By @dnl0x00 A question that often comes into my mind when doing computations with Octave or other numerical libraries is how good do they utilize the CPU. Recently I was interested in estimating the number of GFLOPS that can be achieved with Octave for the element-wise matrix multiplication. For an... [Read More]

By @dnl0x00 Octave is a tool for doing linear algebra. When doing linear algebra one of the most fundamental operations is the matrix multiplication. A straightforward algorithm to do a matrix operation requires O(n³) multiplications. In 196 Strassen developed an algorithm for multiplying big matrices (n > 100 to 1000)... [Read More]

By @dnl0x00 When compiling ATLAS (Automatically Tuned Linear Algebra Software) on modern CPUs it may happen that you get the following error message: CPU Throttling apparently enabled! It appears you have cpu throttling enabled, which makes timings unreliable and an ATLAS install nonsensical. Aborting. This is because today most operating... [Read More]

By @dnl0x00 On my host system I have installed Linux Mint 17 and on that host I’m running Docker with Ubuntu 14.10. Recently I wanted to call lsof in a Docker container and I got the following permission denied error: root@ada77fe6ef60:/tmp $ lsof -p 751 COMMAND PID USER FD TYPE... [Read More]

By @dnl0x00 Recently I played around with Octave that I’ve installed on Ubuntu 14.10 running in a Docker container. The host system on which Docker is running is a Linux Mint 17 (Qiana). I was wondering whether there is a significant overhead induced by Docker when doing big matrix multiplications.... [Read More]