Title: Code Generation for Extreme-Scale Parallel Systems

Description: Speaker: Dr. Karthik Murthy

Time: Thursday, 17 January 2019, 11:30am
Venue: Department of Computer Science and Engineering, Room No. 109, 01st Floor, New CSE/CC Building

Abstract:
Power consumption and fabrication limitations are increasingly playing significant roles in the design of extreme-scale parallel systems. These factors are influencing system designers to support higher on-node computing capability via throughput-optimized processors instead of latency-optimized processors. However, the inter- and intra-processor communication capabilities on such systems are not increasing at the same rate as the on-node computing capability. Consequently, achieving high performance requires careful orchestration of both single- and multi-processor parallelism. Additionally, resilience from failures on these extreme-scale systems is critical to executing applications successfully.

In this talk, I will present compiler and runtime technology that can help applications achieve high-performance on extreme-scale parallel systems. Specifically, I will focus on 1) compiler transformations that enable efficient vector parallelism in the presence of common kinds of complex dependencies [single-processor parallelism], 2) compiler technology to support sophisticated algorithms that minimize interprocessor communication and a novel transformation, EPAC, to overlap communication with computation for systolic computations [multi-processor parallelism], and 3) runtime technology to deploy reboot-able dynamic task graphs [resilience from failures].

The milieu for showcasing the compiler technology will be CA-HPF, a version of the Rice dHPF compiler infrastructure that supports the High Performance Fortran (HPF) programming model, and for the runtime technology will be Legion, a task-based parallel programming model.

Speaker Profile:
Karthik Murthy is a postdoc with Alex Aiken in the CS department at Stanford. He spent several wonderful years at Rice University and University of Texas at Austin working on code generation for distributed memory models, specifically on the compiler side. He is now at Stanford crossing the software bridge to work on the runtime-side of parallel languages.

Organization:
Stanford University



Host:
Prof. Uday Khedker

Date: Thursday, 17 January, 2019
Time: 11:30am IST
Access: Public
Category: Talk*
Created by: Department Calendar
Updated: Saturday, 1 February, 2020 9:51am IST
Send Reminder: Yes  -  308 hours 19 minutes before start
Participants: Department Calendar
<office@cse.iitb.ac.in> (External User)
_NUC_department <all@cse.iitb.ac.in> (External User)