From owner-freebsd-current Fri Oct 22 10:40:18 1999 Delivered-To: freebsd-current@freebsd.org Received: from palrel1.hp.com (palrel1.hp.com [156.153.255.242]) by hub.freebsd.org (Postfix) with ESMTP id 10CC814BD2 for ; Fri, 22 Oct 1999 10:40:16 -0700 (PDT) (envelope-from steveroo@mothra.bri.hp.com) Received: from mothra.bri.hp.com (mothra.bri.hp.com [15.144.1.185]) by palrel1.hp.com (Postfix) with ESMTP id 18F1C1E5; Fri, 22 Oct 1999 10:40:06 -0700 (PDT) Received: from localhost (steveroo@localhost) by mothra.bri.hp.com with ESMTP (8.8.6 (PHNE_17135)/8.7.1) id SAA02526; Fri, 22 Oct 1999 18:38:52 +0100 (BST) Date: Fri, 22 Oct 1999 18:38:52 +0100 (BST) From: Stephen Roome To: Dan Nelson Cc: Ville-Pertti Keinonen , current@FreeBSD.ORG Subject: Re: Anyone adding "support" for Athlons. In-Reply-To: <19991022095759.A16787@dan.emsphone.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Fri, 22 Oct 1999, Dan Nelson wrote: > In the last episode (Oct 22), Stephen Roome said: > > The issues I was trying to raise were really about better > > optimization for Athlons, and later generations of CPU's, as follows > > Optimization issues should be brought up on the egcs mailinglists, > since that is the compiler we use. I agree, although I'd still be concerned wether extra support needs to be added, for example to : /usr/src/sys/i386/i386/support.s (perhaps some Athlon optimized bzero/bcopy - or is it totally unnecessary ?) > > 2) Alignment. By default if code is compiled without -m486 we end up > > with different alignment strategies taken by gcc. Should the 386 > > strategy still be the default ? (i.e. Should 386 still be considered > > the expected hardware for FreeBSD) > > yes. Edit /etc/make.conf and add -march=i586 if you want to compile > your machine with pentium-specific code. I've not got an up to date version of gcc available at the moment (will get -current over the weekend however and see how things change) > > 3) Use of SIMD instructions, (MMX and 3DNow!). Are these used by > > anything ? > > Don't the MMX instructions stomp on the floating-point regs? Think so. Although gas can use them. (Not sure why we have two version of gas in the source tree though ?) -rw-r--r-- 1 root wheel 43061 Sep 6 1998 /usr/src/contrib/binutils/include/opcode/i386.h -rw-r--r-- 1 root wheel 34662 Mar 29 1998 /usr/src/gnu/usr.bin/as/opcode/i386.h One doesn't define the emms instruction and one does (this is, I believe, an mmx instr), it seems the default version does have this, so I can't figure what the other version is for. On the other hand. How about 3DNow! support ? although I've seen some patches for egcs to give this support on Athlons. (Again, I'll try this over the weekend with -current when I can install it..) But they don't yet seem to be in even the egcs cvs tree. (well they weren't on Wednesday) Steve To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message