linear_indexing

Check navigator.gpu?.wgslLanguageFeatures?.has('linear_indexing') before choosing a shader that uses this language feature. It is available automatically when supported; requires linear_indexing; 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.

Compute entry points can declare @builtin(workgroup_index) and @builtin(global_invocation_index) as u32 inputs. They provide linear numbering of workgroups and invocations, complementing the three-component workgroup_id and global_invocation_id built-ins.

Flat buffers and image-processing kernels often need one index per invocation. The global index flattens global_invocation_id across the whole dispatch, with x varying fastest. The workgroup index similarly flattens workgroup_id. For multidimensional workgroups, grouping all invocations of one workgroup together would give a different ordering from this global index.

A fallback can compute global_invocation_id.x + global_invocation_id.y * dispatchWidth + global_invocation_id.z * dispatchWidth * dispatchHeight, where the dispatch width and height are the corresponding workgroup dimensions multiplied by the dispatched workgroup counts. These indices do not specify execution order. Bounds-check accesses when the dispatched invocation count exceeds the number of elements to process.