[PATCH] D13755: Pass to compute dynamic instruction count

Easwaran Raman via llvm-commits llvm-commits at lists.llvm.org
Mon Nov 2 16:47:26 PST 2015


I'm dropping the patch for now. The argument for making this as a pass is
to allow it to be shared by some client other than inliner (LoopUnroller?
something else?). But then, tying it to the cost model used by inliner
doesn't make sense. I can't rewrite this patch as a library now before
switching to the new pass manager, so I'll drop this now and revisit this
later.



On Tue, Oct 27, 2015 at 3:55 PM, hfinkel at anl.gov <hfinkel at anl.gov> wrote:

> hfinkel added inline comments.
>
> ================
> Comment at: include/llvm/Analysis/InlineCost.h:165
> @@ +164,3 @@
> +  // not.
> +  const bool Inlining;
> +
> ----------------
> I like the fact that the inlining-specific behavior can be explicitly
> disabled using this flag, but this needs to be better documented. At least
> give a few examples here of what kind of inlining-specific behavior this
> enables.
>
>
> Repository:
>   rL LLVM
>
> http://reviews.llvm.org/D13755
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151102/abb20d60/attachment.html>


More information about the llvm-commits mailing list