fix: External traversal for raw WGSL implemented functions#2648
fix: External traversal for raw WGSL implemented functions#2648aleksanderkatan wants to merge 33 commits into
Conversation
|
pkg.pr.new packages benchmark commit |
Bundle size comparison (
|
| 🟢 Decreased | ➖ Unchanged | 🔴 Increased | ❔ Unknown |
|---|---|---|---|
| 23 | 297 | 0 | 0 |
import { ... } in PR vs import * as ... in PR (is the library tree-Shakeable?):
Click to reveal the results table (297 entries).
| Test | tsdown |
|---|---|
| d_Void.ts | 87.40 kB ( |
| d_sampler.ts | 87.41 kB ( |
| d_isPtr.ts | 87.41 kB ( |
| d_comparisonSampler.ts | 87.42 kB ( |
| d_isSizeAttrib.ts | 87.41 kB ( |
| d_isWgslArray.ts | 87.41 kB ( |
| d_isAlignAttrib.ts | 87.41 kB ( |
| d_isWgslStruct.ts | 87.41 kB ( |
| d_isAtomic.ts | 87.41 kB ( |
| d_isBuiltinAttrib.ts | 87.42 kB ( |
| d_isLocationAttrib.ts | 87.42 kB ( |
| d_isDecorated.ts | 87.41 kB ( |
| d_isInterpolateAttrib.ts | 87.42 kB ( |
| d_atomic.ts | 87.41 kB ( |
| d_ptrHandle.ts | 87.41 kB ( |
| d_ptrUniform.ts | 87.41 kB ( |
| d_ptrStorage.ts | 87.41 kB ( |
| d_ptrPrivate.ts | 87.41 kB ( |
| d_ptrFn.ts | 87.41 kB ( |
| d_ptrWorkgroup.ts | 87.41 kB ( |
| d_textureExternal.ts | 87.42 kB ( |
| d_textureStorage1d.ts | 87.42 kB ( |
| d_textureStorage2d.ts | 87.42 kB ( |
| d_textureStorage3d.ts | 87.42 kB ( |
| d_textureStorage2dArray.ts | 87.42 kB ( |
| d_isDisarray.ts | 87.41 kB ( |
| d_isUnstruct.ts | 87.41 kB ( |
| d_isLooseDecorated.ts | 87.42 kB ( |
| d_isLooseData.ts | 87.41 kB ( |
| d_isWgslData.ts | 87.41 kB ( |
| d_unstruct.ts | 87.41 kB ( |
| d_isData.ts | 87.41 kB ( |
| d_deepEqual.ts | 87.41 kB ( |
| d_struct.ts | 87.41 kB ( |
| d_ref.ts | 87.40 kB ( |
| std_discard.ts | 103.14 kB ( |
| std_copy.ts | 103.13 kB ( |
| std_arrayLength.ts | 103.14 kB ( |
| std_isBeingTranspiled.ts | 103.15 kB ( |
| std_getTargetShaderLanguage.ts | 103.15 kB ( |
| std_extensionEnabled.ts | 103.15 kB ( |
| std_range.ts | 103.13 kB ( |
| std_dpdx.ts | 103.13 kB ( |
| std_fwidthCoarse.ts | 103.14 kB ( |
| std_dpdxCoarse.ts | 103.14 kB ( |
| std_dpdyCoarse.ts | 103.14 kB ( |
| std_fwidthFine.ts | 103.14 kB ( |
| std_dpdxFine.ts | 103.14 kB ( |
| std_dpdyFine.ts | 103.14 kB ( |
| std_fwidth.ts | 103.14 kB ( |
| std_dpdy.ts | 103.13 kB ( |
| d_bool.ts | 87.40 kB ( |
| d_f16.ts | 87.40 kB ( |
| d_f32.ts | 87.40 kB ( |
| d_i32.ts | 87.40 kB ( |
| d_u32.ts | 87.40 kB ( |
| d_u16.ts | 87.40 kB ( |
| std_atomicLoad.ts | 103.14 kB ( |
| std_textureBarrier.ts | 103.14 kB ( |
| std_atomicStore.ts | 103.14 kB ( |
| std_workgroupBarrier.ts | 103.15 kB ( |
| std_storageBarrier.ts | 103.14 kB ( |
| std_atomicAdd.ts | 103.14 kB ( |
