from Wikipedia, the free encyclopedia

As prefetching ( German  storage anticipation ) is known in the computer science, the heuristic load memory contents up in the memory hierarchy before a need evident has become to achieve so in the case of Actual demand a higher access speed.

In the case of single instruction, multiple data (SIMD) , future demand can easily be predicted with sufficient accuracy, so that prefetching can be initiated with special instructions. Prefetching is also used in pipelining .

The disadvantage is the resulting inefficiency due to actually unnecessary communication and due to the actually unnecessary use of faster storage media, which can have a paradoxical effect (ie: the access times are still increasing). The advantages and disadvantages achieved depend heavily on the heuristic chosen.

The increasing spread of smartphones results in additional problems when optimizing the prefetch, since a change in the cellular network , e.g. B. between UMTS and LTE influences the data rate and the prefetching has to be dynamically adapted accordingly.

See also

Individual evidence

  1. Lee, J., Kim, H., and Vuduc, R .: When Prefetching Works, When It Doesn't, and Why (PDF) Georgia Institute of Technology . P. 2: 4. 2012. Retrieved May 14, 2016.