[PATCH] D98076: [OpenCL][Docs] Release 12.0 notes

[Sven van Haastregt via Phabricator via cfe-commits]

svenvh accepted this revision.
svenvh added inline comments.
This revision is now accepted and ready to land.


================
Comment at: clang/docs/ReleaseNotes.rst:214
+- Added ``global_device`` and ``global_host`` address spaces for USM
+  allocations.
+
----------------
Anastasia wrote:
> svenvh wrote:
> > Anastasia wrote:
> > > svenvh wrote:
> > > > Anastasia wrote:
> > > > > svenvh wrote:
> > > > > > Perhaps one more point to mention:
> > > > > > ```
> > > > > >  - Allow pointer-to-pointer kernel arguments beyond OpenCL 1.2.
> > > > > > ```
> > > > > > (related commit is 523775f96742 ("[OpenCL] Allow pointer-to-pointer kernel args beyond CL 1.2", 2020-12-01)).
> > > > > I have covered it in this but it is combined with another diagnostic:
> > > > > 
> > > > > 
> > > > > ```
> > > > > - Improved diagnostics for nested pointers and unevaluated ``vec_step`` expression.
> > > > > ```
> > > > > 
> > > > > But if you think it is better I can change as you have suggested.
> > > > Ah, that's probably why I missed it, as I was assuming "improved diagnostics" means a nicer error message, not that Clang now accepts something that it rejected before.  So in my opinion it deserves a point on its own.
> > > Ok, do you think we should use `nested pointers` instead of `pointer-to-pointer` because your change seems to work with any number of pointers. Although I appreciate that the latter one is a spec terminology.
> > No preference, but we could even mention both terms for completeness perhaps?
> Do you mean something like:
> 
> ```
> - Allow nested pointers (e.g. pointer-to-pointer) kernel arguments beyond OpenCL 1.2.
> ```
Looks good to me.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D98076/new/

https://reviews.llvm.org/D98076