Accelerated Computing with Chains of Recurrences: light-weight intro and applications

Chains of recurrences technique was introduced as an "aggressive" technique for loop optimization. It was later extended tomultidimensional case that corresponds to the nested loop optimization problem and was used to accelerate massive numeric computations over the regular intervals in such applications as for example mathematical plotting. Recently this technique in very rudimentary form was adopted
by compiler developers. For example it has been adopted in the well-known mainstream GCC 4.0 compiler under the name of "Scalar Evolutions".
In our talk we will briefly introduce this technique and discuss its applications together with several our own (Maple, C, Java, VHDL) implementations of code generators.

 

Category

Lectures, Seminars

Time

Location

CIBC Auditorium, Goldberg Computer Science building

Cost

Free

Additional Information

Dr. Zima received PhD in Computer Science from the Lomonosov Moscow State University (Russia) in 1985 and my MSc in applied mathematics from
the Lomonosov Moscow State University (Russia) in 1980. Prior to joining Laurier, he was an associate professor at the University of Waterloo (1997-2004), assistant and associate professor at Lomonosov Moscow State University (1985-1998), and senior researcher at the Computer Center of the Russian Academy of Sciences (1993-1998).

Dr. Zima's research is a combination of symbolic and numeric computing that includes both theoretical development and practical applications in computer algebra systems, such as Maple. The main focus of his research is in accelerating of popular symbolic and numeric algorithms. This includes accelerated symbolic summation, accelerated high-precision evaluation of
hypergeometric series, accelerated numeric computation over regular intervals with chains of
recurrences, and optimized compiling in general.

 

Contact

David Langstroth dll@cs.dal.ca