Allows the engine to render directly to the screen's back buffer, bypassing intermediate framebuffers when possible.

This change introduces a new mDirectTarget flag to indicate whether the rendering target is the back buffer. When rendering directly, it uses a compose task instead of a blit task and avoids unnecessary framebuffer operations.

This improves rendering performance and reduces memory consumption in scenarios where direct rendering is feasible.

The WASM binding requires modification to enable this feature.

        attrs.depth = true;
        attrs.stencil = true;
        attrs.antialias = true;

Related issues: #3951

Evaluation

Summary

• No regressions in any example.
• Large animation: +0.3% overall FPS, +1.2% steady-state FPS.
• Lottie animations: +3.4% overall FPS, +4.6% steady-state FPS (with higher jitter).
• Simple animation (normal resolution): +3.0% overall FPS, steady-state is the same.
• Simple animation (4K): +44% overall FPS, +36% steady-state FPS (with slightly lower jitter).
• WebGL multi-animation test: After a 1-minute run, both builds stabilize at 60 FPS with similar memory ranges. No meaningful difference observed.

Overall, the feature branch is performance-safe and shows significant improvements at high resolutions, especially in 4K scenarios.

1. Large Animation

Overall FPS

• Origin: 25.66
• Feature: 25.74
• Change: +0.31%

Steady-State FPS (last 20 intervals)

• Origin: 25.49
• Feature: 25.79
• Change: +1.18%

Jitter (stddev/mean)

• Origin: 1.71%
• Feature: 2.18%
• Relative change: +27%

Conclusion: Tiny gain, effectively identical to origin. No regression.

2. Lottie Animation

Overall FPS

• Origin: 34.28
• Feature: 35.45
• Change: +3.41%

Steady-State FPS

• Origin: 34.75
• Feature: 36.36
• Change: +4.62%

Jitter

• Origin: 3.21%
• Feature: 5.51%
• Relative change: +72%

Conclusion: Clear 3–5% FPS gain with higher variability. Still stable.

3. Small Animation (Normal Resolution)

Overall FPS

• Origin: 187.20
• Feature: 192.72
• Change: +2.95%

Steady-State FPS

• Origin: 239.55
• Feature: 239.33
• Change: −0.09%

Jitter

• Origin: 0.10%
• Feature: 0.14%
• Relative change: +43%

Conclusion: +3% overall from faster warm-up; steady-state identical.

4. Small Animation (4K Resolution)

Overall FPS

• Origin: 120.01
• Feature: 173.07
• Change: +44.21%

Steady-State FPS

• Origin: 119.04
• Feature: 161.46
• Change: +35.64%

Jitter

• Origin: 0.85%
• Feature: 0.67%
• Change: −21%

Conclusion: Major performance uplift at 4K — both average and steady-state FPS improve significantly.

5. Resolution Scaling (Normal → 4K)

Origin Branch

• Overall FPS: −35.89%
• Steady-State FPS: −50.31%

Feature Branch

• Overall FPS: −10.20%
• Steady-State FPS: −32.54%

Conclusion: The feature branch is far more resolution-robust.
Origin loses half its steady-state FPS, while the feature retains much more performance.

6. WebGL Multi-Animation Test (Qualitative)

Tested using the WebGL/DevTools HUD for 1 minute.

Feature Build

• FPS: 60 (range 29–60)
• Memory counter: ~20 MB (range 19–25 MB)

Origin Build

• FPS: 60 (range 29–60)
• Memory counter: ~23 MB (range 19–25 MB)

Conclusion:

• FPS stabilizes at 60 on both builds.
• The HUD memory number fluctuates and stays in the same range for both builds.
• Treat this as no meaningful difference and, importantly, no regression.

Notes

• I also make sure that direct rendering is enabled in WebGL with a tool, but the performance remains exactly the same.