This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Rework description of DFA scheduler's max_issue()
Eric Botcazou wrote:
>
> > I've reproduced it too. The major problem is in that some insns have
> > no reservations for hypersparc.
>
> Yes, all the 'ialu' instructions. IMHO this is overly simplified.
>
> > When I wrote the function I did not expect it. My algorithm should be more
> > bulletproof. I am going to submit the patch for the main line today or
> > tomorrow to solve the problem.
>
> Thanks. However I'd like to fix the problem on the 3.3 branch too, because
> GCC 3.3 simply can't bootstrap --with-cpu=hypersparc. What do you think is
> the most convenient way to do that on the branch? Tweaking a bit the DFA
> description ? Then I'll ask DaveM to enter the game because I have no docs on
> HyperSPARC.
I think it is better to constrain the number of tries in the same way
(tries_num should be global because the recursive algorithm is used).
This is the simplest solution therefore it is the best for a release
branch.
If you have difficulties with that, I could make the patch. If it is
not, I'll review your patch in any case.
Vlad