Skip to content

Conversation

@gkjohnson
Copy link
Contributor

Related to #632

In progress PR for queuing up all tiles needed for rendering so we can make sure the jobs are saturated as consistently as possible. Currently tiles can seem slower but it's possible this is because the queue sort is prioritizing tiles that were not previously added to the queue, preventing some parents from loading sooner.

TODO

  • Check if raising the queue size on tiles has a significant improvement before / after
  • Measure time to load completion
  • Ideate on queue sort improvements

@gkjohnson
Copy link
Contributor Author

gkjohnson commented Jul 23, 2024

Generally this seems like an improvement for loading google tiles in the "aerial" example by a couple hundred ms when the parse and download queue max job count has been raised to 1000. Without the increased job cap there is little to no difference. Both queues must be changed to see the benefit.

Before
~ 1969ms

After
~1725ms

We're potentially at risk of filling up the lru cache too quickly causing one branch of the tree to render and not others. Ideally we'd only push items onto the queue if it's an improvement over existing tiles.

@gkjohnson
Copy link
Contributor Author

Closing for now - the impact isn't significant in common cases to be worth it.

@gkjohnson gkjohnson closed this Sep 13, 2024
@gkjohnson gkjohnson deleted the load-improvements branch September 13, 2024 01:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants