Parallel Template Library (PTL)™ Java Matrix Multiplication Comparison

Progeneric's Parallel Template Library (PTL)™ for Java and .NET contains many powerful and easy to use parallel functions that take advantage of multicore processors. In this video, we compare PTL's Java matrix operations to other Java matrix libraries with a multiplication example.

PTL is compared to Linear Algebra for Java (la4j), Efficient Java Matrix Library (EJML) and Parallel Colt. When we multiplied two dense matrices A (5000 x 4000) and B (4000 x 5000), la4j had the slowest performance, so it was used as the baseline. EJML was 19.0x faster than la4j, and Parallel Colt was 20.4x faster than la4j. PTL had the best performance due to its parallel multiplication method and innovative load balancing techniques and was 312.5x faster than la4j.

[ProgenericSystems]

Sorry for the confusion. We recently split Parallel Template Library (PTL) into separate, smaller components, and the Matrices component is not ready yet. It will be available for free trial and purchase in the fourth quarter of 2013. We also updated the video description with this new information.

[ProgenericSystems]

We used la4j version 0.3.0 and the Basic2DMatrix (we briefly show the code that we used around 1:42 in the video).

[cloropingo]

What PTL component provides matrix calculations? Array algorithms?

[spffe]

What version of la4j did you use? And what kind of premitives for dense matrices?

[cloropingo]

Hi, I downloaded the Algorithms and Containers free trials, but I can't find the classes you are using... am I missing something?. ***This is a rather duplicate question, since my previous one was about the same issue***