linux-mips
[Top] [All Lists]

Re: building an elf64 R10k kernel

To: Dave Olson <olson@anchor.engr.sgi.com>, Charles Lepple <clepple@foo.tho.org>
Subject: Re: building an elf64 R10k kernel
From: Ralf Baechle <ralf@uni-koblenz.de>
Date: Wed, 5 May 1999 00:16:06 +0200
Cc: linux@cthulhu.engr.sgi.com
In-reply-to: <199905040354.UAA16791@anchor.engr.sgi.com>; from Dave Olson on Mon, May 03, 1999 at 08:54:28PM -0700
References: <372E6AA0.505A6071@foo.tho.org> <199905040354.UAA16791@anchor.engr.sgi.com>
Sender: owner-linux@cthulhu.engr.sgi.com
On Mon, May 03, 1999 at 08:54:28PM -0700, Dave Olson wrote:

> Charles Lepple wrote: 
> |  Hey all,
> |  Does anyone out there know any details on building elf64 objects? I was
> |  all happy about seeing Andrew's Indigo2 patches, and decided that I
> |  _had_ to try and make it work on an Indigo2 Impact 10000... Suffice it
> |  to say that it isn't straightforward.
> 
> For r10k Indigo2 to work, you will need to hack the compiler,
> for various reasons (the way the r10k works, plus the fact that
> indigo2 doesn't have i/o cache coherency, interact in some "interesting"
> ways.  I would suggest not attempting this port, unless you have a *lot*
> of spare time.

Let me point out that SGI has invented an almost genious workaround for a
R10000 bug that only hits systems without I/O cache coherency, that is the
Indigo2 and O2.

When I first read the description of the bug and it's workaround I thought
I'd be halucinating ...

  Ralf

<Prev in Thread] Current Thread [Next in Thread>