Geo-electrical tomographical software plays a crucial role in geophysical research. However, imported software is expensive and does not provide much customizability, which is essential for more advanced geophysical study. Besides, these programs are unable to exploit the full potential of modern hardware, so the running time is inadequate for large-scale geophysical surveys. It is therefore an essential task to develop domestic software for overcoming all these problems. The development of this software is based on our research in using parallel programming on modern multi-core processors and stream processors for high performance computing. | For a problem of certain size, the number of threads per block and the number of blocks and the grid and block dimension can be manually controlled by the programmer at kernel launch. Choosing these numbers appropriately can have great effects on execution speed. As threads are scheduled in warps of 32 threads each, the number of threads per block should be a multiple of 32. Each generation of processor also has maximum limitations on number of threads, number of threads per stream processor and number of blocks per stream processor. Careful considerations of these parameters can enable the kernel to utilize the full potential of the GPU. For example the G80 architecture can schedule up to 8 blocks and 768 threads per stream processor. If there are 64 threads per block, there are 12 blocks but only 8 blocks can go into the stream processor so only 512 threads is scheduled. With 256 threads per block, each stream processor can execute three blocks and achieve full capacity. For 1024 threads per block, even one block can not fit into a stream processor and cannot be scheduled. Choosing the right parameter for kernel launch is a difficult problem which has different solutions for each situation and optimal execution can only be obtained with some amount of experiments.
