feat: More comprehensive tracking of possible side effects by cieplypolar · Pull Request #2695 · software-mansion/TypeGPU

cieplypolar · 2026-07-07T10:54:19Z

accurate tracking of possibleSideEffects
passing $gpuCallable to the accessors
tpgu.rawCodeSnippet optional possibleSideEffects field

Closes #2646.

- Mark function arguments as side-effect-free - Mark tgpu.const reads as side-effect-free - Propagate operand side-effects in pure binary operators - Add side-effect tracking to dualImpl (matching callableSchema) - Mark buffer reads (uniform/readonly/mutable) as side-effect-free - Add comprehensive ternary runtime tests from issue #2587 - Update existing ternary test for runtime select() generation

… discard sideEffects

…tection

github-actions · 2026-07-07T10:55:29Z

pkg.pr.new

packages
Ready to be installed by your favorite package manager ⬇️

https://pkg.pr.new/software-mansion/TypeGPU/typegpu@df3997a695e8368e6487b7e6c1dfd3da419ba019

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/cli@df3997a695e8368e6487b7e6c1dfd3da419ba019

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/noise@df3997a695e8368e6487b7e6c1dfd3da419ba019

https://pkg.pr.new/software-mansion/TypeGPU/@typegpu/react@df3997a695e8368e6487b7e6c1dfd3da419ba019

https://pkg.pr.new/software-mansion/TypeGPU/unplugin-typegpu@df3997a695e8368e6487b7e6c1dfd3da419ba019

benchmark
view benchmark

commit
view commit

github-actions · 2026-07-07T10:56:40Z

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

🟢 Decreased	➖ Unchanged	🔴 Increased (max 2.93%)	❔ Unknown
0	36	285	0

`import * as ...` in PR vs `import * as ...` in target (did bundle size increase?):

Click to reveal the results table (148 entries).

