|
O-Matrix Statistics Performance
The high performance of the O-Matrix language makes it ideal
for large scale statistical data analysis and simulation. The following
benchmark illustrates performance benefits for some of O-Matrix'
statistics functions.
| Benchmark |
O-Matrix 6.4 |
Matlab 7.01 |
| Sort 3,000,000 random values |
0.456 |
1.02 |
| Sort 30,000 random values |
0.004 |
0.007 |
| Column-wise mean of 100,000x25 random matrix |
0.003 |
0.023 |
| Column-wise standard deviation of 100,000x25 random matrix |
0.003 |
0.147 |
| Norm of 200,000 element random vector |
0.0006 |
0.0003 |
| Create 200,000x20 matrix of uniform distributed random numbers |
0.023 |
0.188 |
| Create 200,000x20 matrix of normally distributed random numbers |
0.095 |
0.164 |
| Create 200,000x20 matrix of log normal distributed random numbers |
0.089 |
NA* |
| Create 200,000x20 matrix of exponentially distributed random numbers |
0.048 |
NA |
| Exponential cumulative distribution function of 200,000x20 matrix |
0.041 |
NA |
| Exponential probability density function of 200,000x20 matrix |
0.050 |
NA |
| Column-wise mean absolute deviation of 100,000x25 random matrix |
0.165 |
NA |
| Column-wise median absolute deviation of 100,000x25 random matrix |
0.007 |
NA |
| Column-wise median of 50,000x25 random matrix |
0.128 |
0.331 |
| Convolution of two 2^14 element vectors |
0.011 |
2.08 |
| Covariance matrix for two 100,000 element vectors |
0.005 |
0.023 |
| Column-wise Kurtosis of 100,000x25 random matrix |
0.009 |
NA |
| Column-wise sum of 100,000x25 random matrix |
0.002 |
0.023 |
All timings are in seconds. - Run on a 3 GHz Pentium 4
NA*, Not available or requires additional toolbox
O-Matrix is my first choice for simulation and data analysis.
I could use C++, VB, or Fortran for this, but O-Matrix provides the
performance and convenience I need to bypass these languages.
- Ryan Franklin
|
