7e6a73959a
Increase fp16 support to allow clspv to continue to be OpenCL compliant following the update of the OpenCL-CTS adding more testing on math functions and conversions with half. Math functions are implemented by upscaling to fp32 and using the fp32 implementation. It garantees the accuracy required for half-precision float-point by the CTS.
27 lines
857 B
C++
27 lines
857 B
C++
#include <utils.h>
|
|
|
|
#define __CLC_SW_FUNC(x) __CLC_CONCAT(__clc_, x)
|
|
|
|
#if __CLC_FPSIZE > 16
|
|
_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE __CLC_FUNC(__CLC_GENTYPE x, __CLC_GENTYPE y) {
|
|
return __CLC_SW_FUNC(__CLC_FUNC)(x, y);
|
|
}
|
|
#elif __CLC_FPSIZE == 16
|
|
#ifdef __CLC_SCALAR
|
|
_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE __CLC_FUNC(__CLC_GENTYPE x,
|
|
__CLC_GENTYPE y) {
|
|
return convert_half(
|
|
__CLC_SW_FUNC(__CLC_FUNC)(convert_float(x), convert_float(y)));
|
|
}
|
|
#else
|
|
_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE __CLC_FUNC(__CLC_GENTYPE x,
|
|
__CLC_GENTYPE y) {
|
|
return __CLC_XCONCAT(convert_half, __CLC_VECSIZE)(__CLC_SW_FUNC(__CLC_FUNC)(
|
|
__CLC_XCONCAT(convert_float, __CLC_VECSIZE)(x),
|
|
__CLC_XCONCAT(convert_float, __CLC_VECSIZE)(y)));
|
|
}
|
|
#endif
|
|
#endif
|
|
|
|
#undef __CLC_SW_FUNC
|