PPT-SASMM: Scalable Analytical Shared Memory Model: Predicting the Performance of Multicore Caches from a Single-Threaded Execution Trace

TL;DR

This paper introduces PPT-SASMM, a scalable analytical model that predicts multicore cache performance from single-threaded traces using probabilistic methods, aiding in computational co-design.

Contribution

It presents a novel, efficient probabilistic model for predicting cache reuse profiles in multicore systems from static single-threaded traces, improving performance estimation accuracy.

Findings

Predicts private L1 cache hit rates with 2.12% error

Predicts shared L2 cache hit rates with 1.50% error

Uses static analysis of single-threaded traces for multicore performance modeling

Abstract

Performance modeling of parallel applications on multicore processors remains a challenge in computational co-design due to multicore processors' complex design. Multicores include complex private and shared memory hierarchies. We present a Scalable Analytical Shared Memory Model (SASMM). SASMM can predict the performance of parallel applications running on a multicore. SASMM uses a probabilistic and computationally-efficient method to predict the reuse distance profiles of caches in multicores. SASMM relies on a stochastic, static basic block-level analysis of reuse profiles. The profiles are calculated from the memory traces of applications that run sequentially rather than using multi-threaded traces. The experiments show that our model can predict private L1 cache hit rates with 2.12% and shared L2 cache hit rates with about 1.50% error rate.