Download Folie 1 - FLI

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
- 2013 -
3D Structures of Biological Macromolecules
Floating Point Operations
and Benchmarking of Computer Performance
Jürgen Sühnel
[email protected]
Leibniz Institute for Age Research, Fritz Lipmann Institute,
Jena Centre for Bioinformatics
Jena / Germany
Supplementary Material: http://www.fli-leibniz.de/www_bioc/3D/
Floating Point Operation
- A Floating Point Operation is a mathematical operation with floating point numbers.
- Floating point number:
23.452 = 23452 x 10-3
significant digits x base exponent
- They cover easily extremely large number ranges between astronomy
and particle physics.
- A Floating Point Unit is a part of a computer system specifically designed to carry out
operations on floating point numbers.
- FLOPS is in computing to a measure of computer performance -floating point operations per second (note plural and per second).
Floating Point Operation
Name
FLOPS
yottaFLOPS
10
zettaFLOPS
10
exaFLOPS
10
petaFLOPS
10
teraFLOPS
10
gigaFLOPS
10
megaFLOPS
10
kiloFLOPS
10
24
21
18
15
12
9
6
3
Floating Point Operation – Theoretical Peak Performance
GFLOPS =
(CPU-Takt in GHz) ×
(Anzahl der CPU-Kerne) ×
(CPU-Instruktionen pro Takt) ×
(Anzahl der CPU im Rechnerknoten)
Beispiel: Zwei Sockel-Server mit 8 Kernen und einer Taktfrequenz von 2.6 GHz
und 8 Instruktionen per Zyklus (z.B. Intel E5-2600-Reihe)
GFLOPS = 2.6 x 8 x 8 x 2 = 332.8 GFLOPS
Adjusted Peak Performance is significantly lower.
Benchmarking
LINPACK –
software library for performing numerical linear algebra on computers,
is now widely used for measuring computer performance
Linpack Benchmark - Java Version´- (http://www.netlib.org/benchmark/linpackjava/)
The problem solved is a dense 500x500 system of linear equations with one right hand side, Ax=b.
The matrix is generated randomly and the right hand side is constructed so the solution has all components
equal to one.
800 – 1100 Mflops/s
TOP500 List
Measured in Pflops.
TOP500 List