Timing simulations for BufferedMatrix 1.3.0 (Simulation 1)
Written by Ben Bolstad
email bmb@bmbolstad.com
|
Background
The goal here was to investigate the speed of the built-ins when only the column buffer is activated (and only one column is stored at a time)
Source code (note takes very long time to run),
Simulation times
createBufferedMatrix() times
Elementwise transformation times
Column Summarization
Row Summarization
Full Matrix Summarization
Memory usage
|
Matrix creation times
|
ReadOnly=True or ReadOnly=False
|
Looking at number of rows and columns (ReadOnly=True or ReadOnly=False)
|
Discussion
With large matrices most of the overhead is taken in loading/de-loading data from the disk. Should batch together summaries where possible.
|
|
|