Talk Abstract
As the computing wo

rld %93goes multicore %94 high performance library development finally beco

mes a nightmare. Optimal programs and their underlying algorithms have to
be adapted to take full advantage of the platform%92s parallelism memory

hierarchy and available instruction set. To make things worse the best im

plementations are often platform-dependent and platforms are constantly evo

lving which quickly renders libraries obsolete. As a consequence develope

rs are forced to permanently re-implement and re-optimize the same function

ality and often even revert to assembly coding just as 50 years ago.

A number of research efforts have started to address this problem in a new
area called Automatic Performance Tuning with the common goal to rethink t

he way libraries are created. In this talk we present Spiral (www.spiral.ne

t) a program generation system for linear transforms. Spiral generates hig

hly optimized platform-tuned implementations of transforms directly from a
problem specification. For a user-specified transform Spiral generates al

ternative algorithms optimizes them compiles them into programs and %93i

ntelligently%94 searches for the best match to the computing platform. The

main idea behind Spiral is a mathematical declarative framework to represe

nt algorithms and the use of rewriting systems to generate and optimize alg

orithms at a high level of abstraction. Optimization includes parallelizati

on for vector architectures shared and distributed memory platforms GPUs
and even FPGAs. Experimental results show that the code generated by Spira

l competes with and sometimes outperforms the best available human-writte

n library code. Further recent research shows that it may be possible to e

xtend Spiral into other domains such as coding or linear algebra. As for th

e question in the title: Spiral shows that at least for well-understood pr

oblem domains a positive answer may be in reach.

Speaker Bio
Mar

kus P%FCschel is an Associate Research Professor of Electrical and Computer
Engineering at Carnegie Mellon University. He received his Diploma (M.Sc.)
in Mathematics and his Doctorate (Ph.D.) in Computer Science in 1995 and

1998 respectively both from the University of Karlsruhe Germany. From 19

98-1999 he was a Postdoctoral Researcher at Mathematics and Computer Scienc

e Drexel University. Since 2000 he has been with Carnegie Mellon Universit

y. He is an Associate Editor for the IEEE Transactions on Signal Processing
and was a Guest Editor of the Journal of Symbolic Computation the Procee

dings of the IEEE and an Associate Editor for the IEEE Signal Processing L

etters. He is a recipient of the Outstanding Research Award of the College

of Engineering at Carnegie Mellon. His research interests include computing
algorithms applied mathematics and signal processing theory/software/ha