| VFu_rsqrt | VDu_rsqrt | VEu_rsqrt |
|
| Function | Reciprocal square root |
|
| Syntax C/C++ | #include <VFmath.h>
int VF_rsqrt( fVector Y, fVector X, ui size );
int VFx_rsqrt( fVector Y, fVector X, ui size, float A, float B, float C ); |
| C++ VecObj | #include <OptiVec.h>
int vector<T>::rsqrt( const vector<T>& X );
int vector<T>::x_rsqrt( const vector<T>& X, const T& A, const T& B, const T& C ); |
| Pascal/Delphi | uses VFmath;
function VF_rsqrt( Y, X:fVector; size:UIntSize ): IntBool;
function VFx_rsqrt( Y, X:fVector; size:UIntSize; A, B, C:Single ): IntBool; |
|
| CUDA function C/C++ | #include <cudaVFmath.h>
int cudaVF_rsqrt( fVector d_Y, fVector d_X, ui size );
int cudaVFx_rsqrt( fVector d_Y, fVector d_X, ui size, float A, float B, float C );
int cusdVFx_rsqrt( fVector d_Y, fVector d_X, ui size, float *d_A, float *d_B, float *d_C );
int VFcu_rsqrt( fVector h_Y, fVector h_X, ui size );
int VFxcu_rsqrt( fVector h_Y, fVector h_X, ui size, float A, float B, float C );
|
| CUDA function Pascal/Delphi | uses VFmath;
function cudaVF_rsqrt( d_Y, d_X:fVector; size:UIntSize ): IntBool;
function cudaVFx_rsqrt( d_Y, d_X:fVector; size:UIntSize; A, B, C:Single ): IntBool;
function cusdVFx_rsqrt( d_Y, d_X:fVector; size:UIntSize; d_A, d_B, d_C:PSingle ): IntBool;
function VFcu_rsqrt( h_Y, h_X:fVector; size:UIntSize ): IntBool;
function VFxcu_rsqrt( h_Y, h_X:fVector; size:UIntSize; A, B, C:Single ): IntBool;
|
|
| Description | simple versions: Yi = 1 / sqrt( Xi )
expanded versions: Yi = C / sqrt( A*Xi+B )
The "unprotected" versions (prefix VFu_, VFux_, etc.) do not perform any error handling, which makes them much faster than the standard versions. On the other hand, any negative or near-zero input number may lead to an uncontrolled programme crash. |
|
| Error handling | Normal versions (VF_, VFx_ etc.): DOMAIN errors occur if, in the real-number versions, the square root of a negative numbers is requested; NAN ("not-a-number") is the default result in this case. SING or OVERFLOW errors are treated with the result HUGE_VAL.
Unprotected versions (VFu_, VFux_ etc.): no error handling. |
|
| Return value | FALSE (0), if no error occurred, otherwise TRUE (non-zero) |
|
|