Module openj9.cuda

Package com.ibm.cuda


package com.ibm.cuda
This package enables access to CUDA-capable devices from Java.

A wealth of background information on CUDA is available here.

  • Class
    Description
    The Cuda class provides general CUDA utilities.
    The CudaBuffer class represents a region of memory on a specific device.
    The CudaDevice class represents a CUDA-capable device.
    CacheConfig identifies the cache configuration choices for a device.
    Limit identifies device limits that may be queried or configured.
     
    The CudaError interface defines the possible values of CudaException.code.
    The CudaEvent class represents an event that can be queued in a stream on a CUDA-capable device.
    The CudaException class represents an unusual or unexpected response from a CUDA-capable device.
    The CudaFunction class represents a kernel entry point found in a specific CudaModule loaded on a CUDA-capable device.
    The CudaGlobal class represents a global variable found in a specific CudaModule loaded on a CUDA-capable device.
    The CudaGrid class represents a kernel launch configuration.
    CudaJitInputType identifies the type of input being provided to CudaLinker#add(...).
    The CudaJitOptions class represents a set of options that influence the behavior of linking and loading modules.
    CacheMode identifies the cache management choices.
    Fallback identifies the available fall-back strategies when an exactly matching object is not available.
    CudaJitTarget identifies the target compute device class for linking and loading modules.
    The CudaKernel class represents a kernel function in a loaded CudaModule.
    The Parameters class represents the actual parameters in a kernel launch.
    The CudaLinker class supports combining one or more code fragments to form a module that can be then loaded on a CUDA-capable device.
    The CudaModule class represents a module that has been loaded on a CUDA-capable device.
    The Cache class provides a simple mechanism to avoid reloading modules repeatedly.
    This class defines CUDA permissions as described in the following table.
    The CudaStream class represents an independent queue of work for a specific CudaDevice.
    The CudaSurface class represents a surface object found in a specific CudaModule loaded on a CUDA-capable device.
    The CudaTexture class represents a texture object found in a specific CudaModule loaded on a CUDA-capable device.
    The Dim3 class represents the dimensions of a cube.