| std_atomicAnd.ts | 103.14 kB ( |
| std_atomicMax.ts | 103.14 kB ( |
| std_atomicMin.ts | 103.14 kB ( |
| std_atomicSub.ts | 103.14 kB ( |
| std_atomicXor.ts | 103.14 kB ( |
| std_atomicOr.ts | 103.14 kB ( |
| d_textureDepth2d.ts | 87.41 kB ( |
| d_textureDepthCube.ts | 87.42 kB ( |
| d_texture1d.ts | 87.41 kB ( |
| d_texture2d.ts | 87.41 kB ( |
| d_texture3d.ts | 87.41 kB ( |
| d_textureDepth2dArray.ts | 87.42 kB ( |
| d_textureCube.ts | 87.41 kB ( |
| d_textureDepthCubeArray.ts | 87.42 kB ( |
| d_textureDepthMultisampled2d.ts | 87.43 kB ( |
| d_texture2dArray.ts | 87.41 kB ( |
| d_textureCubeArray.ts | 87.42 kB ( |
| d_textureMultisampled2d.ts | 87.42 kB ( |
| d_disarrayOf.ts | 87.41 kB ( |
| d_vec2b.ts | 87.41 kB ( |
| d_vec2f.ts | 87.41 kB ( |
| d_vec2h.ts | 87.41 kB ( |
| d_vec2i.ts | 87.41 kB ( |
| d_vec2u.ts | 87.41 kB ( |
| d_vec3b.ts | 87.41 kB ( |
| d_vec3f.ts | 87.41 kB ( |
| d_vec3h.ts | 87.41 kB ( |
| d_vec3i.ts | 87.41 kB ( |
| d_vec3u.ts | 87.41 kB ( |
| d_vec4b.ts | 87.41 kB ( |
| d_vec4f.ts | 87.41 kB ( |
| d_vec4h.ts | 87.41 kB ( |
| d_vec4i.ts | 87.41 kB ( |
| d_vec4u.ts | 87.41 kB ( |
| std_subgroupAdd.ts | 103.14 kB ( |
| std_subgroupBroadcastFirst.ts | 103.15 kB ( |
| std_subgroupExclusiveAdd.ts | 103.15 kB ( |
| std_subgroupExclusiveMul.ts | 103.15 kB ( |
| std_subgroupInclusiveAdd.ts | 103.15 kB ( |
| std_subgroupInclusiveMul.ts | 103.15 kB ( |
| std_subgroupShuffleDown.ts | 103.15 kB ( |
| std_subgroupShuffleXor.ts | 103.15 kB ( |
| std_subgroupBroadcast.ts | 103.15 kB ( |
| std_subgroupShuffleUp.ts | 103.15 kB ( |
| std_subgroupShuffle.ts | 103.14 kB ( |
| std_subgroupBallot.ts | 103.14 kB ( |
| std_subgroupElect.ts | 103.14 kB ( |
| std_subgroupAll.ts | 103.14 kB ( |
| std_subgroupAnd.ts | 103.14 kB ( |
| std_subgroupAny.ts | 103.14 kB ( |
| std_subgroupMax.ts | 103.14 kB ( |
| std_subgroupMin.ts | 103.14 kB ( |
| std_subgroupMul.ts | 103.14 kB ( |
| std_subgroupXor.ts | 103.14 kB ( |
| std_subgroupOr.ts | 103.14 kB ( |
| d_formatToWGSLType.ts | 87.42 kB ( |
| d_uint8.ts | 87.41 kB ( |
| d_unorm10_10_10_2.ts | 87.42 kB ( |
| d_unorm8x4_bgra.ts | 87.41 kB ( |
| d_float16x2.ts | 87.41 kB ( |
| d_float16x4.ts | 87.41 kB ( |
| d_float32x2.ts | 87.41 kB ( |
| d_float32x3.ts | 87.41 kB ( |
| d_float32x4.ts | 87.41 kB ( |
| d_snorm16x2.ts | 87.41 kB ( |
| d_snorm16x4.ts | 87.41 kB ( |
| d_unorm16x2.ts | 87.41 kB ( |
| d_unorm16x4.ts | 87.41 kB ( |
| d_sint16x2.ts | 87.41 kB ( |
| d_sint16x4.ts | 87.41 kB ( |
| d_sint32x2.ts | 87.41 kB ( |
| d_sint32x3.ts | 87.41 kB ( |
| d_sint32x4.ts | 87.41 kB ( |
| d_snorm8x2.ts | 87.41 kB ( |
| d_snorm8x4.ts | 87.41 kB ( |
| d_uint16x2.ts | 87.41 kB ( |
| d_uint16x4.ts | 87.41 kB ( |
