You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Tests by @atrayano found that at high resolution using jemalloc really seemed to help with the memory usage. Indeed, it seems to be true at all resolutions.
For example, at C24 it seems to go from 23% to 18%. At C720, I ran three runs each with and without using jemalloc and:
The higher lines are the 'stock' runs and the lower are 'jemalloc'.
Adding this support to GEOS is not hard at all and I'll use this as a meta-issue to track PRs.
But I suppose the question (as both @tclune and @atrayano have thought as well): Is this "real"? Or are we perhaps not tracking the right thing?
Tests by @atrayano found that at high resolution using jemalloc really seemed to help with the memory usage. Indeed, it seems to be true at all resolutions.
For example, at C24 it seems to go from 23% to 18%. At C720, I ran three runs each with and without using jemalloc and:
The higher lines are the 'stock' runs and the lower are 'jemalloc'.
Adding this support to GEOS is not hard at all and I'll use this as a meta-issue to track PRs.
But I suppose the question (as both @tclune and @atrayano have thought as well): Is this "real"? Or are we perhaps not tracking the right thing?
And then does it work at NAS?
Repos
Run Tests
Trickle Tests
The text was updated successfully, but these errors were encountered: