My favorites | Sign in
Project Home Downloads Wiki Issues Source
Search
for
RuntimePentiumM_2010_08  
Benchmark results for Pentium-M Linux 32bit
Deprecated
Updated Feb 19, 2011 by peter.ab...@gmail.com

Comments:

  • Fixed a bug where overhead was being measured.
    • Did not change results significantly since last benchmark because an earlier change reduced overhead so much.
    • Explains some strange results seen a while ago.
  • To see how different libraries handle memory caches differently three different matrix shapes are transposed.
    • Tall, Wide, Square.
    • In the future all but square will be turned off by default.
  • Changed mult aTb to abT
    • This was done to favor libraries with row-major matrices and specially built operator.
    • mult ab favors column-major libraries.
    • Was a test to see if some mult performance difference was because of row-major/column-major design decision.
  • EJML, ojAlgo, UJMP updated.
  • Explicitly noted that UJMP has extra libraries added to it that are needed to get the performance shown below.

Links to the results:

Test Environment

Test Environment:

Date 2010.08.02
OS Ubuntu 8.04
CPU PentiumM 1.7 GHz - 1 core
JVM Java HotSpot(TM) 64-Bit Server VM 1.6.0_16
Benchmark 0.7

Tested Libraries:

Name Version
Colt 1.2
Commons Math 2.1
EJML 0.14
Jama 1.0.2
JBlas 1.0.2
MTJ 0.9.12
OjAlgo 29.24
Parallel Colt 0.9.4
UJMP 0.2.5 with 3rd party

Pure Java Libraries

These results show the performance of libraries that have code written entirely in Java.

Java: Basic Operation Results

Java: Solving Linear Systems

Java: Matrix Decompositions


Mixed Java and Native Libraries

These results show the performance of libraries that either use pure Java or calls to native libraries.

Java: Basic Operation Results

Java: Solving Linear Systems

Java: Matrix Decompositions


Sign in to add a comment
Powered by Google Project Hosting