🚀 go-pugleaf

Special LCS Seminar
Monday, December 6
Noon - 1 pm
HH 1112
contact:  Karen Lindenfelser (karen@ece) 8-7293

Title:

Prediction Directed Instruction and Data Prefetching

Abstract:

Instruction and data prefetching are effective approaches for reducing
instruction and load latencies.  In this talk I will present new
prediction-based approaches for guiding both instruction and data
prefetching. I will first describe our Fetch-Directed Instruction
Prefetching architecture, which uses a decoupled branch predictor and
instruction cache to guide prefetching.  The branch predictor runs ahead
of the instruction cache producing fetch addresses, which are later used
to perform the normal instruction cache lookup.  Since we have the fetch
addresses early, we use them and filter them to accurately prefetch
instructions from the L2 cache down the branch predicted address stream.
In the last part of the talk I will describe our new Prediction-Directed
Streaming Buffer architecture for data prefetching.  These new streaming
buffers run ahead of the fetch stream for a given load by following a
pc-based address prediction stream, instead of a fixed stride, as in
prior research.  In addition, I will present a new type of address
predictor called Differential Markov, which provides accurate
stride-based, pointer-based, and correlated address prediction, using
less state than previous predictors.

Short Bio:

Brad Calder is currently an Assistant Professor of Computer Science and
Engineering at the University of California, San Diego.  Before joining
UCSD in January of 1997, he worked as a Principal Engineer at Digital
Equipment Corporation's Western Research Lab in Palo Alto. His research
focuses on the interaction between computer architecture and compilers.
His current work focuses on optimizations for reducing memory latency,
fetch prediction architectures, adaptive power optimizations, value
prediction, predicated execution, and multiple path execution. Brad
Calder received his Ph.D. in Computer Science from the University of
Colorado, Boulder in 1995, and an M.S. from Colorado in 1993.  He
obtained a B.S. in Computer Science and a B.S. in Mathematics from the
University of Washington in 1991.  He is a recipient of an NSF CAREER
Award.