[PATCH] D55023: OpenCL: Improve vector printf warnings

[Anastasia Stulova via Phabricator via cfe-commits]

Anastasia accepted this revision.
Anastasia added a comment.
This revision is now accepted and ready to land.

LGTM! Thanks!



================
Comment at: test/SemaOpenCL/printf-format-strings.cl:65
+{
+    printf("%v4f\n", arg); // expected-warning {{format specifies type 'double __attribute__((ext_vector_type(4)))' but the argument has type 'float2' (vector of 2 'float' values)}}
 }
----------------
arsenm wrote:
> Anastasia wrote:
> > arsenm wrote:
> > > arsenm wrote:
> > > > arsenm wrote:
> > > > > Anastasia wrote:
> > > > > > So there is no way to print vector of float? What will happen on architectures that don't support doubles?
> > > > > > 
> > > > > > I guess it's the same for printf in general with the float type?
> > > > > There is, it's converted to a vector of doubles. This case warns because the element count mismatches
> > > > The conversion is to float if doubles aren't supported
> > > The warning message is wrong though without doubles, and still says double.
> > Can we add this explicitly to the test just for a record? I guess the warning message can't be changed to print the type correctly?
> > 
> > Also would it make sense to test other builtin types: double, char...?
> This should be easy to fix, but the way to test for doubles is enabled is making this unnecessarily difficult. It requires threading OpenCLOptions through all of the relevant functions to check for cl_khr_fp64. Why is this separate from LangOptions?
I see. I think the problem is that it is set conditionally based on the targets in earlier CL versions. So it had to be a separated into target specific option. :(


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D55023/new/

https://reviews.llvm.org/D55023