Re: passing regparm floating-point args in FPU registers

[Jim Wilson <wilson at tuliptree dot org>]

Andrew Zabolotny wrote:
> So, now I don't get those 'fld NaN' at the beginning of the function,
> but instead I'm getting random ICE's in different situations (ex. in
> this function: float doit (float x, float y) { return y; }).

There isn't enough info here to help.  We would probably need all of the 
patches you have written.  Since this feature is of limited interest, 
you probably won't find anyone willing to help you for free.  Try 
debugging gcc, or giving more specific info, like stack backtraces for 
the ICEs.  If you can identify a specific gcc bug, then we may be able 
to help.

When showing us patches, you should use "cvs diff" or "diff", rather 
than just showing us your current code.

> And another small problem: _Optlink requires that those args that are
> passed in registers should have a equivalent space reserved on the
> stack.

See REG_PARM_STACK_SPACE.  You may also have to define related macros, 
or tweak other parts of the calling convention support, in order to get 
this working.  See also the code in function.c and calls.c which handle 
receiving function arguments and passing function arguments respectively.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com