There is a signup schedule for this event.

Type of Talk:

Colloquia

Speaker Name: Vijay Menon

Speaker Affiliation: In

tel

Date: Thursday November 2 2006

Start Time: 9:30 a.m.<

br>
Location: PAI 3.14

Host: Keshav Pingali

Talk Title:
Efficient Software Transactional Memory

Talk Abstract:
Program

mers have traditionally used locks to synchronize concurrentaccess to share

d data. Lock-based synchronization however has well-known pitfalls inclu

ding the lack of composability and the possibility of deadlock. Transactio

nal memory provides an alternate concurrency control mechanism that avoids

these pitfalls and significantly eases concurrent programming. Transactiona

l memory language constructs have recently been proposed as extensions to e

xisting languages or included in new concurrent language specifications op

ening the door for new optimizations that target the overheads of transacti

onal memory.
In this talk I discuss compiler and runtime optimization

s for transactional memory language constructs and present the high-perform

ance software transactional memory system (STM) being developed at Intel''s
Programming Systems Lab. Our system which is integrated into a managed ru

ntime environment efficiently implements nested transactions that support

both composition of transactions and partial roll back. Our JIT compiler is
the first to optimize the overheads of STM and we show novel techniques f

or enabling JIT optimizations on STM operations. We measure the performance
of our optimizations on a 16-way SMP running multi-threaded transactional

workloads. Our results show that these techniques enable transactional memo

ry''s performance to compete with that of well-tuned synchronization. I wi

ll conclude with a discussion of ongoing work on supporting extensions such
as strong atomicity and open nested transactions in an STM setting.

Speaker Bio:
Vijay Menon is a research staff member at the Intel Sant

a Clara labs. He received his bachelor''s degree from the University of Cal

ifornia Berkeley and his PhD from Cornell where he worked with Keshav Pin

gali on fractal symbolic analysis. He is known for his contributions to sof

tware transactional memory and JIT compiling. He will give two talks this T

hursday on recent work that has appeared in PLDI and POPL.