Closed JamieJQuinn closed 3 years ago
This is actually not as big a problem as I thought it was. Results of some basic experiments watching top
while running remove-restore:
n update grids | region size (deg) | grid spacing (deg) | approx memory usage (GB) |
---|---|---|---|
2 | 0.5 x 0.2 | 1e-3 | 5 |
2 | 0.5 x 0.2 | 1e-4 | 7 |
2 | 1 x 1 | 1e-3 | 6 |
2 | 1 x 1 | 1e-4 | 8 |
2 | 5 x 5 | 1e-3 | 8 |
2 | 5 x 5 | 1e-4 | Out of memory... (>20GB) |
@Devaraj-G:
This no longer appears to be an issue.
Currently the memory usage of the remove-restore algorithm is very high (on the order of tens of GB) when outputting a grid of very fine resolution (e.g. 10m). This should be more of a priority now that we're polishing the remove-restore functionality.