[PATCH] D20116: Add speculatable function attribute

[Chandler Carruth via llvm-commits]

While I suspect this patch is going in the right direction, I strongly
think this needs a clear RFC on the dev list before it goes in...

The closest I could find with searching for the word "speculatable" was an
over 60 email thread that talked about inaccessible state, and just seemed
to mention speculatable.

There is also the recent thread on llvm-dev by Sanjoy, where something
pretty much precisely the same as speculatable as a function attribute was
discussed.

I think that the dev list should get a clear summary of at least:
- What set of problems are being solved
- The nature of the attribute
- How it impacts existing optimizations and other attributes
- How it is expected to be leveraged going forward

Thanks,
-Chandler

On Fri, Jul 15, 2016 at 4:55 PM Hal Finkel via llvm-commits <
llvm-commits at lists.llvm.org> wrote:

> hfinkel added inline comments.
>
> ================
> Comment at: docs/LangRef.rst:1520
> @@ +1519,3 @@
> +    Note that ``speculatable`` is not enough to conclude that along any
> +    particular exection path the number of calls to this function will be
> +    externally observable.
> ----------------
> Do you mean "will not be"?
>
>
> https://reviews.llvm.org/D20116
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160716/89ccd5f7/attachment.html>