public class cublasMath extends Object
Modifier and Type | Field and Description |
---|---|
static int |
CUBLAS_DEFAULT_MATH |
static int |
CUBLAS_MATH_DISALLOW_REDUCED_PRECISION_REDUCTION
flag to force any reductions to use the accumulator type and not output type in case of mixed precision routines
with lower size output type
|
static int |
CUBLAS_PEDANTIC_MATH
same as using matching _PEDANTIC compute type when using cublas-T-routine calls or cublasEx() calls with
cudaDataType as compute type
|
static int |
CUBLAS_TENSOR_OP_MATH
Deprecated.
|
static int |
CUBLAS_TF32_TENSOR_OP_MATH
allow accelerating single precision routines using TF32 tensor cores
|
Modifier and Type | Method and Description |
---|---|
static String |
stringFor(int n)
Returns a string representation of the given constant
|
public static final int CUBLAS_DEFAULT_MATH
@Deprecated public static final int CUBLAS_TENSOR_OP_MATH
public static final int CUBLAS_PEDANTIC_MATH
public static final int CUBLAS_TF32_TENSOR_OP_MATH
public static final int CUBLAS_MATH_DISALLOW_REDUCED_PRECISION_REDUCTION
public static String stringFor(int n)
Copyright © 2020. All rights reserved.