Subject: Re: borken state of quad/soft float on sparc64
To: James Chacon <>
From: Frederick Bruckman <>
List: tech-toolchain
Date: 02/04/2002 18:11:19
On Mon, 4 Feb 2002, James Chacon wrote:

> There isn't gonna be an easy fix here then (I went through this one day
> with sparc64). The missing function need to be added to libc, libc compiled up
> and installed so that floating point programs gcc emits those calls for will
> link correctly. Basically at this point the "host compiler is working" is no
> longer true due to these library calls. (which is also what bites the sparc64
> on perl at the moment).

I see. So this should only bite folks who are running new toolchain on
m68k, which is not yet the default on mac68k.

To do conversions to floatx80, it looks like you just have to fmove 0
into FPcr and then, fmove the argument into an FP register. The
conversions from floatx80 are just fintrz's like the other conversions
from floats and doubles.

Where is the complete list of operations that gcc is going to require?
I'm grepping over the toolchain sources and not finding it.