We recompute edge values of each block since we don't waste time communicating, but since the block size is so large (256x32), it's a small tradeoff for a huge increase in parallel performance.
This comment was marked helpful 0 times.
bottie
Since there exists overlap between blocks, so we need to recompute the edge values of each block.
We recompute edge values of each block since we don't waste time communicating, but since the block size is so large (256x32), it's a small tradeoff for a huge increase in parallel performance.
This comment was marked helpful 0 times.
Since there exists overlap between blocks, so we need to recompute the edge values of each block.
This comment was marked helpful 0 times.