Date: Sat, 15 Jan 2005 09:40:53 +0100 From: Wilko Bulte <wb@freebie.xs4all.nl> To: John Baldwin <jhb@freebsd.org> Cc: freebsd-alpha@freebsd.org Subject: Re: processor type. Message-ID: <20050115084052.GA22832@freebie.xs4all.nl> In-Reply-To: <20050114224523.GA17854@freebie.xs4all.nl> References: <16866.32790.398095.651691@canoe.dclg.ca> <20050114202618.GA17050@freebie.xs4all.nl> <20050114210716.GA17430@freebie.xs4all.nl> <200501141632.37649.jhb@FreeBSD.org> <20050114224523.GA17854@freebie.xs4all.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Jan 14, 2005 at 11:45:23PM +0100, Wilko Bulte wrote..
> On Fri, Jan 14, 2005 at 04:32:37PM -0500, John Baldwin wrote..
> > On Friday 14 January 2005 04:07 pm, Wilko Bulte wrote:
> > > On Fri, Jan 14, 2005 at 09:26:18PM +0100, Wilko Bulte wrote..
> > >
> > > > On Fri, Jan 14, 2005 at 02:25:07PM -0500, John Baldwin wrote..
> > > >
> > > > > On Friday 14 January 2005 01:26 pm, Wilko Bulte wrote:
> > > > > > On Fri, Jan 14, 2005 at 10:43:00AM -0500, John Baldwin wrote..
> > > > > >
> > > > > > > On Friday 14 January 2005 02:43 am, Wilko Bulte wrote:
> > > > > > > > On Thu, Jan 13, 2005 at 04:26:51PM -0800, David O'Brien wrote..
> > > > > > > >
> > > > > > > > > On Mon, Jan 10, 2005 at 08:16:06AM -0500, David Gilbert wrote:
> > > > > > > > > > I see in the compiler lines crawling by that gcc is asked to
> > > > > > > > > > optimize for 'EV5' while being compatible with 'EV4'. My
> > > > > > > > > > Alpha is an EV4 --- I'm wondering if I would see better
> > > > > > > > > > performance with a different flag there, but the gcc manual
> > > > > > > > > > doesn't even acknowledge
> > >
> > > ....
> > >
> > > > > on "common" machines while still supporting older models. There's
> > > > > certainly no harm in bumping the default -mtune to ev6 if we feel that
> > > > > most Alpha users are ev6 rather than ev5.
> > > >
> > > > Hm, I suppose I could do this in HEAD.
> > > >
> > > > I'll run an experiment with a worldstone built using an ev5 world and
> > > > compare that to a worldstone built using an ev6 world. Stay tuned.
> > >
> > > Things are not entirely clean it seems on RELENG_4_11 and probably RELENG_4
> > > as well:
> > >
> > > cc -c -O -pipe -mcpu=ev6 -Wall -Wredundant-decls -Wnested-externs
> > > -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Winline
> > > -Wcast-qual -fformat-extensions -ansi -nostdinc -I- -I. -I/usr/src/sys
> > > -I/usr/src/sys/../include -I/usr/src/sys/contrib/ipfilter -D_KERNEL
> > > -include opt_global.h -mno-fp-regs -Wa,-mev56
> > > /usr/src/sys/kern/uipc_syscalls.c
> > > cc -c -O -pipe -mcpu=ev6 -Wall -Wredundant-decls -Wnested-externs
> > > -Wstrict-prototypes -Wmissing-prototypes -Wpointer-arith -Winline
> > > -Wcast-qual -fformat-extensions -ansi -nostdinc -I- -I. -I/usr/src/sys
> > > -I/usr/src/sys/../include -I/usr/src/sys/contrib/ipfilter -D_KERNEL
> > > -include opt_global.h -mno-fp-regs -Wa,-mev56
> > > /usr/src/sys/kern/uipc_usrreq.c
> > > {standard input}: Assembler messages:
> > > {standard input}:2145: Error: opcode `minsw4' not supported for target ev56
> > > *** Error code 1
> > >
> > > Stop in /usr/obj/usr/src/sys/DS10.
> > >
> > > This -mev56 should obviously not be there. Duh..
> > >
> > > The buildworld itself ran fine with CPUTYPE=ev6 in /etc/make.conf
> >
> > I can't find where the -mev56 comes from, but I have used CPUTYPE=ev6 in my
> > DS20 for a long time on -CURRENT without a problem.
>
> Looks like
>
> NO_CPU_COPTFLAGS=true # Don't add -march=<cpu> to COPTFLAGS automatically
>
> fixes it on RELENG_4
Well, the elapsed wall clock time for a buildworld running an ev5 world
versus running an ev6 world is only roughly 2 or 3 minutes in favor of a ev6
world on a elapsed time of roughly 75 minutes. This is on a DS10 600MHz
Not too interesting in my book
Wilko
--
Wilko Bulte wilko@FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050115084052.GA22832>