WebGPURenderer: Remove _forceViewport

[cabanier]

This addresses review #31134 (comment)

@Mugen87 , does this diff implement what you were looking for in #31134 (comment)? Or there calls to setSize I can remove?

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

	Before	After	Diff
WebGL	337.26 78.64	337.26 78.64	+0 B +0 B
WebGPU	552.95 153.31	552.92 153.31	-29 B -5 B
WebGPU Nodes	552.3 153.16	552.27 153.15	-29 B -6 B

🌳 Bundle size after tree-shaking

Minimal build including a renderer, camera, empty scene, and dependencies.

	Before	After	Diff
WebGL	468.44 113.28	468.44 113.28	+0 B +0 B
WebGPU	627.97 170	627.94 169.98	-28 B -14 B
WebGPU Nodes	582.82 159.33	582.79 159.32	-28 B -8 B

[cabanier]

I verified that with this change, all the surfaces are correctly sized and multisampled:
Surface 1 | 1500x1000 | color 32bit, depth 24bit, stencil 0 bit, MSAA 1, Mode: 1 (HwBinning) | 4 768x512 bins ( 4 rendered) | 0.33 ms | 10 stages : Binning : 0.147ms Render : 0.142ms StoreColor : 0.008ms Blit : 0.005ms
Surface 2 | 1500x1000 | color 32bit, depth 0 bit, stencil 0 bit, MSAA 1, Mode: 2 (SwBinning) | 2 768x1024 bins ( 2 rendered) | 0.17 ms | 4 stages : Render : 0.162ms StoreColor : 0.004ms
Surface 3 | 800 x800 | color 32bit, depth 24bit, stencil 0 bit, MSAA 1, Mode: 1 (HwBinning) | 2 480x800 bins ( 2 rendered) | 0.10 ms | 6 stages : Binning : 0.025ms Render : 0.054ms StoreColor : 0.004ms Blit : 0.004ms
Surface 4 | 800 x800 | color 32bit, depth 0 bit, stencil 0 bit, MSAA 1, Mode: 2 (SwBinning) | 1 864x800 bins ( 1 rendered) | 0.06 ms | 2 stages : Render : 0.06ms StoreColor : 0.002ms
Surface 5 | 1280x800 | color 32bit, depth 24bit, stencil 0 bit, MSAA 1, Mode: 2 (SwBinning) | 3 480x800 bins ( 4 rendered) | 0.11 ms | 7 stages : Render : 0.08ms StoreColor : 0.007ms Blit : 0.004ms
Surface 6 | 1280x800 | color 32bit, depth 0 bit, stencil 0 bit, MSAA 1, Mode: 2 (SwBinning) | 2 864x864 bins ( 2 rendered) | 0.13 ms | 4 stages : Render : 0.117ms StoreColor : 0.004ms
Surface 7 | 1680x1760 | color 32bit, depth 24bit, stencil 0 bit, MSAA 4, Mode: 1 (HwBinning) | 63 192x256 bins ( 63 rendered) | 2.68 ms | 129 stages : Binning : 0.25ms Render : 1.161ms StoreColor : 0.132ms Blit : 0.005ms Preempt : 0.743ms
Surface 8 | 1680x1760 | color 32bit, depth 24bit, stencil 0 bit, MSAA 4, Mode: 2 (SwBinning) | 66 288x160 bins ( 36 rendered) | 1.03 ms | 71 stages : Render : 0.563ms StoreColor : 0.277ms Blit : 0.006ms
Surface 9 | 1680x1760 | color 32bit, depth 24bit, stencil 0 bit, MSAA 4, Mode: 0 (Direct) | 1 1680x1760 bins ( 1 rendered) | 0.00 ms | 1 stages : Render : 0.002ms

[Mugen87]

Awesome! That is what I was hoping for! The PR fixes the calls of state.viewport() and the dimensions of internal framebuffer target is now kept in sync with the output render target.

Or there calls to setSize I can remove?

With this PR, calls like below should be obsolete since the dimensions are now extracted from the output render target. But I suggest we do this with a second PR (maybe targeting r178).

three.js/src/renderers/common/XRManager.js

Lines 1023 to 1024 in 2dba5a8

	renderer.setPixelRatio( 1 );
	renderer.setSize( glBaseLayer.framebufferWidth, glBaseLayer.framebufferHeight, false );

[cabanier]

With this PR, calls like below should be obsolete since the dimensions are now extracted from the output render target. But I suggest we do this with a second PR (maybe targeting r178).

ok. Will you let me know when that is?

[mrdoob]

You can see the dates for upcoming releases here:

https://github.com/mrdoob/three.js/milestones

(I'm a day late for the one this month. I'll probably be two days late...)