This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: RFC: Patch to libgcc2.h
- To: Jason Merrill <jason at redhat dot com>
- Subject: Re: RFC: Patch to libgcc2.h
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Fri, 04 Aug 2000 13:48:53 -0600
- cc: gcc-patches at gcc dot gnu dot org
- Reply-To: law at cygnus dot com
In message <200006151625.JAA05704@casey.cygnus.com>you write:
> Jakub's December 27 patch to libgcc2.c broke a lot of libgcc bits for
> 16-bit targets; the #defines to change, say, the definition of
> __floatdisf to actually define __floatsisf produced conflicts with
> better versions from libgcc1 and fp-bit.c. In the case I've been
> looking at, that definition was used rather than the one from
> fp-bit.c, and so any code that needed to use __floatsisf ended up
> going into an infinite recursion.
>
> It's wrong to assume that things just scale down to smaller word
> sizes. Or perhaps it's right, but libgcc1 (in all its various
> incarnations) and fp-bit.c and the compiler optabs need to be adjusted
> as well.
>
> This patch returns us to the state of affairs from before
> Jakub's patch, namely that 16-bit targets again get definitions of
> dimode functions that they'll never use, but gcc seems perfectly happy
> to generate code for them, at least for the h8300 (which, admittedly,
> is not a pure 16-bit part).
>
> Any other thoughts as to how to solve this problem?
>
> 2000-06-14 Jason Merrill <jason@redhat.com>
>
> * libgcc2.h: Lose special handling of 8 and 16-bit targets.
Did we ever decide on a fix for this problem? I think 16bit targets are
still blowing up because of libgcc2 issues.
jeff