[Merged by Bors] - Optimize rendering slow-down at high entity counts by TheRawMeatball · Pull Request #5509 · bevyengine/bevy

TheRawMeatball · 2022-07-31T11:50:37Z

Objective

Improve High entity counts cause decreased rendering performance even when rendering nothing. #3953

Solution

The very specific circumstances under which the render world is reset meant that the flush_as_invalid function could be replaced with one that had a noop as its init method.
This removes a double-writing issue leading to greatly increased performance.

Running the reproduction code in the linked issue, this change nearly doubles the framerate.

rparrett · 2022-07-31T23:30:11Z

Neat. This is the most dramatic one of these I've seen.

just the reserve_and_flush span (repro from issue)

cart · 2022-08-01T00:34:25Z

Wow yeah thats a big win right there. Pretty sure this is valid. I'll give this a proper think asap.

TheRawMeatball · 2022-08-01T07:25:31Z

Changed again to a fully custom flush method that directly uses memset, this is getting even larger gains.

TheRawMeatball · 2022-08-01T12:36:33Z

Currently UB, working on a fix.

TheRawMeatball · 2022-08-01T13:11:28Z

Reverted the memset, so this works again, but others investigating why the memset version failed via UB would be appreciated.

TheRawMeatball · 2022-08-01T13:27:18Z

UB fixed with help from @PROMETHIA-27

rparrett · 2022-08-01T13:44:41Z

b02a1d7 (repro from issue)
just the reserve_and_flush span

rparrett · 2022-08-01T14:19:44Z

many_sprites frame time

many_spites frame time (run 2)

many_cubes sphere frame time

In all cases, yellow / this trace is this PR, red is main. Guessing we're just in the noise a bit and this is not a quiet machine at the moment. All traces were ~30 seconds though.

examples/hello_world.rs

cart · 2022-08-08T22:22:01Z

crates/bevy_ecs/src/archetype.rs

 };

 #[derive(Debug, Copy, Clone, Eq, PartialEq, Hash)]
+#[repr(transparent)]


Probably makes sense to add a comment to ArchetypeId::INVALID saying that it must be equivalent to setting all bytes of memory to u8::MAX (and referencing flush_and_reserve_invalid). Don't want anyone changing this here and then implicitly breaking flush_and_reserve_invalid.

StarArawn · 2022-08-11T12:54:31Z

I'm excited for this PR as it seems like it fixes part of the performance issues here! I'm still worried that this isn't a complete fix though. This code is still reserving entities even if they have been despawned which leads to memory leakage(increased memory usage over time) and possibly over time slow performance. I can think of quite a few scenarios where users would spawn high entity counts and despawn said entities. A shmup would be one such example where you would be spawning hundreds or even thousands of bullets which get depsawned.
Looking at it closer it looks like it just reserves the number of entities that currently exist. I still think performance could be an issue for me, but I think a better solution might be to have a separate world.

mbolt35 · 2022-08-27T23:38:27Z

Just checking on the general status of this PR. I noticed it was removed from the 0.8.1 release. Curious as to when we might expect this to be part of mainline. Thanks!

hankjordan · 2022-09-03T18:44:08Z

Noticeable performance improvement for my game as well - uses extremely high entity count, yellow is this PR, red is 0.8.0:

cart · 2022-10-24T19:24:30Z

bors r+

# Objective - Improve #3953 ## Solution - The very specific circumstances under which the render world is reset meant that the flush_as_invalid function could be replaced with one that had a noop as its init method. - This removes a double-writing issue leading to greatly increased performance. Running the reproduction code in the linked issue, this change nearly doubles the framerate. Co-authored-by: Carter Anderson <mcanders1@gmail.com>

bors · 2022-10-24T19:40:46Z

Pull request successfully merged into main.