Test	tsdown
d_ref.ts	4.94 kB ($${\color{red}+2.9\%}$$)
d_bool.ts	13.20 kB ($${\color{red}+1.3\%}$$)
d_f16.ts	13.20 kB ($${\color{red}+1.3\%}$$)
d_f32.ts	13.20 kB ($${\color{red}+1.3\%}$$)
d_i32.ts	13.20 kB ($${\color{red}+1.3\%}$$)
d_u32.ts	13.20 kB ($${\color{red}+1.3\%}$$)
d_u16.ts	13.22 kB ($${\color{red}+1.3\%}$$)
std_discard.ts	14.49 kB ($${\color{red}+1.3\%}$$)
std_copy.ts	14.73 kB ($${\color{red}+1.2\%}$$)
std_arrayLength.ts	14.74 kB ($${\color{red}+1.2\%}$$)
d_textureDepth2d.ts	13.64 kB ($${\color{red}+1.2\%}$$)
d_textureDepthCube.ts	13.65 kB ($${\color{red}+1.2\%}$$)
d_texture1d.ts	13.65 kB ($${\color{red}+1.2\%}$$)
d_texture2d.ts	13.65 kB ($${\color{red}+1.2\%}$$)
d_texture3d.ts	13.65 kB ($${\color{red}+1.2\%}$$)
d_textureCube.ts	13.66 kB ($${\color{red}+1.2\%}$$)
d_textureDepth2dArray.ts	13.66 kB ($${\color{red}+1.2\%}$$)
d_textureDepthCubeArray.ts	13.66 kB ($${\color{red}+1.2\%}$$)
d_textureDepthMultisampled2d.ts	13.67 kB ($${\color{red}+1.2\%}$$)
d_texture2dArray.ts	13.67 kB ($${\color{red}+1.2\%}$$)
d_textureCubeArray.ts	13.67 kB ($${\color{red}+1.2\%}$$)
d_textureMultisampled2d.ts	13.68 kB ($${\color{red}+1.2\%}$$)
std_dpdx.ts	15.45 kB ($${\color{red}+1.2\%}$$)
std_dpdxCoarse.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_dpdxFine.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_dpdy.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_dpdyCoarse.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_dpdyFine.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_fwidth.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_fwidthCoarse.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_fwidthFine.ts	15.46 kB ($${\color{red}+1.2\%}$$)
std_atomicLoad.ts	16.25 kB ($${\color{red}+1.1\%}$$)
std_atomicStore.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_textureBarrier.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicAdd.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicAnd.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicMax.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicMin.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicOr.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicSub.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_atomicXor.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_storageBarrier.ts	16.26 kB ($${\color{red}+1.1\%}$$)
std_workgroupBarrier.ts	16.26 kB ($${\color{red}+1.1\%}$$)
d_vec2b.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec2f.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec2h.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec2i.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec2u.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec3b.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec3f.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec3h.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec3i.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec3u.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec4b.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec4f.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec4h.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec4i.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_vec4u.ts	19.63 kB ($${\color{red}+0.9\%}$$)
d_formatToWGSLType.ts	21.12 kB ($${\color{red}+0.8\%}$$)
d_uint8.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float16.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float16x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float16x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float32.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float32x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float32x3.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_float32x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint16.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint16x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint16x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint32.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint32x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint32x3.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint32x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint8.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint8x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_sint8x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_snorm16.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_snorm16x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_snorm16x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_snorm8.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_snorm8x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_snorm8x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint16.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint16x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint16x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint32.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint32x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint32x3.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint32x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint8x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_uint8x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm10_10_10_2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm16.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm16x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm16x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm8.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm8x2.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm8x4.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_unorm8x4_bgra.ts	21.13 kB ($${\color{red}+0.8\%}$$)
d_packedFormats.ts	21.15 kB ($${\color{red}+0.8\%}$$)
d_isPackedData.ts	21.19 kB ($${\color{red}+0.8\%}$$)
d_alignmentOf.ts	22.08 kB ($${\color{red}+0.8\%}$$)
std_subgroupAdd.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupAll.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupAnd.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupAny.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupBallot.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupBroadcast.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupBroadcastFirst.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupElect.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupExclusiveAdd.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupExclusiveMul.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupInclusiveAdd.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupInclusiveMul.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupMax.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupMin.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupMul.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupOr.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupShuffle.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupShuffleDown.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupShuffleUp.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupShuffleXor.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_subgroupXor.ts	24.58 kB ($${\color{red}+0.7\%}$$)
std_textureDimensions.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureGather.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureLoad.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSample.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSampleBaseClampToEdge.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSampleBias.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSampleCompare.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSampleCompareLevel.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSampleGrad.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureSampleLevel.ts	26.19 kB ($${\color{red}+0.7\%}$$)
std_textureStore.ts	26.19 kB ($${\color{red}+0.7\%}$$)
d_isBuiltin.ts	24.81 kB ($${\color{red}+0.7\%}$$)
d_sizeOf.ts	24.86 kB ($${\color{red}+0.7\%}$$)
d_isContiguous.ts	24.86 kB ($${\color{red}+0.7\%}$$)
d_getLongestContiguousPrefix.ts	24.87 kB ($${\color{red}+0.7\%}$$)
d_size.ts	26.54 kB ($${\color{red}+0.6\%}$$)
d_align.ts	26.54 kB ($${\color{red}+0.6\%}$$)
d_location.ts	26.55 kB ($${\color{red}+0.6\%}$$)
d_interpolate.ts	26.55 kB ($${\color{red}+0.6\%}$$)
d_invariant.ts	26.92 kB ($${\color{red}+0.6\%}$$)
std_pack2x16float.ts	36.29 kB ($${\color{red}+0.5\%}$$)
std_pack4x8unorm.ts	36.29 kB ($${\color{red}+0.5\%}$$)
std_unpack2x16float.ts	36.29 kB ($${\color{red}+0.5\%}$$)
std_unpack4x8unorm.ts	36.29 kB ($${\color{red}+0.5\%}$$)

