Synthetic‐perturbation techniques for screening shared memory programs

Abstract

The synthetic‐perturbation screening (SPS) methodology is based on an empirical approach; SPS introduces artificial perturbations into the MIMD program and captures the effects of such perturbations by using the modern branch of statistics called design of experiments. SPS can provide the basis of a powerful tool for screening MIMD programs for performance bottlenecks. This technique is portable across machines and architectures, and scales extremely well on massively parallel processors. The purpose of this paper is to explain the general approach and to extend it to address specific features that are the main source of poor performance on the shared memory programming model. These include performance degradation due to load imbalance and insufficient parallelism, and overhead introduced by synchronizations and by accessing shared data structures. We illustrate the practicality of SPS by demonstrating its use on two very different case studies: a large image understanding benchmark and a parallel quicksort.

C oz

“…Mytkowicz et al use delays to validate the output of profilers on single-threaded Java programs [35]. Snelick, JáJá et al use delays to profile parallel programs [38]. This approach measures the effect of slowdowns in combination, which requires a complete execution of the program for each of an exponential number of configurations.…”

Evaluation of applications on a loosely-coupled cluster

“…Algorithm NEWS: send to all neighbors /* 8 msgs (sends)*/ for ( each neighbor rcvd a msg ) /* [8][9][10][11][12]…”

“…We used the S-Check Barrier test[12,13] to determine communication delays. This test provides a measure of the time variation between nodes completing their computation phase, and thus delaying the transmission of their data.…”

Using S-Check, alpha release 1.0

“…Note that screening experiments and barrier tests must be conducted sep-Using S-Check S-Check's Directory/Experiment Layout arately as factor treatments in these two cases are not comparable. See the Reference [2], "Synthetic perturbation techniques for screening shared-memory programs", for a thorough explanation of the technique. Scaling test.…”

“…However, a brief superficial overview of the process is given in the next section, SPS Basics. A detailed description of SPS can be found in the following publications: "Synthetic-perturbation tuning of MIMD programs"[l], "Synthetic perturbation techniques for screening shared-memory programs" [2], and "A simple scalability test for parallel code" [3].…”