[RFC] chore: Don't name externals in mergeExternals#2616
Merged
Conversation
|
pkg.pr.new packages benchmark commit |
📊 Bundle Size Comparison
👀 Notable resultsStatic test results:No major changes. Dynamic test results:No major changes. 📋 All resultsClick to reveal the results table (355 entries).
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 [0.90, 1.87, 4.10, 6.66, 8.00, 11.11, 21.55, 25.33]
line [0.90, 1.78, 4.10, 5.96, 7.23, 11.76, 20.98, 24.46]
line [0.93, 1.76, 4.24, 6.51, 7.50, 11.65, 21.56, 23.71]
---
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 [0.33, 0.74, 0.72, 0.84, 1.15, 1.19, 1.45, 1.52]
line [0.35, 0.48, 0.66, 0.85, 1.11, 1.14, 1.58, 1.57]
line [0.28, 0.58, 0.76, 0.87, 1.18, 1.26, 1.51, 1.55]
---
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 [0.78, 2.12, 4.34, 7.27, 12.81, 26.99, 55.93, 118.22]
line [0.77, 2.06, 3.71, 6.16, 12.08, 24.74, 53.10, 108.88]
line [0.98, 2.12, 3.88, 6.16, 12.17, 25.49, 55.44, 110.19]
|
aleksanderkatan
commented
Jun 16, 2026
Comment on lines
+571
to
+577
| it("resolved externals's names stay the same", () => { | ||
| const c1 = (() => tgpu.const(d.vec2u, d.vec2u(1)))(); // unnamed | ||
| tgpu.resolve([tgpu.fn([])`() { let a = myConst; }`.$uses({ myConst: c1 })]); | ||
| tgpu.resolve([tgpu.fn([])`() { let a = otherName; }`.$uses({ otherName: c1 })]); | ||
|
|
||
| expect(getName(c1)).toMatchInlineSnapshot(`"myConst"`); | ||
| }); |
Contributor
Author
There was a problem hiding this comment.
This makes me wonder if we should do anything at all, though I don't think it's that much of an issue
Collaborator
There was a problem hiding this comment.
I think it's fine, I think it's unusual for a thing to be references by other names in different WGSL templates, and even if it is, it's still a valid name for it.
iwoplaza
reviewed
Jun 16, 2026
Collaborator
There was a problem hiding this comment.
This was a very useful case for ambient naming (through external keys), as time is the result of calling useUniform, a helper API that's not instrumented by unplugin-typegpu to be auto-named.
What we could do instead, is add a separate mechanism of ambient naming to 'use gpu' shader functions. When an external is retrieved using getById, and it is namable, we can name it. Same whenever we access a prop, we can name the value using the key if its namable. Wdyt?
Contributor
Author
There was a problem hiding this comment.
Great idea! I thought we could just add more cases to plugin autonaming, but this seems to do the job
![pullfrog[bot]](https://avatars.githubusercontent.com/in/1768019?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
ℹ️ Minor suggestion inline — otherwise looks solid.
Reviewed changes — removes the auto-naming logic from mergeExternals and shifts responsibility to replaceExternalsInWgsl for raw WGSL and the plugin for TGSL, with a follow-up fix for the guarded compute pipeline.
- Remove
setNamefrommergeExternals— the function no longer names unnamed externals by map key, avoiding depth-dependent naming bugs and the perf cost ofisResolvablein that path. - Add
setNameinreplaceExternalsInWgsl— for raw WGSL functions, unnamed resolvable externals are now named from the last component of their external key (e.g.ext.n2→n2) before resolution. - Explicit
$nameincreateGuardedComputePipeline— thewrappedCallbackwas previously named bymergeExternals; added an explicit$name('wrappedCallback')call. - Snapshot updates — doc example tests now show
item(themakeUniqueIdentifierdefault) instead oftimefor unnamed externals.
Note: 1 inline comment(s) dropped because they did not anchor to lines inside the PR diff:
packages/typegpu/src/core/resolve/externals.ts:18(RIGHT) — line 18 (RIGHT) is not inside a diff hunk
DeepSeek Pro (free via Pullfrog for OSS) ｜ 𝕏
Contributor
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes — adds TGSL-side externals naming in getById and accessProp, completing externals naming coverage that was missing from the prior review (which only covered raw WGSL in replaceExternalsInWgsl).
- Name externals in
getById— whenItemStateStackImpl.getSnippetByIdfetches a namelessTgpuNamableexternal fromlayer.externalMap[id], it now names it with that ID (e.g.,myConst). - Name externals in
accessProp— when accessing a property on a comptime-known target that yields a namelessTgpuNamablevalue, it is named with the property name. - New tests for naming coverage — added
"names used externals"and"names used nested externals"totgslFn.test.ts, plus"names resolved externals"/"resolved externals's names stay the same"torawFn.test.ts. - Removed doc snapshot churn — the
time→itemsnapshot updates in doc example tests were reverted, since unnamed externals now carry proper names.
DeepSeek Pro (free via Pullfrog for OSS) ｜ 𝕏
iwoplaza
approved these changes
Jun 18, 2026
iwoplaza
left a comment
iwoplaza
left a comment
Collaborator
There was a problem hiding this comment.
I like these changes! 💜
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Currently, whether an unnamed item is autonamed in
mergeExternalsdepends on its depth in plugin-generated externals:As seen in #2593, an actual deep pass for proper autonaming is not only difficult, but also hits the performance due to how long
isResolvabletakes.I think removing this may be a good step in order to make our externals both correct and performant.