A new project has started concerning hardware acceleration of calculations, including means of modern graphics cards. The first card, I tested, was ATI Radeon HD4850. Its processor RV770 has 10 SIMD engines built-in, each engine has got 16 thread processors. Taking into account the fact that each thread processor has got 5 stream cores (i.e. shaders), one RV770 unit has got 800 cores! So, a parallelization of computation can be easily applied using the stream cores. It is so called Stream Computing. All I need is Microsoft Visual Express Edition 2005 + platform SDK and compiler with Brook+ libraries - available online on Microsoft and ATI/AMD web pages for free.
The debug version of the project: "Slope stability by variational method with PGA application" is just available:
The next stage - NVidia chipsets: GeForce GTX285, which is equipped with 240 fast cores. The cards with CUDA only (i.e. Compute Unified Device Architecture) can be programmed.
Finished (but still as a debug version) CUDA project: Slope stability by variational method with PGA application" is just available:
The works with CUDA are continued...