Advanced GPU Programming Using Mathematica and CUDA
Presented by Wolfram Research

Join us in 2011 on May 4 at noon, May 11 at 4pm, or May 13 at 1pm for our "S71: Advanced GPU Programming Using Mathematica and CUDA" webinar. All times are U.S. Eastern Daylight Time (EDT).

The seminar is a continuation of "S70: GPU Computation Using Mathematica and CUDA", taking an in-depth look at the new GPU programming functionality in Mathematica 8 through CUDA.


Topics:

Overview of new GPU programming functionality in Mathematica
Using CUDAQ and CUDAInformation to find information about GPUs installed on the system; using built-in CUDA accelerated functions to speed up common image or array operations

Compiling CUDA programs
Using the NVCCCompiler and CCompilerDriver to compile CUDA programs into binary, library, or object files

Loading CUDA programs into Mathematica
Using CUDAFunctionLoad to load CUDA programs from source or binary into Mathematica

Using CUDAMemory to optimize memory usage
Using CUDAMemoryLoad and CUDAMemoryAllocate to minimize GPU memory usage; using CUDAMemoryUnload to delete memory

CUDA programming workflow
Using Mathematica parallel tools and symbolic code generation to speed up and simplify computation by using CUDALink