On Fri, Sep 26, 2014 at 5:39 PM, reed kotler <rkotler at mips.com> wrote:
> Not that I think there is much chance that anyone else is thinking of
> submitting fast-isel patches for Mips but...
>
> I am very far a long in this port; almost to where the ARM port for
> example was when they first enabled fast-isel as the default for -O0.
> It's not really our plan to make this the default right now but I'm
just
> saying where we are.
>
> I'm starting to catch up on the patch push and review so hopefully I
will
> get it all in over the next few weeks.
>
> Just a warning to anyone wanting to add to what is there that it would for
> sure conflict with what is in my
> plan.
>
> I have reorganized it all now to be as close as possible to the AARch64
> fast-isel port which seems to be the
> cleanest to me and especially in call lowering and call return and some of
> the other more complex areas.
> This reorganization is a later one in my queue of patches but that is
> where it is all at right now internally at Mips.
>
> I'm so close to AARch64 port that I think that seriously in the future
> this could all be merged into target independent
> code with some small amount of virtual functions or other configuration
> methods.
>
> All the hard things have been implemented so my belief is that the rest
> will go quickly with the AArch64 as the main
> reference design; of course consulting the other ports too.
>
>
Not a bad idea, however, why aren't you doing this incrementally?
-eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140926/7d7bac89/attachment.html>