The single-precision floating-point vector value is converted to a half vector value using the appropriate rounding mode. The halfn
value is then written to address computed as (p
+ (offset
* n
)). The write address computed as (p
+ (offset
* n
)) must be aligned to sizeof (half) * number of half elements
being read (given by n
) bytes.
vstorea_half
uses the current rounding mode. The default current rounding mode is round to nearest even.
n
If extended with cl_khr_fp64, the function is extended to allow a double precision scalar or vector value to be written to memory as half values. In this case, the functions use double instead of float.