2.1. Half Arithmetic Constants
To use these constants, include the header file cuda_fp16.h in your program.
Macros
- CUDART_INF_FP16
-
Defines floating-point positive infinity value for the
halfdata type. - CUDART_MAX_NORMAL_FP16
-
Defines a maximum representable value for the
halfdata type. - CUDART_MIN_DENORM_FP16
-
Defines a minimum representable (denormalized) value for the
halfdata type. - CUDART_NAN_FP16
-
Defines canonical NaN value for the
halfdata type. - CUDART_NEG_ZERO_FP16
-
Defines a negative zero value for the
halfdata type. - CUDART_ONE_FP16
-
Defines a value of 1.0 for the
halfdata type. - CUDART_ZERO_FP16
-
Defines a positive zero value for the
halfdata type.
2.1.1. Macros
-
CUDART_INF_FP16 __ushort_as_half((unsigned short)0x7C00U)
-
Defines floating-point positive infinity value for the
halfdata type.
-
CUDART_MAX_NORMAL_FP16 __ushort_as_half((unsigned short)0x7BFFU)
-
Defines a maximum representable value for the
halfdata type.
-
CUDART_MIN_DENORM_FP16 __ushort_as_half((unsigned short)0x0001U)
-
Defines a minimum representable (denormalized) value for the
halfdata type.
-
CUDART_NAN_FP16 __ushort_as_half((unsigned short)0x7FFFU)
-
Defines canonical NaN value for the
halfdata type.
-
CUDART_NEG_ZERO_FP16 __ushort_as_half((unsigned short)0x8000U)
-
Defines a negative zero value for the
halfdata type.
-
CUDART_ONE_FP16 __ushort_as_half((unsigned short)0x3C00U)
-
Defines a value of 1.0 for the
halfdata type.
-
CUDART_ZERO_FP16 __ushort_as_half((unsigned short)0x0000U)
-
Defines a positive zero value for the
halfdata type.