Closed KatieWoe closed 4 years ago
With published version: Start: 13 MB 1 min: 14.8 MB 2 min: 15.5 MB 3 min: 15.9 MB 4 min: 16.1 MB 5 min: 16.5 MB
@jbphet requested the test be done on recently published States of Matter. Start: 27.2 1 min: 56.6 2 min: 74.9 3 min: 94.2 4 min: 119 5 min: 137
I did a memory analysis and found the leaks. There were three nodes for which there were no dispose functions, so they have been added. Kinda a lame of me to have missed these, but it's great that QA ran this test and identified the problem.
Running a RequireJS version of Atomic Interactions, I now see the following memory heap size:
Time (minutes) | Memory Usage (MB) |
---|---|
0 | 17.3 |
1 | 22.6 |
2 | 22.8 |
3 | 24.5 |
4 | 25.3 |
5 | 25.6 |
6 | 25.5 |
7 | 25.7 |
8 | 26.0 |
9 | 26.2 |
10 | 26.2 |
This is a pretty reasonable profile. The memory is increasing very gradually, and decreases between minutes 5 and 6, and holds steady between minutes 9 and 10. I did some comparisons between the heap snapshots, and there was nothing obvious from the sim that was leaking. If there is a slow leak, it's probably somewhere in common code, and it's not enough to cause any problems.
Bottom line: I think this is now fixed.
For https://github.com/phetsims/QA/issues/216 Start: 17.5 MB 1 min: 128 MB 2 min: 231 MB 3 min: 334 MB 4 min: 441 MB 5 min: crash before heap snapshot can finish. Will try test on published version per request of @jbphet