Debug
Namespace: SymTensor.Compiler.Cuda
Functions and values
| Function or value | Description |
DebugCompile
Signature: bool
|
compiles kernels with debug information and no optimizations |
DisableElementsWorkOrdering
Signature: bool
|
disables ordering of element expressions work to increase coalesced memory access |
DisableEvents
Signature: bool
|
disables all events |
DisableKernelCache
Signature: bool
|
disables the caching of CUDA kernels |
DisableStreams
Signature: bool
|
redirects all stream calls to the null stream |
DumpCode
Signature: bool
|
dumps kernel code before it is compiled |
FastKernelMath
Signature: bool
|
enable the use of fast math operations in kernels |
GenerateLineInfo
Signature: bool
|
compiles kernels with source line-number information (for nVidia profiler) |
KeepCompileDir
Signature: bool
|
keeps the compile temporary directory |
PtxasInfo
Signature: bool
|
prints ptxas verbose information during compilation |
ResourceUsage
Signature: bool
|
prints CUDA memory usage during CUDA function compilation |
RestrictKernels
Signature: bool
|
tells nvrtc that all pointers are restricted |
SyncAfterEachCudaCall
Signature: bool
|
synchronizes the CUDA context after each call to detect errors |
TerminateWhenNonFinite
Signature: bool
|
terminates the program when a non-finite tensor was found by the CheckFinite op |
Timing
Signature: bool
|
prints timing information during CUDA function compilation |
TraceCalls
Signature: bool
|
outputs messages when a function / kernel is launched |
TraceCompile
Signature: bool
|
prints each compile step during compilation |
