This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: cr logical insn implementation for rs6000
- To: dje at watson dot ibm dot com
- Subject: Re: cr logical insn implementation for rs6000
- From: Geoff Keating <geoffk at cygnus dot com>
- Date: Wed, 6 Sep 2000 14:11:26 -0700
- CC: gcc-patches at gcc dot gnu dot org
- References: <200009062019.QAA23812@mal-ach.watson.ibm.com>
> cc: gcc-patches@gcc.gnu.org
> Date: Wed, 06 Sep 2000 16:19:31 -0400
> From: David Edelsohn <dje@watson.ibm.com>
>
> >>>>> Geoff Keating writes:
>
> Geoff> They're execute-synchronized according to the docs on the 604 and 750.
> Geoff> I expect that they will usually have a latency of 3-4 cycles, as it'll
> Geoff> take 2-3 cycles for the pipeline to flush. There doesn't seem to be
> Geoff> an easy way to explain execute-synchronization to the old scheduler,
> Geoff> and I'm unwilling to spend much time on it when it'll all go away RSN
> Geoff> with the DFA-based pipeline descriptions.
>
> The execution serialization of those instructions on the 604 and
> 750 cannot be generalized. On the processors with a separate CR Logic
> Unit, they are not serialized -- only those where they are in the Branch
> Unit have the serialization. I will fix this as well.
>
> What is the new DFA-based pipeline scheduler?
It's a more accurate way of describing instruction timings. It can
cope with things like function unit dependencies, limits on
instruction dispatch and/or completion, and so on. Of course, it
needs much more detailed information than the current scheduling model
to be done right.
> Geoff> If you happen to have a POWER3 data book handy, one that describes the
> Geoff> instruction timings in detail, that'd be very useful. Especially in
> Geoff> PDF form...
>
> Unfortunately, the documents are IBM Confidential. I work across
> the hall from the person who maintains the documents for all IBM PowerPC
> architectures, so it is easy for me to look up any details for any IBM
> chip implementation.
Hmmm. I probably couldn't use it then anyway. I'll leave the
scheduling for the IBM-specific chips to you then, as you're best
placed to deal with IBM confidentiality rules.
--
- Geoffrey Keating <geoffk@cygnus.com>