Nakul Jindal

Office E362 CSE Building University of Florida P.O. Box 116120 Gainesville FL 32611, USA

Research

I work on Performance Modeling of Parallel Applications.

ACESIII, developed at the QTP, is a Parallel Computational Chemistry package that consists of a ideas from the Super Instruction Architecture (SIA). The SIA consists of a parallel runtime system - SIP (Super Instruction Processor) and a Domain Specific Language - SIAL (Super Instruction Assembly Language). ACESIII has been run on many large machines - Jaguar (now Titan), Bluewaters, etc. - utilizing thousands of cores. Our goal is to use these resources efficiently.

Performance Modeling allows us to predict wall clock time, perform optimizations by recognizing hot spots, choose various parameters, do load balancing more effectively, etc. We can do all this on expensive parallel machines without making repeated wasted runs and get on with the actual science. 

In the ACESIII package, various Computational Chemistry methods are coded in the Domain Specific Language - SIAL. Modeling the performance of each SIAL code by hand would be tedious. We automate the performance modeling process using our existing compiler infrastructure. We've also built an eclipse plugin using LPG and IMP for the SIAL language.

We are currently working on improving the performance modeling process itself and on its various applications.

Publications

Our paper titled "SIPMaP: A Tool for Modeling Irregular Parallel Computations in the Super Instruction Architecture" has been accepted to IPDPS 2013.

Teaching

• TA for Programming Language Principles (COP555) – Spring ‘12
• TA for Concurrent Programming (COP5618) – Fall ‘12
• Teaching TA for Programming Fundamentals for CIS Majors 2 Using Java (COP3503) – Spring ‘13

Service

Reviewer for IPDPS 2013

Misc.

• LinkedIn Page