Perceptron Based Consumer Prediction in Shared-Memory Multiprocessors

Abstract

Abstract-Recent research has shown that forwarding speculative data to other processors before it is requested can improve the performance of multiprocessor systems. The most recent work in speculative data forwarding places all of the processors on a single bus, allowing the data to be forwarded to all of the processors at the same cost as any subset of the processors. Modern multiprocessors however often employ more complex switching networks in which broadcast is expensive. Accurately predicting the consumers of data can be challenging, especially in the case of programs with many shared data structures.Past consumer predictors rely on simple prediction mechanisms, a single table lookup followed by a static mapping of the table values onto a prediction. We make two main contributions in this paper. First, we show how to reduce the design space of consumer predictors to a set of interesting predictors, and how previous consumer predictors can be tuned to expand the range of available performance. Second, we propose a perceptron consumer predictor that dynamically adapts its reaction to the system behavior, and uses more history information than previous consumer predictors. This predictor outperforms the previous predictors by 21% while using only 1KByte more storage than previous predictors.

Invalidate or Update? Revisiting Coherence for Tomorrow's Cache Hierarchies

“…Instruction-based predictors [52], [53] are proposed as an alternative to normal address-based predictors. There are also some coherence predictors based on perceptron [54]. All of these optimizations are costly and may require some major changes to a processor design.…”

Learning Forward Reuse Distance

“…The perceptron is almost the simplest possible neural networks, which enables hardware implementation. Because of its simplicity, the perceptron technique is used later for several other systems problems [22], [42]. LSTM is a recurrent neural networks that is getting popular in the application in systems problems.…”

Tree‐based scheme for reducing shared cache miss rate leveraging regional, statistical and temporal similarities

“…In contrast, there are several methods in the literature to address the miss rate concern for off‐chip multi‐processor systems [3, 4, 11–13]. Recent off‐chip researches reduce miss in sets of cache blocks that are accessed in consistent streams [4, 13].…”

“…Some of the other existing off‐chip methods moderate the number of remote misses by updating invalidated blocks. To avoid generating excessive update messages to nodes that no longer wish to consume the data, the consumer set has to be predicted [11, 12].…”