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

Test	tsdown
tgpu_init.ts	256.85 kB ($${\color{green}-3.2\%}$$)
tgpu_initFromDevice.ts	256.33 kB ($${\color{green}-3.4\%}$$)
tgpu_resolve.ts	187.56 kB ($${\color{green}-29.3\%}$$)
tgpu_resolveWithContext.ts	187.50 kB ($${\color{green}-29.4\%}$$)
tgpu_bindGroupLayout.ts	69.09 kB ($${\color{green}-74.0\%}$$)
tgpu_mutableAccessor.ts	66.09 kB ($${\color{green}-75.1\%}$$)
tgpu_accessor.ts	66.08 kB ($${\color{green}-75.1\%}$$)
tgpu_privateVar.ts	65.38 kB ($${\color{green}-75.4\%}$$)
tgpu_workgroupVar.ts	65.38 kB ($${\color{green}-75.4\%}$$)
tgpu_const.ts	64.60 kB ($${\color{green}-75.7\%}$$)
tgpu_fragmentFn.ts	38.04 kB ($${\color{green}-85.7\%}$$)
tgpu_fn.ts	38.01 kB ($${\color{green}-85.7\%}$$)
tgpu_vertexFn.ts	37.85 kB ($${\color{green}-85.7\%}$$)
tgpu_computeFn.ts	37.56 kB ($${\color{green}-85.9\%}$$)
tgpu_vertexLayout.ts	26.78 kB ($${\color{green}-89.9\%}$$)
tgpu_comptime.ts	14.47 kB ($${\color{green}-94.5\%}$$)
tgpu_unroll.ts	1.66 kB ($${\color{green}-99.4\%}$$)
tgpu_slot.ts	1.57 kB ($${\color{green}-99.4\%}$$)
tgpu_lazy.ts	1.22 kB ($${\color{green}-99.5\%}$$)

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

github-actions · 2026-07-07T10:57:20Z

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 [0.91, 1.88, 4.14, 6.33, 7.40, 11.52, 22.06, 23.11]
  line [0.98, 1.91, 4.04, 5.82, 7.72, 11.60, 21.99, 23.88]
  line [0.86, 1.79, 3.82, 6.46, 7.70, 10.18, 21.75, 23.31]

---
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 [0.30, 0.50, 0.69, 0.84, 1.12, 1.13, 1.39, 1.60]
  line [0.32, 0.56, 0.69, 0.81, 1.13, 1.20, 1.42, 1.57]
  line [0.28, 0.52, 0.73, 0.84, 1.15, 1.23, 1.38, 1.59]

---
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 [0.82, 2.08, 4.15, 6.54, 12.55, 27.12, 54.52, 111.19]
  line [0.87, 2.13, 4.11, 6.17, 12.11, 26.31, 52.73, 108.80]
  line [0.83, 2.06, 3.52, 6.64, 12.23, 26.34, 54.24, 109.77]

Copilot

Pull request overview

This PR extends the TGSL snippet system to track possibleSideEffects more precisely, enabling additional safe codegen optimizations (notably allowing runtime-value ternaries when both branches are side-effect-free) and adds/expands tests around side-effect classification.

Changes:

Thread possibleSideEffects through snippet creation/conversion paths (generator, conversion helpers, resource handles, raw snippets, etc.).
Relax ternary restrictions by permitting runtime conditions when both branches are proven side-effect-free (compiled to select()).
Expand side-effect test coverage across literals, resources, accessors, indexing, conversions, refs, and logical operators.

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
packages/typegpu/tests/utils/parseResolved.ts	Improves return-expression extraction by using typed expression generation when return type is known; broadens `expectSideEffects` input types.
packages/typegpu/tests/tgsl/sideEffects.test.ts	Adds extensive side-effect classification tests for many TGSL constructs and resource reads.
packages/typegpu/src/tgsl/wgslGenerator.ts	Propagates `possibleSideEffects` across generated snippets and updates conditional/select and operator logic.
packages/typegpu/src/tgsl/generationHelpers.ts	Preserves `possibleSideEffects` through concretization/coercion helpers.
packages/typegpu/src/tgsl/forOfUtils.ts	Marks range/default iteration snippets as side-effect-free constants.
packages/typegpu/src/tgsl/conversion.ts	Preserves `possibleSideEffects` across snippet conversions/actions.
packages/typegpu/src/std/boolean.ts	Updates default precision snippet to be explicitly side-effect-free.
packages/typegpu/src/resolutionCtx.ts	Ensures snippet resolution preserves `possibleSideEffects`.
packages/typegpu/src/data/snippet.ts	Clarifies semantics/docs for `possibleSideEffects` and documents `snip`’s default behavior.
packages/typegpu/src/data/ref.ts	Threads `possibleSideEffects` through ref/deref operations.
packages/typegpu/src/core/variable/tgpuVariable.ts	Marks variable `.$` snippet generation as side-effect-free.
packages/typegpu/src/core/unroll/tgpuUnroll.ts	Preserves `possibleSideEffects` when resolving/unrolling iterables.
packages/typegpu/src/core/texture/texture.ts	Marks texture-view handle snippets as side-effect-free.
packages/typegpu/src/core/texture/externalTexture.ts	Marks external texture handle snippets as side-effect-free.
packages/typegpu/src/core/slot/accessor.ts	Avoids eagerly executing GPU-callable functions during accessor resolution; propagates `possibleSideEffects` when calling functions.
packages/typegpu/src/core/sampler/sampler.ts	Marks sampler handle snippets as side-effect-free.
packages/typegpu/src/core/rawCodeSnippet/tgpuRawCodeSnippet.ts	Adds `possibleSideEffects` parameter to raw snippets and preserves it in `.$` and resolution.
packages/typegpu/src/core/function/dualImpl.ts	Sets function call snippet `possibleSideEffects` using `options.sideEffects` and/or argument side effects.
apps/typegpu-docs/src/content/docs/apis/utils.mdx	Documents ternary relaxation for side-effect-free branches and the new `possibleSideEffects` raw snippet parameter.

