Graduate Exam Abstract

Tejasi Pimpalkhute

M.S. Final
October 15, 2013, 3:30 PM
Application and Memory-Aware NoC Optimizations in Multi-core Computing

Abstract: In chip multiprocessor (CMP) systems, communication and memory access both play an important role in influencing the performance achievable by the system. For any system executing multi-programmed workloads, it is essential for memory requests from each core to be handled efficiently inorder to ensure optimal system level performance. In a multi-core environment with several applications executing in parallel, system performance is significantly impacted by network and memory performance. The manner in which network packets and off-chip memory bound packets are handled determines end-to-end latencies across the network and memory. Several techniques have been proposed in the past that schedule packets in an application-aware manner or memory requests in a DRAM row/bank locality aware manner. Prioritization of memory requests is a major factor in increasing the overall system throughput. Moreover, with the increasing diversity in multi-programmed systems, applying the same prioritization rules to all packets traversing the NoC as is done in current implementations may no longer be a viable approach. In this thesis, we propose a holistic framework that integrates novel scheduling techniques for both network and memory accesses and operates cohesively in an application-aware and memory-aware manner to optimize overall system performance. Our application-aware technique makes fine grain classification of applications. A new ranking scheme for classifying an application’s criticality is also proposed. We propose two memory scheduling algorithms, out of which one is specifically tuned for high-speed memories. We analyze the fairness issues which crop up in a multi-programmed environment and to ensure fairness in the system and propose a fairness strategy to be employed system-wide. We evaluate our framework using a detailed cycle-accurate full system event-driven simulator and achieve significant improvement over the previous work.

Adviser: Dr. Sudeep Pasricha
Co-Adviser: N/A
Non-ECE Member: Dr. Wim Bohm
Member 3: Dr. Anura Jayasumana
Addional Members: N/A


Program of Study:
ECE 452
GRAD 511
ECE 561
ECE 554
CS 575