| d_uint32x2.ts | 87.41 kB ( |
| d_uint32x3.ts | 87.41 kB ( |
| d_uint32x4.ts | 87.41 kB ( |
| d_unorm8x2.ts | 87.41 kB ( |
| d_unorm8x4.ts | 87.41 kB ( |
| d_float16.ts | 87.41 kB ( |
| d_float32.ts | 87.41 kB ( |
| d_sint8x2.ts | 87.41 kB ( |
| d_sint8x4.ts | 87.41 kB ( |
| d_snorm16.ts | 87.41 kB ( |
| d_uint8x2.ts | 87.41 kB ( |
| d_uint8x4.ts | 87.41 kB ( |
| d_unorm16.ts | 87.41 kB ( |
| d_sint16.ts | 87.41 kB ( |
| d_sint32.ts | 87.41 kB ( |
| d_snorm8.ts | 87.41 kB ( |
| d_uint16.ts | 87.41 kB ( |
| d_uint32.ts | 87.41 kB ( |
| d_unorm8.ts | 87.41 kB ( |
| d_sint8.ts | 87.41 kB ( |
| d_packedFormats.ts | 87.41 kB ( |
| d_isPackedData.ts | 87.41 kB ( |
| std_textureDimensions.ts | 103.15 kB ( |
| std_textureGather.ts | 103.14 kB ( |
| std_textureSampleBaseClampToEdge.ts | 103.16 kB ( |
| std_textureSampleCompareLevel.ts | 103.15 kB ( |
| std_textureSampleCompare.ts | 103.15 kB ( |
| std_textureSampleLevel.ts | 103.15 kB ( |
| std_textureSampleBias.ts | 103.15 kB ( |
| std_textureSampleGrad.ts | 103.15 kB ( |
| std_textureSample.ts | 103.14 kB ( |
| std_textureStore.ts | 103.14 kB ( |
| std_textureLoad.ts | 103.14 kB ( |
| d_alignmentOf.ts | 87.41 kB ( |
| std_translation4.ts | 103.14 kB ( |
| std_rotationX4.ts | 103.14 kB ( |
| std_rotationY4.ts | 103.14 kB ( |
| std_rotationZ4.ts | 103.14 kB ( |
| std_identity2.ts | 103.14 kB ( |
| std_identity3.ts | 103.14 kB ( |
| std_identity4.ts | 103.14 kB ( |
| std_scaling4.ts | 103.14 kB ( |
| d_isBuiltin.ts | 87.41 kB ( |
| d_sizeOf.ts | 87.41 kB ( |
| d_isContiguous.ts | 87.41 kB ( |
| d_getLongestContiguousPrefix.ts | 87.43 kB ( |
| d_arrayOf.ts | 87.41 kB ( |
| d_size.ts | 87.40 kB ( |
| d_align.ts | 87.41 kB ( |
| d_location.ts | 87.41 kB ( |
| d_interpolate.ts | 87.41 kB ( |
| d_invariant.ts | 87.41 kB ( |
| d_mat4x4f.ts | 87.41 kB ( |
| d_mat2x2f.ts | 87.41 kB ( |
| d_mat3x3f.ts | 87.41 kB ( |
| d_matToArray.ts | 87.41 kB ( |
| d_builtin.ts | 87.41 kB ( |
| std_unpack2x16float.ts | 103.14 kB ( |
| std_unpack4x8unorm.ts | 103.14 kB ( |
| std_pack2x16float.ts | 103.14 kB ( |
| std_pack4x8unorm.ts | 103.14 kB ( |
| std_bitcastU32toF32.ts | 103.14 kB ( |
| std_bitcastU32toI32.ts | 103.14 kB ( |
| std_neg.ts | 103.13 kB ( |
| std_bitShiftRight.ts | 103.14 kB ( |
| std_bitShiftLeft.ts | 103.14 kB ( |
| std_add.ts | 103.13 kB ( |
| std_div.ts | 103.13 kB ( |
| std_mod.ts | 103.13 kB ( |
| std_mul.ts | 103.13 kB ( |
| std_sub.ts | 103.13 kB ( |
| std_translate4.ts | 103.14 kB ( |
| std_rotateX4.ts | 103.14 kB ( |
| std_scale4.ts | 103.14 kB ( |
| std_rotateY4.ts | 103.14 kB ( |
| std_rotateZ4.ts | 103.14 kB ( |
| d_memoryLayoutOf.ts | 87.41 kB ( |
| std_and.ts | 103.13 kB ( |