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

pullfrog

ℹ️ No critical issues — minor suggestions inline.

Reviewed changes — this PR threads possibleSideEffects through TypeGPU's snippet resolution and codegen so runtime ternaries can safely lower to select() when both branches are pure.

Propagate side-effect bits across snip() call sites in accessors, refs, conversions, struct/array constructors, unary/binary operators, loops, raw snippets, and DualImpl calls.
Mark pure sources as side-effect-free: literals, constants, parameters, variable reads, buffer/texture/sampler handles, and pure builtins.
Keep impure calls conservative: user-defined functions and builtins like discard/atomics remain possibleSideEffects: true.
Expand test coverage in sideEffects.test.ts for pure and impure expression shapes, and update docs to reflect that runtime ternary conditions are allowed when branches are pure.

^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

pullfrog

ℹ️ No new issues in the latest push; prior feedback still stands.

Reviewed changes since last review — the latest push restores the atomicLoad example in DualImplOptions.sideEffects docs and switches the punctuation back to ASCII arrows/hyphens.

Docs-only polish in packages/typegpu/src/core/function/dualImpl.ts.

^{｜ Fix it ➔ ｜ View workflow run ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

pullfrog

Important

The latest push addresses the prior ! constant-fallback feedback, but the updated float NaN guard now misclassifies 0.0 as NaN.

Reviewed changes — the latest commit switches the unary ! constant fallback to possibleSideEffects: false and polishes the side-effect test suite, including renaming tests, using std.isBeingTranspiled, broadening expectSideEffects, and routing returns through _typedExpression in the test helper.

Addressed prior feedback in packages/typegpu/src/tgsl/wgslGenerator.ts by returning possibleSideEffects: false for the constant ! fallback.
Polished sideEffects.test.ts with clearer test names, reorganized cases, additional ! and conversion coverage, and std.* builtins.
Updated parseResolved.ts to type return expressions against the function return type and accept TgpuFn in expectSideEffects.

^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

pullfrog

ℹ️ No new issues in the latest push; prior feedback still stands.

Reviewed changes since last review — added coverage for zero-argument dualFn accessors and polished a side-effect test name.

Added allows zero-argument dualFn test in packages/typegpu/tests/accessor.test.ts to verify tgpu.accessor(d.bool, std.subgroupElect).$ resolves correctly.
Renamed side-effect test in packages/typegpu/tests/tgsl/sideEffects.test.ts from "pure function access from accessor" to "pure function call from accessor".

^{｜ Fix it ➔ ｜ View workflow run ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

pullfrog

✅ No new issues found.

Reviewed changes — since the last review the branch picked up the latest main build-output changes and finished the mechanical possibleSideEffects refactor.

Merged latest main build-output changes into the branch, switching the typegpu package to a Bun-based build.mjs and removing tsdown.
Refactored snippet construction to use withValue and withDataType helpers across accessor.ts, resolutionCtx.ts, tgpuUnroll.ts, ref.ts, conversion.ts, and generationHelpers.ts.
Updated RefOperator and derefSnippet to propagate the operand's possibleSideEffects bit instead of hard-coding it.
Addressed oxlint feedback by moving an inline disable comment in resolutionCtx.ts.

