Re: define "internal" visibility for i386

[Richard Henderson <rth at redhat dot com>]

On Sun, Mar 03, 2002 at 12:17:38PM +0100, Jan Hubicka wrote:
> Shouldn't we also modify the call instruction pattern to call the
> function directly w/o PIC tables?

The linker will take care of that already.

> But about the i386, that is very symetric about what can be load and
> thus it is manageable to make visible all the loads before reload, perhaps we
> can just rely on reload to not introduce any and unfix the register before
> register allocation if no static memory storage references are present?

I think this is dangerous.  Much better to fix this properly
and teach reload about allocating the pic register.


r~