Build succeeded:

# Objective - Improve bevyengine#3953 ## Solution - The very specific circumstances under which the render world is reset meant that the flush_as_invalid function could be replaced with one that had a noop as its init method. - This removes a double-writing issue leading to greatly increased performance. Running the reproduction code in the linked issue, this change nearly doubles the framerate. Co-authored-by: Carter Anderson <mcanders1@gmail.com>

better flush

dcaefaf

TheRawMeatball added A-Rendering Drawing game state to the screen C-Performance A change motivated by improving speed, memory usage or compile times labels Jul 31, 2022

superdump requested a review from cart July 31, 2022 14:34

cart added this to the Bevy 0.8.1 milestone Aug 1, 2022

TheRawMeatball added 2 commits August 1, 2022 09:12

more docs

aeb5230

Add assert for added safety

a5f8c03

TheRawMeatball force-pushed the optimize-large-entity-counts branch from 038f2a5 to a5f8c03 Compare August 1, 2022 06:16

TheRawMeatball added 2 commits August 1, 2022 10:18

switch to ridiculously fast memset

a745ad6

fix docs

9b0e298

TheRawMeatball added 3 commits August 1, 2022 10:30

final fixes

9f106d8

mirror safety comment on EntityMeta

e283fc6

add repr C

8a32577

add DO_UB flag

becb9e7

fix ub

b02a1d7

mockersf reviewed Aug 1, 2022

View reviewed changes

examples/hello_world.rs Show resolved Hide resolved

revert example

e8be324

cart reviewed Aug 8, 2022

View reviewed changes

add safety comment.

d500cbe

StarArawn mentioned this pull request Aug 11, 2022

Memory Leak + Reduced Framerate in 0.7 StarArawn/bevy_ecs_tilemap#218

Closed

cart removed this from the Bevy 0.8.1 milestone Aug 18, 2022

inodentry pushed a commit to BroovyJammy/bevy that referenced this pull request Aug 19, 2022

PR bevyengine#5509

3df47e1

inodentry added a commit to BroovyJammy/bevy that referenced this pull request Aug 19, 2022

Merge PR bevyengine#5509

a76290d

alice-i-cecile added P-High This is particularly urgent, and deserves immediate attention X-Needs-SME This type of work requires an SME to approve it. labels Sep 3, 2022

terhechte mentioned this pull request Sep 20, 2022

Really slow rendering of tiled 3d floors #6032

Closed

cart added this to the Bevy 0.9 milestone Sep 22, 2022

Merge remote-tracking branch 'origin/main' into pr/TheRawMeatball/5509

2fea9ed

bors bot changed the title ~~Optimize rendering slow-down at high entity counts~~ [Merged by Bors] - Optimize rendering slow-down at high entity counts Oct 24, 2022

bors bot closed this Oct 24, 2022

JMS55 mentioned this pull request Oct 26, 2022

High entity counts cause decreased rendering performance even when rendering nothing. #3953

Closed

Uh oh!

Conversation

TheRawMeatball commented Jul 31, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Objective

Solution

Uh oh!

rparrett commented Jul 31, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cart commented Aug 1, 2022

Uh oh!

TheRawMeatball commented Aug 1, 2022

Uh oh!

TheRawMeatball commented Aug 1, 2022

Uh oh!

TheRawMeatball commented Aug 1, 2022

Uh oh!

TheRawMeatball commented Aug 1, 2022

Uh oh!

rparrett commented Aug 1, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rparrett commented Aug 1, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

cart Aug 8, 2022

Choose a reason for hiding this comment

Uh oh!

StarArawn commented Aug 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mbolt35 commented Aug 27, 2022

Uh oh!

hankjordan commented Sep 3, 2022

Uh oh!

cart commented Oct 24, 2022

Uh oh!

bors bot commented Oct 24, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

TheRawMeatball commented Jul 31, 2022 •

edited

Loading

rparrett commented Jul 31, 2022 •

edited

Loading

rparrett commented Aug 1, 2022 •

edited

Loading

rparrett commented Aug 1, 2022 •

edited

Loading

StarArawn commented Aug 11, 2022 •

edited

Loading