News for hackers. Stuff that matters.
Re:Show HN: A high-performance framework for high-level para... by rcorcs on Friday August 14, 2015 @05:11PM (#10062970)
For the stencil computation, if the input and output array together are greater than the GPU memory, there is an automatic and transparent method for tiling the data array into the GPU's memory, computing the stencil pattern in several chunks.

There are no comments on this comment yet.