CUDA Toolkit 2.2 (May 2009)

Release Highlights

• CUDA Visual Profiler now reports memory transactions
• Improved OpenGL interop performance
• Support for texturing from pitch linear memory delivers up to 2x bandwidth savings for video processing and other applications
• New Zero-copy feature allows kernel functions to read and write directly from pinned system memory
• Support for pinned shared memory applications to use less total system memory and achieve better performance when using multiple GPUs
• Asynchronous memcopy on Vista allows applications to realize significant performance improvements (previously supported on other OS's)
• CUDA GDB hardware debugger helps debug 1000s of threads running on the GPU, complete with breakpoints, watch variables, and state inspection

New CUDA SDK code samples:

• Added Mandelbrot, deviceQueryDrv, radixSort, SobolQRNG, and threadFenceReduction samples
• Updated code samples so demonstrate use of CUDA Event Blocking Synchronization (CU_CTX_BLOCKING_SYNC)

Windows XP/Vista

Linux

MacOS