fix: Lazy buffer views in bitcast to allow for polyfill

[iwoplaza]

No description provided.

[Copilot]

Pull request overview

This PR adjusts the CPU-side bitcast implementation to instantiate typed-array buffer views lazily, preventing eager Float16Array construction so environments relying on a late-loaded polyfill don’t fail during module initialization.

Changes:

• Introduces a small cache object to hold lazily-created typed array views.
• Replaces eagerly-created bufViews typed arrays with getters that allocate on first access.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

pkg.pr.new

packages
Ready to be installed by your favorite package manager ⬇️

https://pkg.pr.new/software-mansion/TypeGPU/eslint-plugin-typegpu@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/tgpu-gen@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/tinyest-for-wgsl@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/typegpu@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/cli@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/color@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/gl@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/noise@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/radiance-cascades@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/react@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/sdf@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/three@df7067aedbba2558f15ada4c338483ed861bb1d2

https://pkg.pr.new/software-mansion/TypeGPU/unplugin-typegpu@df7067aedbba2558f15ada4c338483ed861bb1d2

benchmark
view benchmark

commit
view commit

[github-actions]

Resolution Time Benchmark

---
config:
  themeVariables:
    xyChart:
      plotColorPalette: "#E63946, #3B82F6, #059669"
---
xychart
  title "Random Branching (🔴 PR | 🔵 main | 🟢 release)"
  x-axis "max depth" [1, 2, 3, 4, 5, 6, 7, 8]
  y-axis "time (ms)"
  line [0.72, 1.30, 2.60, 4.39, 4.83, 8.57, 15.45, 17.06]
  line [0.68, 1.34, 2.56, 4.27, 4.92, 6.96, 14.06, 16.29]
  line [0.66, 1.31, 2.85, 4.21, 5.54, 8.59, 14.32, 16.67]

Loading

---
config:
  themeVariables:
    xyChart:
      plotColorPalette: "#E63946, #3B82F6, #059669"
---
xychart
  title "Linear Recursion (🔴 PR | 🔵 main | 🟢 release)"
  x-axis "max depth" [1, 2, 3, 4, 5, 6, 7, 8]
  y-axis "time (ms)"
  line [0.21, 0.45, 0.52, 0.65, 0.82, 0.88, 1.01, 1.10]
  line [0.20, 0.35, 0.50, 0.55, 0.77, 0.82, 0.96, 1.08]
  line [0.19, 0.41, 0.48, 0.62, 0.80, 0.85, 0.99, 1.07]

Loading

---
config:
  themeVariables:
    xyChart:
      plotColorPalette: "#E63946, #3B82F6, #059669"
---
xychart
  title "Full Tree (🔴 PR | 🔵 main | 🟢 release)"
  x-axis "max depth" [1, 2, 3, 4, 5, 6, 7, 8]
  y-axis "time (ms)"
  line [0.67, 1.59, 2.58, 4.15, 7.96, 16.23, 36.43, 72.83]
  line [0.67, 1.68, 2.60, 4.34, 7.88, 16.46, 37.14, 74.30]
  line [0.65, 1.53, 3.07, 4.22, 8.09, 17.28, 37.19, 76.38]

Loading

[github-actions]

Bundle size comparison (import * as ... in PR vs import * as ... in target):

🟢 Decreased	➖ Unchanged	🔴 Increased (max 0.29%)	❔ Unknown
0	296	26	0

import { ... } in PR vs import * as ... in PR (is the library tree-Shakeable?):

Test	tsdown
tgpu_init.ts	259.04 kB ($${\color{green}-3.2\%}$$)
tgpu_initFromDevice.ts	258.52 kB ($${\color{green}-3.4\%}$$)
tgpu_resolve.ts	164.59 kB ($${\color{green}-38.5\%}$$)
tgpu_resolveWithContext.ts	164.53 kB ($${\color{green}-38.5\%}$$)
tgpu_bindGroupLayout.ts	68.97 kB ($${\color{green}-74.2\%}$$)
tgpu_mutableAccessor.ts	65.97 kB ($${\color{green}-75.3\%}$$)
tgpu_accessor.ts	65.96 kB ($${\color{green}-75.3\%}$$)
tgpu_privateVar.ts	65.31 kB ($${\color{green}-75.6\%}$$)
tgpu_workgroupVar.ts	65.30 kB ($${\color{green}-75.6\%}$$)
tgpu_const.ts	64.55 kB ($${\color{green}-75.9\%}$$)
tgpu_fn.ts	38.14 kB ($${\color{green}-85.7\%}$$)
tgpu_fragmentFn.ts	38.14 kB ($${\color{green}-85.7\%}$$)
tgpu_vertexFn.ts	37.96 kB ($${\color{green}-85.8\%}$$)
tgpu_computeFn.ts	37.66 kB ($${\color{green}-85.9\%}$$)
tgpu_vertexLayout.ts	26.79 kB ($${\color{green}-90.0\%}$$)
tgpu_comptime.ts	14.48 kB ($${\color{green}-94.6\%}$$)
tgpu_unroll.ts	1.66 kB ($${\color{green}-99.4\%}$$)
tgpu_slot.ts	1.54 kB ($${\color{green}-99.4\%}$$)
tgpu_lazy.ts	1.19 kB ($${\color{green}-99.6\%}$$)

If you wish to run a comparison for other, slower bundlers, run the 'Tree-shake test' from the GitHub Actions menu.

[pullfrog]

Reviewed changes — converts eagerly-created TypedArray buffer views into lazy getters with ??= caching so that importing the module doesn't throw when Float16Array is unavailable at load time (e.g. before a polyfill initializes).

• Lazy bufViews in bitcast.ts — replaces 4 eagerly-allocated TypedArray views (f32, u32, i32, f16) with getter-backed lazy initialization. Each view is still allocated at most once per module lifetime via ??= in bufViewCache.

✅ No new issues found.

  ^{｜ View workflow run ｜ Using DeepSeek Pro (free via Pullfrog for OSS) (Kimi K2 not used — the program covers this model; add its provider key to run your pick) ｜ 𝕏}