This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Flops flop is an acronym for floating point operation. Counting the number of flops an algorithm requires to solve a problem allows us to compare – at least roughly – the relative speed of methods. (The term flops is also used by computer marketers to mean “floating point operations per second”, and is a measure of the speed of the computer). In older papers and textbooks flop stood for an operation like c ij = s + a ik * b kj , which included a floating point multiply, a floating point add, and some indexing work. Today (since the mid 1990’s) flop means one floating point operation (+ ,- , * , /, √ , > , etc.). (The older version of flop is thus about 2 of the newer flops). One could argue that + is faster than, e.g. √ , and thus shouldn’t be counted the same. This is true, and if an algorithm had a relatively large amount of √ ’s, then we would count them separately. In practice, counting flops gives only a rough way to compare algorithms, and since a > and a √ each require bringing floats into...
View Full Document
This note was uploaded on 12/18/2010 for the course PHYS 5073 taught by Professor Mark during the Fall '10 term at Arkansas.
- Fall '10