Date: Mon, 11 Nov 2002 12:23:15 -0500 (EST) From: John Baldwin <jhb@FreeBSD.org> To: Poul-Henning Kamp <phk@critter.freebsd.dk> Cc: alpha@FreeBSD.ORG, Andrew Gallatin <gallatin@cs.duke.edu> Subject: Re: ev4/5/6 issue ? Message-ID: <XFMail.20021111122315.jhb@FreeBSD.org> In-Reply-To: <55809.1037029315@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On 11-Nov-2002 Poul-Henning Kamp wrote: > In message <15823.51765.171947.796322@grasshopper.cs.duke.edu>, Andrew Gallatin > writes: > >>Your initial assertion that this might be happening because the >>libraries are built for ev6 on beast could be true. Can you verify >>that that 12001f4e4 is indeed inside a library function using nm? > > Well, GDB said that much already. > >>That doesn't solve the initial problem of why your make was busted, >>though. If its in the libs, then your libs are likely busted too. >>Perhaps you'll also need to build a libc on beast, explicitly >>setting your cpuflags to ev4. Or perhaps the compiler is emitting >>FIX instructions when it should not. > > I think ev4 should be the default on beast. This release wasn't built on beast. It was built on a DS20 (which is an EV6). However, it should be built with '-mcpu=ev4 -mtune=ev6' so it should not be using any !ev4 instructions. Perhaps -mtune is broken? Hmm, actually, the whole release was compiled with '-mcpu=ev5 -mtune=ev5'. I do know that on alpha gcc "adjusts" itself to whatever CPU it is built on, so maybe because the gcc in the release make world was built on an EV6 it thinks it can emit EV6 instructions? Weird. -- John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.20021111122315.jhb>