This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: reload1: detect chained reloads
> If we actually allocate new reload registers there, that could be
> fatal. I can't see this in the source, though. AFAICT
> allocate_reload_reg calls set_reload_reg, which in turn calls
> mark_reload_reg_in_use, which then updates the appropriate regset.
If we were allocating our own reload register at that point, wouldn't
it then discard the recommendations from the first pass? It would
leave us no worse than we already are without my patch.
> Hmm. I wonder if for the mc68030, we might also see auto-increment
> reloads here.
> That's easy enough to check, though:
> if (rld[r1].out || rld[r2].out)
> return 0;
I added that.