Skip to content

Minor CPU optimizations in cpu_froxel_parallel scope - #602

Closed
Baplar wants to merge 3 commits into
DoubleStyx:masterfrom
Baplar:froxel-cpu-parallel-opt-1
Closed

Baplar wants to merge 3 commits into
DoubleStyx:masterfrom
Baplar:froxel-cpu-parallel-opt-1

Conversation

@Baplar

@Baplar Baplar commented May 30, 2026

Copy link
Copy Markdown
Collaborator

No description provided.

@Baplar
Baplar force-pushed the froxel-cpu-parallel-opt-1 branch 2 times, most recently from 1a6a61e to 9356373 Compare May 31, 2026 17:08
Baplar added 3 commits June 2, 2026 08:27
It turns out that the amount of effort needed to parallelize this
operation and then transpose the Vec<Vec<_>> is more costly in time than
just looping sequentially, even (especially?) with large numbers of
iterations.
This allows us to do parallelize the 16 depth slices, which do not have
dependencies on each other. We also do a bit of intermediate values
caching as a cherry on top.
@Baplar
Baplar force-pushed the froxel-cpu-parallel-opt-1 branch from 9356373 to 19a9e56 Compare June 2, 2026 06:27
@DoubleStyx

Copy link
Copy Markdown
Owner

Merged, thank you! Sorry for the delays, I'll try to get to these in less than a day next time. Working on the other PRs as well. Was quite busy lately.

Apologies again for the hash change not causing a clean merge. I usually check out prs and then work on them over a day or two, so if they get updated during that time then the hash will be stale.

@DoubleStyx DoubleStyx closed this Jun 3, 2026
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