From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <775b8d190601190905i49095f52ga631ff633adde974@mail.gmail.com>
Date: Fri, 20 Jan 2006 04:05:42 +1100
From: Bruce Ellis <bruce.ellis@gmail.com>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
Subject: Re: [9fans] 5a and optimization
In-Reply-To: <386F3D23-0B29-4781-B9AC-87DB921101B7@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline
References: <acb178fec4c3be89cd56a9b7ef1ff45f@plan9.bell-labs.com>
	<386F3D23-0B29-4781-B9AC-87DB921101B7@gmail.com>
Topicbox-Message-UUID: e03bd076-ead0-11e9-9d60-3106f5b1d025

the mips compilers (at least 0c and ec etc.) support psuedo instrs
(NOSCHED, SCHED i think) which really stops your low level
code being screwed with.

brucee

On 1/20/06, Noah Evans <noah.evans@gmail.com> wrote:
> Yeah, I put it in l.s(probably should put it in rom.s but who knows).
> The linker eats the text if I leave the branch in. So I put the
> branch in a WORD and(thanks to Charles' advice) use -E to go to the
> initial TEXT _rom section.
>
> Very true about the rom weirdness.
>
> Thank you very much for the advice. I really appreciate the support.
>
> Noah
>
> On Jan 20, 2006, at 12:49 AM, jmk@plan9.bell-labs.com wrote:
>
> > One way is to make it TEXT, e.g. the multiboot header and GDT stuff
> > in /sys/src/9/pc/l.s.
> >
> > Agreed, the loaders don't make such things easy to do, but writing
> > assembler and making odd format binaries is the exception, not the
> > rule.
> >
> > --jim
> >
> > On Thu Jan 19 06:31:30 EST 2006, noah.evans@gmail.com wrote:
> >> How do you keep the linker from optimizing away your rom headers and
> >> other data? I had to put a noop in front of my first text section to
> >> keep the linker from putting the main loop in front of it. I also had
> >> to code the branch past the rom header in hex to keep the linker from
> >> eating the header(it's unreachable).
> >>
> >> Best,
> >>
> >> Noah
>
>