Tutorial by Examples

Consider a grid working on some task, e.g. a parallel reduction. Initially, each block can do its work independently, producing some partial result. At the end however, the partial results need to be combined and merged together. A typical example is a reduction algorithm on a big data. A typica...

Consider an array of work items. The time needed for an each work item to complete varies greatly. In order to balance the work distribution between blocks it may be prudent for each block to fetch the next item only when previous one is complete. This is in contrast to a-priori assigning items to...