Some unusual profiling data
Posted: Sat May 14, 2011 9:14 pm
I've collected some profiling data in a rather unusual way and I'm not even sure whether it's any use at all.
Anyway, what I did is measuring the frame times without the rendering, for every single frame. I've collected between 50'000 and 1'000'000 data points for each scenario.
Then Matlab was kind enough to plot a histogram of these 'tick times'. And I would like to share this data with you:
The first one was an empty level without any ship at all, but with the match started:
Data plus 1 on a log10 scale with wider x axis:
Second one was a match with 50 bots, but I didn't participate:
Data plus 1 on a log10 scale with wider x axis:
The last scenario is similar to the first one, but PresentationDM instead of EmptyLevel:
Data plus 1 on a log10 scale with wider x axis:
Anyway, what I did is measuring the frame times without the rendering, for every single frame. I've collected between 50'000 and 1'000'000 data points for each scenario.
Then Matlab was kind enough to plot a histogram of these 'tick times'. And I would like to share this data with you:
The first one was an empty level without any ship at all, but with the match started:
- empty.png (6.4 KiB) Viewed 15972 times
- empty_log.png (7.89 KiB) Viewed 15972 times
- 50bots.png (7.58 KiB) Viewed 15972 times
- 50bots_log.png (7.04 KiB) Viewed 15972 times
- dm.png (7.33 KiB) Viewed 15970 times
- dm_log.png (7.25 KiB) Viewed 15970 times