: <200203271333.g2RDXsZ96421@green.bikeshed.org> In-Reply-To: Your message of "Wed, 27 Mar 2002 11:55:20 %2B0100." <200203271055.g2RAtLh3014236@Magelan.Leidinger.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Alexander Leidinger <Alexander@Leidinger.net> wrote: > On 26 M=E4r, Brian F. Feldman wrote: > = > >> > Can we build kernel with this (with gcc's help) ? > >> = > >> Feel free to try it and to send patches if it doesn't. :-) > > = > > I think we'd really have to move every bit of __asm__() out and only = use = > > actual .[Ss] files, or provide some alternative form of in-line assem= bly. > > I haven't determined if icc provides in-line assembly at all; even if= it = > > did, it would be totally backward from the syntax we currently use in= the = > > most basic ways, so it would be impossible to write to both of them. > = > icc understands parts of gcc specific attributes and the gcc assembly > syntax (at least parts of it, if I remember correctly). To use MS synta= x > you have to specify the "-use_msasm" option. So it may not be that > impossible to support both... > = > But the Linux kernel isn't icc compilable too, so it's not that much of= > a drawback... Hm, I didn't know it would actually use AT&T syntax instead of Intel! = That's definitely unexpected. > >> > What about servers like Apache (it would be great for busy servers= ) ? > >> = > >> I've seen this speed improvement in FPU intensive code (no MMX or SI= MD > >> used, plain FPU code, testet on an AMD Duron). I don't expect such a= > >> large speed improvement in "ordinary server code". And trust me, it > >> would be a lot of hassle to get apache compiled with icc. > > = > > I got Ruby compiled with icc some weeks back and I seem to recall aro= und a = > > 20% speed improvement, as well. An interpreter is a good benchmark f= or a = > > compiler's ability to generate both fast and correct code :) > = > Awesome. Now... what about "cd .../ruby; make -DUSE_ICC"? ;-) I didn't test it with FreeBSD headers building FreeBSD executables, so I'= ll = have to try that and make sure binary compatibility exists, too :) -- = Brian Fundakowski Feldman \'[ FreeBSD ]''''''''= ''\ <> green@FreeBSD.org <> bfeldman@tislabs.com \ The Power to Serv= e! \ Opinions expressed are my own. \,,,,,,,,,,,,,,,,,,= ,,,,\ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200203271333.g2RDXsZ96421>