BWT in javascript
As an idea for improvement of save file compression in js-like game, I implemented Burrows-Wheeler transform in javascript. For now, the code is very simple and needs further optimization.
Table of Contents
Blog
FFT continued
I've implemented the slight change that reorders twiddle factors, so they are accessed sequentially in the recombination phase of the FFT. As expected, the data locality improved the performance drastically.
script load / error handling
While modifying OAT.Map API to use new loading scheme Ondrej Zara wrote for OAT javascript toolkit, we ran upon issues and inconsistencies regarding script elements events handling in several common browsers (MSIE 6, 7, Firefox 3, Safari / Google Chrome and Opera). This entry summarizes the results.
OAT Fisheye widget optimization
Instead of working on my research assignment, I spend a while optimizing OAT toolkit's fisheye widget (oh yeah, procrastinating again…)
Installing Redmine on debian etch (4.0)
It's a pain. IMHO it is much easier to dist-upgrade and try again (if it helps?) than to go through this procedure. This document should serve as a how-to with working Apache 2.2/fcgi + MySQL 5 + Redmine 0.8.3 site on debian etch. Its more focused on getting this to work on a standalone server. If you know you have all the required packages and a database installed, then good guide is at the official redmine site [1].
Discrete laplacian eigenvalues, part 2
In the last entry, we ended having eigenfunctions and eigenvalues for continuous case. In this section, we plug this result into a discrete heat equation and obtain the stability restriction for explicit scheme.
Discrete laplacian eigenvalues
For a research assignment, which includes some numerical math, I needed to find largest eigenvalue of the discrete laplacian in two dimensions.
A not so fast Fourier transform
During evenigns of last week or so, I've been trying to write an FFT in C. The algo is pretty straighforward, so it won't seem to be a surprise it's been known for ~ 200 years. The original transform of k'th component of vector x0..x_N-1 looks like this:
<m>X_k = \sum_{n=0}^{N-1} x_n e^{-\frac{2 \pi i}{N} k}</m>
Complexity is O(n^2), but FFT can be done in O(n log2 n).
Archive
