gentype
half_divide
(
| gentype x, |
gentype y) |
gentype
native_divide
(
| gentype x, |
gentype y) |
half_divide
computes x
/
y
.
native_divide
computes x
/
y
over an implementation-defined range.
The maximum error is implementation-defined.
The vector versions of the math functions operate component-wise. The description is per-component.
The built-in math functions are not affected by the prevailing rounding mode in the calling environment, and always return the same value as they would if called with the round to nearest even rounding mode.
For any specific use of a function, the actual type has to be the same for all arguments and the return type, unless otherwise specified.
A subset of Math functions (table 6.8) are defined
with the half_
prefix. These
functions are implemented with a minimum of
10-bits of accuracy i.e. an ULP value ≤
8192 ulp.
A subset of Math functions (table 6.8) are
defined with the native_
prefix. These
functions may map to one or more native device
instructions and will typically have
better performance compared to the corresponding
functions (without the native__
prefix). The accuracy (and in some cases the
input range(s)) of
these functions is implementation-defined.
For the half and native forms, we use the generic type name gentype to indicate that the functions (table 6.9) can take float, float2, float3, float4, float8 or float16 as the type for the arguments.
Support for denormal values is optional for
half_
functions. The half_
functions may return any
result allowed by section 7.5.3, even when
-cl-denorms-are-zero
(see section 5.8.4.2) is not in
force. Support for denormal values is
implementation-defined for native_
functions.