fragment_depth

Check navigator.gpu?.wgslLanguageFeatures?.has('fragment_depth') before choosing a shader that uses this language feature. It is available automatically when supported; requires fragment_depth; documents the dependency and rejects the shader on unsupported implementations. This language token is not a requiredFeatures entry for requestDevice(); any additional device requirements are described below.

Writing frag_depth typically stops the GPU from depth-testing a fragment before its shader runs, because the final depth is unknown until then. With this language feature, the output can be declared as @builtin(frag_depth, less) - a promise that the written value is less than or equal to the fragment’s rasterized depth - or @builtin(frag_depth, greater) for greater than or equal.

Depth-correct impostors such as sphere and capsule billboards, parallax occlusion mapping, and decals that nudge depth all move depth in a known direction. Pairing the promise with the depth comparison is what helps: with a "less" compare function, a greater promise means a fragment that already fails the test can never pass after shading, so it can be rejected early.

Breaking the promise is not validated; the spec allows an indeterminate depth value to be used instead, so clamp the written value against @builtin(position).z when in doubt. Plain @builtin(frag_depth) remains valid everywhere, so the fallback is simply to omit the depth mode and accept the slower path.