NVIDIA CUDA 2.2 available

[Stefan]

NVIDIA CUDA
Windows XP and Vista Release Notes
Version 2.2
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------


--------------------------------------------------------------------------------
New Features
--------------------------------------------------------------------------------

  Hardware Support
  o  See http://www.nvidia.com/object/cuda_learn_products.html

  Platform Support
  o  Additional OS support
     - Microsoft Windows Vista Server 2008

  API Features
  o Pinned Memory Support
     - These new memory management functions (cuMemHostAlloc() and
       cudaHostAlloc()) enable pinned memory to be made "portable" (available
       to all CUDA contexts), "mapped" (mapped into the CUDA address space),
       and/or "write combined" (not cached and faster for the GPU to access).
     - cuMemHostAlloc
     - cuMemHostGetDevicePointer
     - cudaHostAlloc
     - cudaHostGetDevicePointer
  o Function attribute query
     - This function allows applications to query various function properties.
     - cuFuncGetAttribute
  o 2D Texture reads from pitch linear memory
     - You can bind linear memory that you get from cuMemAlloc() or
       cudaMalloc() directly to a 2D texture. In previous releases, you were
       only able to bind cuArrayCreate() or cudaMallocArray() arrays to 2D
       textures.
     - cuTexRefSetAddress2D
     - cudaBindTexture2D
  o Flags for event creation
     - Applications can now create events that use blocking synchronization.
     - cudaEventCreateWithFlags
  o New device management and context creation flags
     - The function cudaSetDeviceFlags() allows the application to specify
       attributes such as mapping host memory and support for blocking
       synchronization.
     - cudaSetDeviceFlags
  o Improved runtime device management
     - The runtime now defaults to attempting context creation on other
       devices in the system before returning any failure messages. The new
       call cudaSetValidDevices() allows the application to specify a list of
       acceptable devices for use.
     - cudaSetValidDevices
  o Driver/runtime version query functions
     - Applications can now directly query version information about the
       underlying driver/runtime.
     - cuDriverGetVersion
     - cudaDriverGetVersion
     - cudaRuntimeGetVersion
  o New device attribute queries
     - CU_DEVICE_ATTRIBUTE_INTEGRATED
     - CU_DEVICE_ATTRIBUTE_CAN_MAP_HOST_MEMORY
     - CU_DEVICE_ATTRIBUTE_COMPUTE_MODE

  Documentation
  o Doxygen-generated and cross-referenced html, pdf, and Windows help files.
     - Runtime API
     - Driver API

  Performance Enhancements
  o Asynchronous memcpy support for Windows Vista
     - Asynchronous memory copy operations can now overlap GPU execution

Download from CUDA zone.