| std_or.ts | 103.13 kB ( |
| std_select.ts | 103.14 kB ( |
| std_allEq.ts | 103.13 kB ( |
| std_all.ts | 103.13 kB ( |
| std_not.ts | 103.13 kB ( |
| std_eq.ts | 103.13 kB ( |
| std_lt.ts | 103.13 kB ( |
| std_ne.ts | 103.13 kB ( |
| std_isCloseTo.ts | 103.14 kB ( |
| std_any.ts | 103.13 kB ( |
| std_ge.ts | 103.13 kB ( |
| std_gt.ts | 103.13 kB ( |
| std_le.ts | 103.13 kB ( |
| std_dot4U8Packed.ts | 103.14 kB ( |
| std_length.ts | 103.14 kB ( |
| std_countTrailingZeros.ts | 103.15 kB ( |
| std_distance.ts | 103.14 kB ( |
| std_dot.ts | 103.13 kB ( |
| std_countLeadingZeros.ts | 103.15 kB ( |
| std_firstTrailingBit.ts | 103.15 kB ( |
| std_firstLeadingBit.ts | 103.14 kB ( |
| std_fract.ts | 103.13 kB ( |
| std_frexp.ts | 103.13 kB ( |
| std_modf.ts | 103.13 kB ( |
| std_quantizeToF16.ts | 103.14 kB ( |
| std_mix.ts | 103.13 kB ( |
| std_countOneBits.ts | 103.14 kB ( |
| std_dot4I8Packed.ts | 103.14 kB ( |
| std_determinant.ts | 103.14 kB ( |
| std_extractBits.ts | 103.14 kB ( |
| std_faceForward.ts | 103.14 kB ( |
| std_inverseSqrt.ts | 103.14 kB ( |
| std_reverseBits.ts | 103.14 kB ( |
| std_insertBits.ts | 103.14 kB ( |
| std_smoothstep.ts | 103.14 kB ( |
| std_normalize.ts | 103.14 kB ( |
| std_transpose.ts | 103.14 kB ( |
| std_saturate.ts | 103.14 kB ( |
| std_degrees.ts | 103.14 kB ( |
| std_radians.ts | 103.14 kB ( |
| std_reflect.ts | 103.14 kB ( |
| std_refract.ts | 103.14 kB ( |
| std_acosh.ts | 103.13 kB ( |
| std_asinh.ts | 103.13 kB ( |
| std_atan2.ts | 103.13 kB ( |
| std_atanh.ts | 103.13 kB ( |
| std_clamp.ts | 103.13 kB ( |
| std_cross.ts | 103.13 kB ( |
| std_floor.ts | 103.13 kB ( |
| std_ldexp.ts | 103.13 kB ( |
| std_round.ts | 103.13 kB ( |
| std_trunc.ts | 103.13 kB ( |
| std_acos.ts | 103.13 kB ( |
| std_asin.ts | 103.13 kB ( |
| std_atan.ts | 103.13 kB ( |
| std_ceil.ts | 103.13 kB ( |
| std_cosh.ts | 103.13 kB ( |
| std_exp2.ts | 103.13 kB ( |
| std_log2.ts | 103.13 kB ( |
| std_sign.ts | 103.13 kB ( |
| std_sinh.ts | 103.13 kB ( |
| std_sqrt.ts | 103.13 kB ( |
| std_step.ts | 103.13 kB ( |
| std_tanh.ts | 103.13 kB ( |
| std_abs.ts | 103.13 kB ( |
| std_cos.ts | 103.13 kB ( |
| std_exp.ts | 103.13 kB ( |
| std_fma.ts | 103.13 kB ( |
| std_log.ts | 103.13 kB ( |
| std_max.ts | 103.13 kB ( |
| std_min.ts | 103.13 kB ( |
| std_pow.ts | 103.13 kB ( |
| std_sin.ts | 103.13 kB ( |
| std_tan.ts | 103.13 kB ( |
| common_fullScreenTriangle.ts | 57.14 kB ( |
| common_writeSoA.ts | 57.13 kB ( |
| STATIC_allImports.ts | 287.16 kB ( |
| STATIC_d.ts | 87.40 kB ( |
| STATIC_std.ts | 103.13 kB ( |
If you wish to run a comparison for other, slower bundlers, run the 'Tree-shake test' from the GitHub Actions menu.
