This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: 2nd try for patch for automaton based pipeline hazard recognizer (part #1)
- To: Richard Henderson <rth at redhat dot com>
- Subject: Re: 2nd try for patch for automaton based pipeline hazard recognizer (part #1)
- From: Vladimir Makarov <vmakarov at redhat dot com>
- Date: Thu, 14 Jun 2001 17:09:36 -0400
- CC: Vladimir Makarov <vmakarov at toke dot toronto dot redhat dot com>, gcc-patches at gcc dot gnu dot org, geoffk at cygnus dot com
- References: <200106131859.OAA30718@toke.toronto.redhat.com> <20010614132854.B28982@redhat.com>
Richard Henderson wrote:
>
> On Wed, Jun 13, 2001 at 02:59:09PM -0400, Vladimir Makarov wrote:
> > (haifa_insn_data): Brace definitions `blockage' and `units' by
> > preprocessor conditionals with OLD_PIPELINE_INTERFACE.
>
> I wonder if it is possible to get rid of OLD_PIPELINE_INTERFACE.
> There seems to be quite a lot of code conditionalized on it.
>
Bernd wrote me about it. I tried to minimize it (in comparison with our
repository it looks much better).
> Would it be possible to interpret the old define_function_unit
> bits as a degenerate automata? Or do you think the over-use of
> ADJUST_COST makes this infeasable?
>
Yes, we could make process of translation old description to new one
automatic. We also could generate automaton from old description. The
current ADJUST_COST code could be and should be placed in automaton
based descriptions. It can not be done automatically.
I think that the process of transition from old descriptions to new
ones should be done manually also because of the old descriptions are
too simplified ones because the old model was not good.
> Assuming we can't get rid of the old interface, I wonder if it
> would be possible instead to split the code out of haifa-sched.c
> instead of conditionalizing it. It appears that there is quite
> a lot of stuff dependant on this.
I think it is a good point to design and extract this code. It would
simplify writing a new insn scheduler if we decide to rewrite the
interblock scheduler sometime.
I only would like to make one step at time.
Vlad