[cfe-commits] RFC: Half floating point support
eli.friedman at gmail.com
Thu Oct 6 14:25:32 PDT 2011
On Thu, Oct 6, 2011 at 2:16 PM, Anton Korobeynikov
<anton at korobeynikov.info> wrote:
> Hello Everyone,
>> If you follow John's advice and do the promotion in
>> UsualUnaryConversions, it should pick up this case without extra code.
> Attached is the second version of the patch. The differences are (I
> believe I addressed all the comments):
> 1. Added sema checks for __fp16 arguments and return types
> 2. "Main" promotion code moved to UsualUnaryConversions
> 3. Several special cases are handled separately (e.g. unary ops,
> half-to-bool conversions, etc)
> 4. Aggregate assignments now work (the intrinsic emission was moved
> inside EmitScalarConversion)
> 5. half initializers now use standard initialization logic
> Im finishing the 'feature' testsuite now.
> Ok to commit? (with testcases)
It looks like you attached the old patch.
More information about the cfe-commits