This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Midlevel RTL part 1
- From: law at redhat dot com
- To: Jason Merrill <jason at redhat dot com>
- Cc: Jan Hubicka <jh at suse dot cz>, rth at cygnus dot com, patches at x86-64 dot org, gcc-patches at gcc dot gnu dot org, gcc-pdo at atrey dot karlin dot mff dot cuni dot cz
- Date: Tue, 22 Jan 2002 09:43:48 -0700
- Subject: Re: Midlevel RTL part 1
- Reply-to: law at redhat dot com
In message <wvln0z6qrhu.fsf@prospero.cambridge.redhat.com>, Jason Merrill
write
s:
> > Yes. My patch already contain enum rtl_form that represent the form the
> > RTL is done in. Currently it is 3 stage - MID, lowering, LOW, but I plan
> > more passes. I do use relative comparisons to it to make adding more
> > of levels easy.
>
> Sounds good, but my question is, do we need to define levels at all? I'm
> not sure that lowering needs to be monotonic; optimization passes can be
> useful multiple times in the compilation, why not lowering passes?
My gut tells me that we're going to find that we'll eventually want to
lower certain aspects of the IR at different times. But I suspect once
we've lowered a certain set of constructs/concepts we won't need to do the
same lowering again.
Basically I don't really forsee a need to perform a specific lowering more
than once. But hey, maybe things will chance once we start using multiple
level IRs and we see specific issues.
jeff