8715 articles  [version française]

hal-00758713, version 1

Revisiting Value Prediction

Arthur Perais () a1, André Seznec () a1

N° RR-8155 (2012)

Abstract: Value prediction was proposed in the mid 90's to enhance the performance of high-end microprocessors. Unfortunately, to the best of our knowledge, there are no Value Prediction implementations available on the market. Moreover, the research on Value Prediction techniques almost vanished in the early 2000's as it was more effective to increase the number of cores than to dedicate silicon to Value Prediction. However, high-end processor chips currently feature 8-16 high-end cores and the technology will allow to implement 50-100 of such cores on a single die in a foreseeable future. Amdahl's law suggests that the performance of most workloads will not scale to that level. Therefore, dedicating more silicon area to single high-end core will be considered as worthwhile for future multicores, either in the context of heterogeneous multicores or homogeneous multicore. In particular, spending transistors on specialized, performance and/or power optimized units, such as a value predictor. In this report, we first build on the concept of value prediction. We introduce a new value predictor VTAGE harnessing the global branch history. VTAGE directly inherits the structure of the indirect jump predictor ITTAGE. We show that VTAGE is able to predict with a very high accuracy many values that were not correctly predicted by previously proposed predictors, such as the FCM predictor and the stride predictor. Compared with these previously proposed solutions, VTAGE can accommodate very long prediction latencies. The introduction of VTAGE opens the path to the design of new hybrid predictors. Three sources of information can be harnessed by these predictors: the global branch history, the differences of successive values and the local history of values. We show that the predictor components using these %three sources of information are all amenable to very high accuracy at the cost of some prediction coverage. %On SPEC 2006 Using SPEC 2006 benchmarks, our study shows that with a large hybrid predictor, in average 56.76% of the values can be predicted with a 99.48% accuracy against respectively 55.50% and 98.62% without advanced confidence estimation and the VTAGE component.

  • a –  INRIA
  • 1:  ALF (INRIA - IRISA)
  • INRIA – Université de Rennes 1
  • Domain : Computer Science/Hardware Architecture
  • Keywords : Value Prediction – VTAGE – hybrid predictors
  • Internal note : RR-8155
 
  • hal-00758713, version 1
  • oai:hal.inria.fr:hal-00758713
  • From: 
  • Submitted on: Thursday, 29 November 2012 11:03:35
  • Updated on: Thursday, 29 November 2012 14:17:46