^{｜ View workflow run ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

aleksanderkatan

Great work! 🐳⚔️

aleksanderkatan · 2026-07-08T13:07:29Z

  }

-  return snip(stitch`(*${snippet})`, innerType, snippet.origin);
+  return snip(stitch`(*${snippet})`, innerType, snippet.origin, false);


is this always an identifier? why not repeat the snippet possibleSideEffects?

maybe it is not

let ptr = *(&(array[impureInt()]))

pullfrog

✅ No new issues found.

Reviewed changes since the last review — the latest commit refined side-effect propagation in accessor resolution and ref dereferencing.

Used ctx.resolveSnippet(snippet) in AccessorBase[$resolve] so the resolved snippet preserves dataType, origin, and possibleSideEffects through the central helper.
Updated derefSnippet to propagate the operand's possibleSideEffects bit instead of forcing false, matching the RefOperator resolution paths.

Targeted tests and the typecheck both pass.

^{｜ View workflow run ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

iwoplaza

Awesome! 💜
Left a few nits

iwoplaza · 2026-07-09T13:46:57Z

  return new SnippetImpl(snippet.value, dataType, snippet.origin, snippet.possibleSideEffects);
 }

+export function withValue(value: string, snippet: Snippet): ResolvedSnippet;


It could be that snippet is a comptime-known string

Suggested change

export function withValue(value: string, snippet: Snippet): ResolvedSnippet;

export function withValue(value: string, snippet: ResolvedSnippet): ResolvedSnippet;

iwoplaza · 2026-07-09T13:57:05Z

@@ -1048,11 +1054,7 @@ export class ResolutionCtxImpl implements ResolutionCtx {
  }

  resolveSnippet(snippet: Snippet): ResolvedSnippet {


Could you update the documentation for resolveSnippet in types.ts? I believe it no longer reflects the new implementation.

iwoplaza · 2026-07-09T13:58:15Z

+const impureVec = () => {
+  'use gpu';
+  return d.vec3f(6, 6, 6);
+};
+const impureInt = () => {
+  'use gpu';
+  return 666;
+};
+const impureMat = () => {
+  'use gpu';
+  return d.mat2x2f(6, 6, 6, 6);
+};
+const impureStruct = () => {
+  'use gpu';
+  return Boid();
+};
+const impureBool = () => {
+  'use gpu';
+  return true;
+};


These might soon become pure, once we can determine that a function call is side-effect free. I would make these truly impure by just updating a shared private variable.

iwoplaza · 2026-07-09T14:25:57Z

 }

-export function expectSideEffects(cb: () => unknown): Assertion<boolean> {
+export function expectSideEffects(cb: TgpuFn | (() => unknown)): Assertion<boolean> {


Suggested change

export function expectSideEffects(cb: TgpuFn | (() => unknown)): Assertion<boolean> {

export function expectSideEffects(cb: () => unknown): Assertion<boolean> {

pullfrog Bot and others added 16 commits June 15, 2026 19:16

style: apply oxfmt formatting

59ebd23

fix: add sideEffects prop to dualImpl for correct side-effect tracking

faee545

make sideEffects required in DualImplOptions

5bc6920

chore: merge ternaryRuntime tests into ternaryOperator.test.ts

6538f87

chore: address review feedback — use snip param for side-effects, fix…

0d6d656

… discard sideEffects

fix: user-defined functions should assume side-effects are possible

3b6467e

Merge branch 'main' into pullfrog/2587-improve-ternary-side-effect-de…

c797858

…tection

chore: clarify docs for sideEffects and possibleSideEffects

2dca023

work

96cd096

work + tests + docs

8660054

test fix

e9c7882

docs update

5c29027

Merge branch 'main' into fix/possible-side-effects-flow

a878f01

final polish

bbb2dfc

Copilot AI review requested due to automatic review settings July 7, 2026 10:54

Copilot started reviewing on behalf of cieplypolar July 7, 2026 10:54 View session

cieplypolar force-pushed the fix/possible-side-effects-flow branch from 7435193 to befe8de Compare July 7, 2026 10:55

relict of pullfrog

6d5ee16

cieplypolar force-pushed the fix/possible-side-effects-flow branch from befe8de to 6d5ee16 Compare July 7, 2026 10:56

