RE: gcc 3.3.2/mips/mabi=64 undefined refs to __udivti3 and__umodti3

["Gary Funck" <gary at intrepid dot com>]

> -----Original Message-----
> From: gcc-owner@gcc.gnu.org [mailto:gcc-owner@gcc.gnu.org]On Behalf Of
> Eric Christopher
> Sent: Monday, October 25, 2004 4:53 PM
>
> On Mon, 2004-10-25 at 16:43, Gary Funck wrote:
> > Running some tests with -mabi=64 on an SGI Octane, Irix 6.5.
> > Working with GCC 3.3.2 baseline compiler. We've added some code that
> > creates new tree code and rtl; one of those mods may (or may not)
> > be triggering this behavior:
>
> Since you haven't provided your modifications or a testcase it's very
> hard to help you.
>

We're working on experimental dialect of C targeting distributed
shared memory computers, called UPC. Described here:
  http://intrepid.com/upc

Once I can find the source line that is causing the __udivti3
and __umodti3 calls to be emitted, I'll be glad to provide further
details on the tree and RTL code we're emitting.

> > - is __udivti3 and __umodti3 supported on mabi=64 SGI/MIPS targets?
>
> A few TImode functions are supported by the backend for TFmode emulated
> arithmetic, look at _tilib.c in config/mips/ in mainline sources.
>

OK. thanks. I thought I grep'd around in the CVS head tree and
didn't find those entries mentioned, but will look in more detail.

> > - should the MIPS code generator be emitting those library calls at all?
> > - what's the best/easiest method for finding out which source code line
> >   in the compiled program corresponds the generated RTL and/or assembly
> >   language?
>
> For mainline I'd suggest looking at the tree level dumps and figuring
> out which expands to the rtl that's generating the library calls.

Is there something in those dumps that will tell me the source and file
they were likely derived from? For a large program it can be problematic
trying to guess which RTL corresponds to a particular line in the
source.