Re: Using a DFA for insn-recog.c

[Paolo Bonzini <bonzini at gnu dot org>]

> I'm on the fence because the whole thing seems a bit of a wash at 
> present; not much improvement, not much harm, some risk just because 
> change is risky.  Does anyone else have an opinion as to whether or not 
> this patch is sufficiently compelling to include?

Personally, I think the patch is very interesting and it would be a pity 
to not get it in.  What happens if you compile insn-recog.o at -O1? 
Since time spent in recog is almost never measurable, it should improve 
bootstrap time and not make the compiler much slower.

Paolo