This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [MIPS] PR target/22209: Support TImode arithmetic on MIPS
- From: Richard Sandiford <richard at codesourcery dot com>
- To: Roger Sayle <roger at eyesopen dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 13 Feb 2006 20:00:18 +0000
- Subject: Re: [MIPS] PR target/22209: Support TImode arithmetic on MIPS
- References: <Pine.LNX.4.44.0602130841300.5091-100000@www.eyesopen.com>
Roger Sayle <roger@eyesopen.com> writes:
> On Mon, 13 Feb 2006, Richard Sandiford wrote:
>> * config/floatunditf.c: Use if __LDBL_MANT_DIG__ == 113.
>> * config/fixtfdi.c: Likewise.
>> * config/fixunstfdi.c: Likewise.
>> * config/floatditf.c: Likewise.
>
> These changes look good to me. Technically, libgcc isn't part of
> libbackend.a so not covered by my middle-end maintainer rights, but
> as a MIPS port maintainer you can approve this patch yourself (as
> it only affects MIPS targets). One might even claim these fixes are
> obvious.
OK, in that case, I'll hope I'm covered by at least one of those. ;)
Patch installed to trunk.
> Sorry for not realizing "long double" hand a different number of mantissa
> bits between IRIX and Linux/GNU on MIPS. I tried to avoid the original
> MIPS-specific ABI checks in _tilib.c, so that these files would be useful
> on other platforms (using __LDBL_MANT_DIG__ changes such as yours).
No prob, and good idea.
> p.s. Are you able to check where PR middle-end/24998 still affects
> MIPS16? It looks related and may have been resolved by these changes.
Not in any special position that others aren't. Joseph's talking about
mips16.S, which is the MIPS16 -mhard-float support code. FWIW, I've
never seen that code used, and I can't remember seeing a single patch
related to it in the time that I've been working on MIPS gcc. The mips.c
support code seems to be quite subtle, and relies on all sorts of stub magic,
so I suspect it has bitrotted terribly over the years...
Still, it might be interesting to try it one rainy day. I think it
would mean hacking up a custom gdb simulator, as I don't recall any
out-of-the-box configurations combining mips16 and FPU support.
Richard