Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Jump to bottom

internal/graphicscommand: optimize prependPreservedUniforms #3144

Open
11 tasks
hajimehoshi opened this issue Oct 25, 2024 · 2 comments
Open
11 tasks

internal/graphicscommand: optimize prependPreservedUniforms #3144

hajimehoshi opened this issue Oct 25, 2024 · 2 comments
Labels
performance
Milestone
v2.9.0

Comments

@hajimehoshi
Copy link
Owner

hajimehoshi commented Oct 25, 2024 •
edited
Loading

Operating System

  • Windows
  • macOS
  • Linux
  • FreeBSD
  • OpenBSD
  • Android
  • iOS
  • Nintendo Switch
  • PlayStation 5
  • Xbox
  • Web Browsers

What feature would you like to be added?

prependPreservedUniforms tends to be a hot spot for performance, especially when rendering a lot of sprites. Here is the profiling result of examples/sprites with 10000 sprites:

image

atlas.Image.drawTriangles is the heaviest process, but this does a little complicated things. Let's focus on prependPreservedUniforms.

There are some ideas to optimize this.

Why is this needed?

No response
@hajimehoshi hajimehoshi added this to the v2.9.0 milestone Oct 25, 2024
@hajimehoshi
Copy link
Owner Author

BCE result:

% go build -gcflags=-d=ssa/check_bce ./internal/graphicscommand
# github.com/hajimehoshi/ebiten/v2/internal/graphicscommand
internal/graphicscommand/command.go:51:17: Found IsSliceInBounds
internal/graphicscommand/command.go:228:16: Found IsInBounds
internal/graphicscommand/command.go:229:16: Found IsInBounds
internal/graphicscommand/image.go:179:28: Found IsInBounds
internal/graphicscommand/image.go:183:28: Found IsInBounds
internal/graphicscommand/image.go:185:62: Found IsSliceInBounds
internal/graphicscommand/image.go:217:7: Found IsInBounds
internal/graphicscommand/image.go:234:16: Found IsInBounds
internal/graphicscommand/image.go:234:31: Found IsInBounds
internal/graphicscommand/commandqueue.go:139:23: Found IsInBounds
internal/graphicscommand/commandqueue.go:180:19: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:264:14: Found IsInBounds
internal/graphicscommand/commandqueue.go:275:17: Found IsInBounds
internal/graphicscommand/commandqueue.go:297:43: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:297:52: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:300:11: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:301:11: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:304:23: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:326:10: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:351:15: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:367:11: Found IsInBounds
internal/graphicscommand/commandqueue.go:368:11: Found IsInBounds
internal/graphicscommand/commandqueue.go:409:11: Found IsInBounds
internal/graphicscommand/commandqueue.go:410:11: Found IsInBounds
internal/graphicscommand/commandqueue.go:416:11: Found IsInBounds
internal/graphicscommand/commandqueue.go:417:11: Found IsInBounds
internal/graphicscommand/commandqueue.go:461:14: Found IsInBounds
internal/graphicscommand/commandqueue.go:463:19: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:536:10: Found IsSliceInBounds
internal/graphicscommand/image_notjs.go:100:14: Found IsSliceInBounds
internal/graphicscommand/command.go:247:53: Found IsInBounds
internal/graphicscommand/command.go:248:53: Found IsInBounds
internal/graphicscommand/commandqueue.go:138:36: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:152:37: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:155:29: Found IsSliceInBounds
internal/graphicscommand/commandqueue.go:350:34: Found IsSliceInBounds

hajimehoshi added a commit that referenced this issue Oct 26, 2024
@hajimehoshi
internal/graphicscommand: add BenchPrependPreservedUniforms
9f5f53c 
Updates #3144
hajimehoshi added a commit that referenced this issue Oct 26, 2024
@hajimehoshi
internal/graphicscommand: improve performance of prependPreservedUnif…
1fe988e 
…orms

```
% benchstat old.txt new.txt
goos: darwin
goarch: arm64
pkg: github.com/hajimehoshi/ebiten/v2/internal/graphicscommand
cpu: Apple M3 Pro
                            │   old.txt   │               new.txt               │
                            │   sec/op    │   sec/op     vs base                │
PrependPreservedUniforms-12   23.07n ± 2%   19.96n ± 0%  -13.48% (p=0.000 n=10)
```

Updates #3144
hajimehoshi added a commit that referenced this issue Oct 26, 2024
@hajimehoshi
internal/graphicscommand: move the const check outside of the function
ecc42d4 
old.txt: 1fe988e
new.txt: this commit

```
benchstat old.txt new.txt
goos: darwin
goarch: arm64
pkg: github.com/hajimehoshi/ebiten/v2/internal/graphicscommand
cpu: Apple M3 Pro
                            │   old.txt   │              new.txt               │
                            │   sec/op    │   sec/op     vs base               │
PrependPreservedUniforms-12   19.79n ± 4%   18.84n ± 2%  -4.80% (p=0.001 n=10)
```

Updates #3144
# for free to join this conversation on GitHub. Already have an account? # to comment
Assignees
No one assigned
Labels
performance
Projects
None yet
Milestone
v2.9.0
Development

No branches or pull requests
2 participants
@hajimehoshi and others