---

CUDAAccel.C File Reference

VMD CUDA GPU and topology enumeration and management interface. More...

#include <stdio.h>
#include <stdlib.h>
#include "config.h"
#include "Inform.h"
#include "ResizeArray.h"
#include "CUDAAccel.h"
#include "CUDAKernels.h"
#include "WKFThreads.h"
#include "ProfileHooks.h"

Go to the source code of this file.

---

Detailed Description

VMD CUDA GPU and topology enumeration and management interface.

Class to enumerate and initialize CUDA GPU accelerator devices, while providing an abstraction layer that allows a user-defined subset of GPU devices to be enabled, while excluding others based on their hardware capabilities, software configuration properties, or to facilitate conflict free sharing of GPUs among multiple application processes on the same workstation or compute node. In order for the abstraction to work properly, all of the CUDA code in VMD needs to make use of the abstracted query functions rather than the native CUDA functions, so that device masks and selection criteria are always honored, and to avoid needless duplication of GPU device selection logic, and the like.

Definition in file CUDAAccel.C.

---