HIP: Heterogenous-computing Interface for Portability
Loading...
Searching...
No Matches
HIP_INTRINSIC_BFLOAT16_STRUCT

Struct to represent a 16 bit brain floating point number. More...

Typedefs

typedef _Float16 __2f16 __attribute__((ext_vector_type(2)))
 

Detailed Description

Struct to represent a 16 bit brain floating point number.

Typedef Documentation

◆ __attribute__

__device__ __attribute__

raw representation of bfloat16

create __hip_bfloat16 from an unsigned int

create __hip_bfloat16 from a int

create __hip_bfloat16 from an unsigned short

create __hip_bfloat16 from a short

create __hip_bfloat16 from a double

create __hip_bfloat16 from a float

create __hip_bfloat16 from a __hip_bfloat16_raw

default constructor

return a __hip_bfloat16_raw

return a __hip_bfloat16_raw cv qualifier

return false if bfloat value is +0.0 or -0.0, returns true otherwise

return a casted char from underlying float val

return a float

return a casted int casted from float of underlying bfloat16 value

return a casted long casted from float of underlying bfloat16 value

return a casted long long casted from float of underlying bfloat16 value

return a casted short casted from float of underlying bfloat16 value

return a casted signed char from float of underlying bfloat16 value

return a casted unsigned char casted from float of underlying bfloat16 value

return a casted unsigned int casted from float of underlying bfloat16 value

return a casted unsigned from float of underlying bfloat16 value

return a casted unsigned long long from float of underlying bfloat16 value

return a casted unsigned short from float of underlying bfloat16 value

assign value from an unsigned int

assign value from a int

assign value from an unsigned short

assign value from a short int

assign value from a double

assign value from a float

assign value from a __hip_bfloat16_raw

assign value from a __hip_bfloat16_raw volatile

assign value from a __hip_bfloat16_raw cv qualifier