Offline Phase Analysis and Optimization for Multi-configuration Processors

Vandeputte, Frederik; Eeckhout, Lieven; Bosschere, Koen De

doi:10.1007/11512622_22

Cited by 4 publications

(7 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [8], Vandeputte et al proposed an offline phase classification method based on [7], that is able to efficiently deal with multi-configuration processors. In [10], Vandeputte et al analyzed and compared a number of state-of-the-art phase predictors.…”

Section: Previous Workmentioning

confidence: 99%

“…Being aware of this large-scale time-varying behavior is key to understanding the behavior of a program as a whole. Phase behavior can be exploited for various purposes, like performance modeling [1], compiler optimizations [1], hardware adaptation [2,3,4,5,6,1,7,8], etc. For example in phase-based hardware adaptation, if we know that particular parts of the processor are nearly unused during some program phase, we could turn off those parts during that phase, resulting in a reduced energy consumption without affecting overall performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Analysis of Program Phase Behavior and its Predictability

Vandeputte¹,

Eeckhout²,

Bosschere³

2006

AIP Conference Proceedings

Self Cite

View full text Add to dashboard Cite

The execution of a computer program typically consists of a number of phases where each phase exhibits its own behavior. This phase behavior can be exploited for various purposes such as hardware adaptation, performance modeling, etc. Recently, a number of phase identification and classification schemes have been proposed to capture this phase behavior. For some of these phase classification schemes, phase predictors are needed to accurately predict phase transitions.In this paper, we analyze program phase behavior by constructing a model for representing the phase sequence of a program execution. We validate this model by generating a probabilistic phase sequence from it and by comparing the phase prediction accuracy of this phase sequence with the original program phase sequence for a number of phase predictors. From this model, we derive a compact model which we use to evaluate the performance of various types of phase predictors. This gives us a better understanding in program phase behavior.

show abstract

Section: Previous Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

An Analysis of Program Phase Behavior and its Predictability

Vandeputte¹,

Eeckhout²,

Bosschere³

2006

AIP Conference Proceedings

Self Cite

View full text Add to dashboard Cite

show abstract

“…In [7], Vandeputte et al propose an offline phase analysis methode that is capable of efficiently dealing with multi-configuration hardware where a large number of hardware units can be configured adaptively. This offline phase analysis technique determines the phases based on a fused metric that incorporates both phase predictability and phase homogeneity.…”

Section: Previous Workmentioning

confidence: 99%

“…Note that the number of unique phase IDs is fairly small here compared to [1] [9]. The reason is that our offline phase analysis approach [7] balances phase predictability and phase homogeneity, whereas in [1] [9], the main objective is phase homogeneity.…”

Section: Previous Workmentioning

confidence: 99%

“…Being aware of this large-scale timevarying behavior is key to understanding the behavior of the program as a whole. Phase behavior can be exploited for various purposes, ranging from performance modeling [1], compiler optimizations [2], hardware adaptation [3][4] [5][6] [7], etc. For example in phase-based hardware adaptation, if we know that particular parts of the processor are unused during some program phase, we can turn off those parts during that phase resulting in a reduced energy consumption without affecting overall performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Detailed Study on Phase Predictors

Vandeputte

Eeckhout

Bosschere

2005

Euro-Par 2005 Parallel Processing

Self Cite

View full text Add to dashboard Cite

Abstract. Most programs are repetitive, meaning that some parts of a program are executed more than once. As a result, a number of phases can be extracted in which each phase exhibits similar behavior. These phases can then be exploited for various purposes such as hardware adaptation for energy efficiency. Temporal phase classification schemes divide the execution of a program into consecutive (fixed-length) intervals. Intervals showing similar behavior are grouped into a phase. When a temporal scheme is used in an on-line system, phase predictors are necessary to predict when the next phase transition will occur and what the next phase will be. In this paper, we analyze and compare a number of existing state-of-the-art phase predictors using the SPEC CPU2000 benchmarks. The design space we explore is huge. We conclude that the 2-level burst predictor with confidence and conditional update is today's most accurate phase predictor within reasonable hardware budgets.

show abstract