[llvm-commits] Cost Table, take 2

[Renato Golin Linaro]

On 24 January 2013 08:45, Chandler Carruth <chandlerc at google.com> wrote:

> At this point, I'm not really seeing the benefit of this patch.
>

As usual, the patch is changing its original intent from reviews, so it may
well be a mess because it's trying to accommodate everyone's ideas.

My original proposal was to *just* move the two cost structures out of x86
code, so that the ARM code could use it, but many constraints appeared,
including making it a full ADT type. So, it's bigger than it should because
it may be expecting to work on areas it wasn't originally planned.


1) A TargetCostTable class template in include/llvm/Target which accepts a
> T type parameter that can be either MVT, std::pair<MVT, MVT>, or
> std::tuple<MVT, MVT, ...>. Then the ISD and the cost value can be concrete
> things.
>

Yet another possible place, yet another slightly different implementation
detail... We're going in circles and I won't accommodate every one's
designs because they're all incompatible.

The efforts in changing this have already outweighed its costs 10-fold. I'm
tempted to just leaving as it is and move the two tables and their two
functions with generic names to lib/Target and call it a day.

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130124/e3feecb0/attachment.html>