mtl12 Devlog, Seite 3
Ältere Einträge aus dem Devlog. Projekt, Stand und Galerie stehen auf der ersten Seite.
Einträge
Die Einträge sind auf Englisch verfasst. 72 Einträge, neueste zuerst. Diese Seite zeigt 052 bis 041.
- Entry 052
Setting only what changed: 10000 draws from 20 ms to 7.5 ms, and a CPU clock that was 40 times too low
The D3D12-to-Metal layer now sets per draw only the GPU state that differs from the previous draw, which cut a 10000-draw benchmark frame from 20.3 to 7.5 ms, and a Rosetta clock quirk had hidden where the time went.
- Entry 051
A benchmark harness, and the GPU timestamps it caught lying
The D3D12-to-Metal layer now has a repeatable performance harness with a first baseline, and building it exposed GPU timestamps that returned a value from an earlier frame.
- Entry 050
Fullscreen and HDR: Windows' window rules from Wine, HDR10 from the Mac's own EDR
D3D12 games on the Mac layer can now go exclusive fullscreen and output HDR10 or scRGB, which macOS shows as extended dynamic range on an XDR screen; Wine's own DXGI conformance test now runs against the layer.
- Entry 049
Work graphs on a Mac: a GPU that schedules its own work, one level at a time
D3D12 work graphs, where shaders launch other shaders by writing records, now run on an M1 Mac through a level-by-level scheduler ported from vkd3d-proton; all ten vkd3d-proton work graph tests pass.
- Entry 048
Wide and antialiased lines on an M1: every line becomes a rectangle
D3D12's wide, narrow and alpha-antialiased lines are now drawn on an M1 Mac as rectangles made by a generated object function, and D24 depth buffers get D3D12's depth bias units.
- Entry 047
Variable-rate shading on an M1: four threads share the work of a coarse pixel
D3D12 variable-rate shading, where one pixel-shader run colours a block of up to 4x4 pixels, is now emulated on an M1 Mac by having each 2x2 group of shader threads shade coarse pixels and share the results.
- Entry 046
Conservative rasterization on an M1: draw a bigger polygon
D3D12 conservative rasterization, which Metal lacks on an M1 Mac, is now emulated by replacing each triangle with a slightly larger polygon before Metal rasterizes it.
- Entry 045
Depth bounds on an M1: test against a copy of the depth buffer
The D3D12 depth bounds test, which only Apple's newest GPUs have natively, is now emulated on an M1 Mac by reading a copy of the depth buffer in the pixel shader.
- Entry 044
Sampler feedback with 32-bit atomics: one region, two words
D3D12 sampler feedback, which records which parts of a texture a shader sampled at which detail level, now works on an M1 Mac without 64-bit texture atomics.
- Entry 043
Sharing a fence: a handle that knows when it is closed
D3D12 fences can now be shared through Windows handles inside one process, with the fence kept alive exactly as long as a handle or a fence object refers to it.
- Entry 042
One draw, four views: view instancing by replay
D3D12 view instancing, where one draw renders the same geometry for up to four cameras, now works on Metal by replaying each draw once per view.
- Entry 041
Rasterizer-ordered views, and a pixel shader that Metal runs four times
D3D12's rasterizer-ordered views now work on Metal, after finding that Metal runs a pixel shader that writes memory once per covered sample in multisampled passes.