|
Doxygen
1.9.1
Toolkit for Adaptive Stochastic Modeling and Non-Intrusive ApproximatioN: Tasmanian v8.2 (development)
|
|
Doxygen
1.9.1
Toolkit for Adaptive Stochastic Modeling and Non-Intrusive ApproximatioN: Tasmanian v8.2 (development)
|
Common methods for manipulating acceleration options and reading CUDA environment properties. More...
Functions | |
| TypeAcceleration | getIOAccelerationString (const char *name) |
| Convert the string (coming from C or Python) into an enumerated type. More... | |
| std::map< std::string, TypeAcceleration > | getStringToAccelerationMap () |
| Creates a map with std::string rule names (used by C/Python/CLI) mapped to TypeAcceleration enums. | |
| const char * | getIOAccelerationString (TypeAcceleration accel) |
| Convert the enumerated type to a string, the inverse of getIOAccelerationString() More... | |
| int | getIOAccelerationInt (TypeAcceleration accel) |
| Convert the integer (coming from Fortran) into an enumerated type. More... | |
| TypeAcceleration | getIOIntAcceleration (int accel) |
| Convert the enumerated type to an integer, the inverse of getIOAccelerationInt() More... | |
| bool | isAccTypeGPU (TypeAcceleration accel) |
| Returns true if accele is cuda, cublas or magma. More... | |
| bool | isAvailable (TypeAcceleration accel) |
| Identifies whether the acceleration mode is available. | |
| TypeAcceleration | getAvailableFallback (TypeAcceleration accel) |
| Implements fallback logic, if accel has been enabled through CMake then this returns accel, otherwise it returns the "next-best-thing". More... | |
| int | getNumGpuDevices () |
| Return the number of visible GPU devices. More... | |
| void | setDefaultGpuDevice (int deviceID) |
| Selects the active device for this CPU thread, not supported for DPC++. More... | |
| unsigned long long | getTotalGPUMemory (int deviceID) |
| Return the memory available in the device (in units of bytes). More... | |
| std::string | getGpuDeviceName (int deviceID) |
| Returns the name of the selected GPU device, empty string if no device is available or the index is out of bounds. More... | |
| template<typename T > | |
| void | recvGpuArray (AccelerationContext const *, size_t num_entries, const T *gpu_data, std::vector< T > &cpu_data) |
| Copy a device array to the main memory, used for testing only, always favor using GpuVector (if possible). More... | |
| template<typename T > | |
| void | delGpuArray (AccelerationContext const *, T *x) |
| Deallocate device array, used primarily for testing, always favor using GpuVector (if possible). More... | |
| void * | createCublasHandle () |
| Creates a new cuBlas handle, used in unit-testing only. | |
| void | deleteCublasHandle (void *) |
| Destroys the cuBlas handle, used in unit-testing only. | |
Common methods for manipulating acceleration options and reading CUDA environment properties.