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 [0.86, 1.66, 3.51, 5.06, 6.37, 10.32, 19.08, 21.75]
line [0.83, 1.65, 3.60, 5.77, 6.42, 8.75, 18.06, 20.54]
line [0.80, 1.63, 3.54, 5.41, 6.35, 10.14, 17.95, 22.20]
---
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 [0.31, 0.50, 0.66, 0.78, 1.00, 1.02, 1.22, 1.41]
line [0.32, 0.48, 0.68, 0.74, 1.00, 1.03, 1.19, 1.32]
line [0.30, 0.51, 0.59, 0.72, 0.96, 1.00, 1.26, 1.37]
---
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 [0.71, 1.79, 3.32, 5.39, 10.67, 22.65, 46.46, 96.52]
line [0.80, 1.87, 3.63, 5.53, 10.72, 22.53, 47.59, 95.35]
line [0.77, 1.80, 3.42, 5.53, 10.69, 23.73, 47.29, 96.56]
|
|
@pullfrog review |
There was a problem hiding this comment.
Important
The new in externalMap check on packages/typegpu/src/core/resolve/externals.ts:98 walks the prototype chain, so WGSL identifiers like toString, constructor, or valueOf are incorrectly treated as missing externals and trigger spurious warnings. See the inline comment for the one-line fix.
Reviewed changes — a rewrite of replaceExternalsInWgsl that resolves external property chains in a single regex pass instead of recursing per external key.
- Single-pass external resolution —
replaceExternalsInWgslnow finds all bounded identifier chains, walks them until a resolvable value is found, and caches resolved values. - Removed unused-external warning — the call to warn when an external key isn't used has been removed.
- Test and snapshot updates —
declare.test.ts,resolve.test.ts, and several console-log snapshots were updated for the new resolution ordering.
openrouter/moonshotai/kimi-k2.7-code (free via Pullfrog for OSS) | 𝕏
Co-authored-by: pullfrog[bot] <226033991+pullfrog[bot]@users.noreply.github.com>
There was a problem hiding this comment.
ℹ️ One behavioral change worth confirming — details inline.
Reviewed changes — rewrote raw-WGSL external resolution to traverse identifier chains in a single scan and resolve at the first resolvable link.
- Rewrite
replaceExternalsInWgsl— uses one regex scan over WGSL, walks each identifier chain through the external map, and resolves at the first resolvable value or warns when the chain ends without one. - Remove the unused-external warning — the old
The external 'X' wasn't used...console warning is dropped because externals may be applied broadly across functions. - Emit resolved WGSL in source order — inline snapshots updated to reflect declarations appearing in the order they are referenced rather than in
$usesinsertion order. - Update
tgpu.declaretests — declarations are now explicitly referenced in templates; unreferenced declarations are no longer injected.
Kimi K2 (free via Pullfrog for OSS) | 𝕏
…l-for-raw-wgsl-implemented-functions

Before: given an external, if resolvable then replace occurrences with resolved, otherwise look which keys appear in the code with a prop following them, and call recursively with
external.prop.Now: given externals, look for prop chains starting with external keys, then follow them until either a resolvable appears, or the chain ends.
Other changes: