[PATCH] D26593: [X86][GlobalISel] Add minimal call lowering support to the IRTranslator

[Davide Italiano via llvm-commits]

davide added a comment.

Picking this one (but the same applies to the ARM global isel skeleton commit).
I'm a little bit concerned to see various targets moving to global isel without proper scrutiny, as already pointed out by @echristo on llvm-dev. Don't get me wrong, I'm personally very excited about the whole idea, and it's likely that my organization will put some resources on it in the foreseeable future, but I would be less worried if Eric's concerns are addressed before moving forward.

In https://reviews.llvm.org/D26593#593922, @zvi wrote:

> In https://reviews.llvm.org/D26593#593918, @davide wrote:
>
> > Picking this one (but the same applies to the ARM global isel skeleton commit).
> >  I'm a little bit concerned to see various targets moving to global isel without proper scrutiny, as already pointed out by @echristo on llvm-dev. Don't get me wrong, I'm personally very excited about the whole idea, and it's likely that my organization will put some resources on it in the foreseeable future, but I would be less worried if Eric's concerns are addressed before moving forward.
>
>
> I believe that @echristo's concern was about running in a mixed mode of existing ISel (SelectionDAG) and Global ISel, but better let him speak for himself.
>  At the moment I am not aware of plans to run the X86 ISel in such a mixed mode. This patch is intended to jumpstart implemention of the X86 part of the Global Isel as designed by the team lead by Quentin and in par with the documentation. The development team presented the current status of the work on Global ISel at the recent dev meeting, and called out to developers of other targets to join this effort. We agreed that we could start development of the X86 part in low gear and speed-up after some additional work will be completed, mainly related to tablegen uses.


OK, I apologize if I missed the whole plan. So, is the goal that of growing this prototype into something ready for production? I remember there were some aspects that, although valuable, were voluntarily ignored for the prototype timeframe because there wasn't enough time (e.g. reuse of instcombine for the new selector).
I'm pretty sure there are others (which I don't have off the top of my head).
Are these aspects being reconsidered now (out of curiosity)?


Repository:
  rL LLVM

https://reviews.llvm.org/D26593