Discovering and exploiting program phases

Abstract

Understanding program behavior is at the foundation of computer architecture and program optimization. Many programs have wildly different behavior on even the largest of scales (that is, over the program's complete execution). During one part of the execution, a program can be completely memory bound; in another, it can repeatedly stall on branch mispredicts. Average statistics gathered about a program might not accurately picture where the real problems lie. This realization has ramifications for many architecture and compiler techniques, from how to best schedule threads on a multithreaded machine, to feedback-directed optimizations, power management, and the simulation and test of architectures. Taking advantage of time-varying behavior requires a set of automated analytic tools and hardware techniques that can discover similarities and changes in program behavior on the largest of time scales.The challenge in building such tools is that during a program's lifetime it can execute billions or trillions of instructions. How can high-level behavior be extracted from this sea of instructions?The reality is this: The way a program's execution changes over time is not totally random; in fact, it often falls into repeating behaviors, called phases. Automatically identifying this phase behavior is the goal of our research and key to unlocking many new optimizations. We define a phase as a set of intervals (or slices in time) within a program's execution that have similar behavior, regardless of temporal adjacency. Recent research has shown that it is indeed possible to accurately identify and predict these phases in program behavior to capture meaningful phase behavior. 1-8The key observation for phase recognition is that any program metric is a direct function of the way a program traverses the code during execution. We can find this phase behavior and classify it by examining only the ratios in which different regions of code are being executed over time. We can simply and quickly collect this information using basic block vector profiles for off-line classification 4,6 or through dynamic branch profiling for online classification. 7 In addition, accurately capturing phase behavior through the computation of a single metric, independent of the underlying architectural details, means that it is pos-

SACR: Scheduling-Aware Cache Reconfiguration for Real-Time Embedded Systems

“…However, it is inappropriate for real-time systems as it either imposes unpredictable performance or significant energy overhead, both due to the exploration of suboptimal cache configurations. During static analysis, various cache alternatives are explored and the best cache configuration is selected for each application in its entirety (application-based tuning) [7] or for each phase of execution within an application (phase-based tuning) [17]. Regardless of the tuning method, the predetermined best cache configuration (based on design requirements) may be stored in a look-up table or encoded into specialized instructions.…”

Explaining Dynamic Cache Partitioning Speed Ups

“…We collected traces of the most representative 300 million instruction segment of each program, following the SimPoint methodology [7]. We use the FAME simulation methodology proposed in [10] with a Maximum Allowable IPC Variance of 5%.…”

An Inductive Method to Select Simulation Points

“…Selection of simulation points is usually performed based on the same idea of phase detection [3], [4]. SimPoint [5]- [8], which is the most recognized method, selects simulation points from the program counter (PC) sequence of executed instructions obtained by emulating the target processor.…”