Fix OpenCL float/double conversion usage.
https://www.khronos.org/registry/cl/sdk/1.0/docs/man/xhtml/tan.html
https://www.khronos.org/registry/cl/sdk/1.0/docs/man/xhtml/cl_khr_fp64.html
Differential D2188
Fix 'warning: double precision constant requires cl_khr_fp64, casting to single precision' in Cycles Authored by Edward O'Callaghan (funfunctor) on Sep 2 2016, 2:13 PM.
Details Fix OpenCL float/double conversion usage. https://www.khronos.org/registry/cl/sdk/1.0/docs/man/xhtml/tan.html
Diff Detail Event TimelineComment Actions We don't intentionally use doubles in the kernel. Rather than enabling this extension, we probably just need to change a few constants or functions to use the f postfix. Comment Actions Can I leave you with that? My concern is the opensource radeon drivers so (mesa/clover, llvm & kernel). Comment Actions Ok, I've fixed a couple of those cases in rBe76e8fcdcc53. If there's more please provide a build log with the warnings. Comment Actions Thanks!
Yea no worries. |