Copilot AI reviewed Jul 7, 2026

View reviewed changes

Comment thread packages/typegpu/src/tgsl/wgslGenerator.ts

Comment thread packages/typegpu/tests/tgsl/sideEffects.test.ts

pullfrog Bot reviewed Jul 7, 2026

View reviewed changes

Comment thread packages/typegpu/src/tgsl/wgslGenerator.ts Outdated

Comment thread packages/typegpu/tests/tgsl/sideEffects.test.ts Outdated

pullfrog Bot reviewed Jul 7, 2026

View reviewed changes

review changes

731cb19

pullfrog Bot reviewed Jul 7, 2026

View reviewed changes

Comment thread packages/typegpu/src/tgsl/wgslGenerator.ts

Comment thread packages/typegpu/tests/tgsl/sideEffects.test.ts

more pull frog relicts

1bc02b3

cieplypolar assigned cieplypolar and unassigned cieplypolar Jul 7, 2026

cieplypolar requested review from aleksanderkatan, iwoplaza and reczkok July 7, 2026 11:46

accessor dualFn test

c6ee8ec

pullfrog Bot reviewed Jul 7, 2026

View reviewed changes

iwoplaza reviewed Jul 7, 2026

View reviewed changes

Comment thread packages/typegpu/src/core/unroll/tgpuUnroll.ts Outdated

Comment thread packages/typegpu/src/core/unroll/tgpuUnroll.ts Outdated

Comment thread packages/typegpu/src/data/ref.ts Outdated

review changes

ad575e8

cieplypolar requested a review from iwoplaza July 7, 2026 16:26

cieplypolar and others added 2 commits July 7, 2026 18:50

oxlint

81a30b1

Merge branch 'main' into fix/possible-side-effects-flow

b58883a

pullfrog Bot reviewed Jul 8, 2026

View reviewed changes

aleksanderkatan approved these changes Jul 8, 2026

View reviewed changes

cieplypolar and others added 2 commits July 8, 2026 16:26

Merge branch 'main' into fix/possible-side-effects-flow

f62da06

review changes

df3997a

pullfrog Bot reviewed Jul 8, 2026

View reviewed changes

iwoplaza approved these changes Jul 9, 2026

View reviewed changes

iwoplaza changed the title ~~impr: tracking possibleSideEffects~~ feat: More comprehensive tracking of possible side effects Jul 10, 2026

	export function withValue(value: string, snippet: Snippet): ResolvedSnippet;
	export function withValue(value: string, snippet: ResolvedSnippet): ResolvedSnippet;

		@@ -1048,11 +1054,7 @@ export class ResolutionCtxImpl implements ResolutionCtx {
		}

		resolveSnippet(snippet: Snippet): ResolvedSnippet {

	export function expectSideEffects(cb: TgpuFn \| (() => unknown)): Assertion<boolean> {
	export function expectSideEffects(cb: () => unknown): Assertion<boolean> {

Uh oh!

Uh oh!

Conversation

cieplypolar commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

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

import * as ... in PR vs import * as ... in target (did bundle size increase?):

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

Uh oh!

github-actions Bot commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Resolution Time Benchmark

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

pullfrog Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

pullfrog Bot left a comment

Choose a reason for hiding this comment

Uh oh!

pullfrog Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

pullfrog Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pullfrog Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aleksanderkatan left a comment

Choose a reason for hiding this comment

Uh oh!

aleksanderkatan Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

cieplypolar Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

pullfrog Bot left a comment

Choose a reason for hiding this comment

Uh oh!

iwoplaza left a comment

Choose a reason for hiding this comment

Uh oh!

iwoplaza Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

iwoplaza Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

iwoplaza Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

iwoplaza Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

cieplypolar commented Jul 7, 2026 •

edited

Loading

github-actions Bot commented Jul 7, 2026 •

edited

Loading

github-actions Bot commented Jul 7, 2026 •

edited

Loading

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

`import * as ...` in PR vs `import * as ...` in target (did bundle size increase?):

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

github-actions Bot commented Jul 7, 2026 •

edited

Loading