From owner-freebsd-stable Sat Oct 28 0:32:25 2000 Delivered-To: freebsd-stable@freebsd.org Received: from krycek.zoominternet.net (krycek.zoominternet.net [63.67.120.9]) by hub.freebsd.org (Postfix) with SMTP id 7AD5F37B4CF for ; Sat, 28 Oct 2000 00:32:21 -0700 (PDT) Received: (qmail 16138 invoked from network); 28 Oct 2000 07:31:08 -0000 Received: from lcl12.cvzoom.net (208.226.155.12) by krycek.zoominternet.net with SMTP; 28 Oct 2000 07:31:08 -0000 Date: Sat, 28 Oct 2000 03:32:20 -0400 (EDT) From: Donn Miller To: "Jeffrey J. Mountin" Cc: stable@FreeBSD.ORG Subject: Re: "Malloc type lacks magic" show-stopper solved In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I forgot to finish. See inline. On Sat, 28 Oct 2000, Donn Miller wrote: > Well, I believe the "default" instruction set generated by gcc/gas is > i386. If you specifify -march=486, you will get i386+i486 instructions > generated, because i486 contains i386 instructions as a subset. But if > one specifies gcc -march=pentium, you will get i586 instructions, which > includes 586+486+386 machine code. > > I guess it depends on the code being compiled. If there is a case where > the asm code can collapse a series of i386 instructions into one i586 instruction, for example, the resulting code will be slightly smaller. But, I think the slight bloating by spec'ing a higher ARCH with -march=ARCH comes from the ARCH-specific instruction scheduling, which may pad the machine code on certain boundaries in memory to achieve optimum speed. In addtion, extra asm code is sometimes needed to take advantage of the pipelined architecture of the Pentium. - Donn To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message