Gostovanje američkog profesora na Računarskom faklutetu

U ponedeljak 11.01.2010. u 10 časova (učionica broj 4), na Računarskom fakultetu, prof. Wayne Hayes, sa američkog Univerziteta u Kaliforniji, održaće predavanje sa temom: “Fast and Exact Summation of Floating-Point Numbers“.

Tematsko predavanje: Fast and Exact Summation of Floating-Point Numbers

Professor Wayne Hayes, Computer Science Department, University of California, Irvine, USA


Floating-point calculations contain error because every computation includes a roundoff error. For example, given two floating-point numbers a and b, a program that assigns “c=a+b” will find that c
is close, but not exactly equal to, the exact sum a+b. In some systems (such as chaotic systems), such small “roundoff” error accumulate to give solutions that eventually go far from the exact solution.

In this talk, prof. Hayes will give some examples of fast, exact floating point calculations. In particular, he will give a brief history of exact floating-point summation, and then introduce several new algorithms for fast, efficient, exact summation.

Leave a Reply