Stuart Brady

cl_half is currently defined as a 16-bit integer, which allows code that might _appear_ to assign floating point values to a cl_half to compile and give the wrong result. Issue...

_The OpenCL Extension Specification_ § 38.10 _Clustered Reductions_ states: > The specified cluster size must be an integer constant expression that is a power-of-two. Behavior is undefined if the specified...