Class Cfg
Options for configuring operations performed on CUDA tensors.
Inheritance
System.Object
Cfg
Namespace: Tensor.Cuda
Assembly: Tensor.dll
Syntax
type CfgRemarks
All settings are local to the calling thread.
Properties
| Name | Description |
|---|---|
| static property DebugCompile: bool | If set to true, all CUDA kernels are compiled with debug flags. |
| static property DisableKernelCache: bool | If set to true, the CUDA kernel cache is disabled. |
| static property FastKernelMath: bool | If set to true, CUDA uses fast math functions with lower accuracy. |
| static property RestrictKernels: bool | If set to true, all arguments are passed as restriced to CUDA kernels (experimental). |
| static property Stacktrace: bool | If set to true, CUDA operations produce an acurate stack trace when an error is encountered. |
| static property Stream: CUstream | The CUDA stream to execute CUDA operations on. |