Repository navigation
fix: report the deck.gl render backend in the GPU panel (#8) - #20
Merged
Merged
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
…nd (#8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
RENDER now reads the luma.gl device type passed to deck's onDeviceInitialized instead of the WebGPU adapter tier, and the compute readout key becomes WEBGPU COMPUTE. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
#8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
…nknown (#8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
…opup (#8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
…ean (#8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf
Closed
4 tasks
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.
Summary
The GPU panel said
RENDER WebGPUwhenevernavigator.gpu.requestAdapter()succeeded, while deck.gl was actually drawing with WebGL2. This PR reports the truth (user decision; WebGPU rendering is not wired):onDeviceInitializedcallback passed in theMapboxOverlayconstructor props (src/components/MapView.jsx).webgl(andwebgl2) showsWebGL2,webgpushowsWebGPU,Detecting...until deck has a device,Unknownfor anything else. The WebGPU adapter tier no longer decides RENDER; it only gates the compute readout.WEBGPU COMPUTE, so WebGPU compute availability (--or the RUN button from GPU compute benchmark runs a 64 MB dispatch every second indefinitely, including in hidden tabs #11) reads apart from the render backend.device, thendescription, thenvendor architecturejoined, then deck's WebGL renderer string (device.info.renderer), thenUnknown. Long names are truncated with an ellipsis at 32ch with the full name intitle(the span opts intopointer-eventsso the tooltip shows; the panel stayspointer-events: none).max-widthis nowcalc(100vw - 340px)so it always clears the 288px desktop popup, and the|separators are hidden at every width with a 14px column gap (as Layout breaks on mobile: header text overlaps and panels overflow at 375px #5 already did at 1024px and below), so a wrapped row never starts or ends with one. The 1024px and narrow rules from Layout breaks on mobile: header text overlaps and panels overflow at 375px #5 are unchanged.src/components/gpuLabels.js(backend labels in aMap, so prototype names like__proto__showUnknown).Plan
docs/plans/2026-11-07-issue-8-gpu-panel-render-backend.md, committed in b33b130 before any code. Revised in the same commits as the review fixes that departed from it: 3838b74 (AC1, prototype names), c1f9551 (AC5 truncation and title, AC7 popup clearance above 1024px), 212cde3 (AC5 tooltip pointer-events), 886f4cf (AC7 separators and column gap at every width), ff04f8e (README scope).Test evidence
Gate at 886f4cf, run verbatim in the worktree (README-only commit ff04f8e after):
npm ci: added 330 packages, found 0 vulnerabilitiesnpm run lint: cleannpm test: Test Files 30 passed (30), Tests 249 passed (249) (baseline on main: 24 files, 211 tests)npm run build: built (pre-existing >500 kB chunk warning)npm audit --omit=dev --audit-level=high: found 0 vulnerabilitiesNew and updated tests:
gpuLabels.test.jsandgpuLabels.protoKeys.test.js(RENDER forwebgl,webgl2,webgpu, not initialized, unknown and prototype names; every GPU label fallback step),GPUInfoPanel.test.jsx(RENDER cases incl. tierwebgpuwith devicewebglshowingWebGL2, key rename),GPUInfoPanel.benchmarkTier.test.jsx(key rename),GPUInfoPanel.gpuName.test.jsx,MapView.test.jsx(callback in theMapboxOverlayconstructor props;Detecting...thenWebGL2),styles.test.js,styles.gpuPanel.test.js,styles.gpuSeparators.test.js. Each defect's test was committed red before its fix.Live drive (
npm run build+npm run preview, Playwright with headless Chromium; the chrome-devtools MCP profile was locked):Detecting...(warn) at about 190 ms, thenWebGL2(ok) at about 450 to 620 ms.maplibregl-canvas,#deckgl-overlay) hold awebgl2context and nowebgpucontext; the deck device read from the panel props is{ type: 'webgl', renderer: 'ANGLE (Google, Vulkan 1.3.0 (SwiftShader Device (Subzero) ...' }. RENDER matches.requestAdapter()is null): WEBGPU COMPUTE--, no RUN, GPU is the ANGLE renderer string, truncated with fulltitle.--enable-unsafe-webgpu): tierwebgpu, GPUgoogle swiftshader, RUN shown and a run gave~3.5 GB/s est., RENDER stillWebGL2.Review findings
High tier: three Claude finders (bugs/compliance, security, outside-the-diff) and two Fable finders (bugs/compliance, edge cases), a confirming finder, two delta rounds, a fresh verifier, and two doc finders.
Fixed:
renderBackendLabelused a plain object, so__proto__crashed React andtoStringshowed ok. Fixed with aMap(33fe11e, 3838b74).titletooltip was unreachable because the panel ispointer-events: none. Fixed (264840a, 212cde3).|. Fixed (8e04e05, 886f4cf).Dismissed and why:
Left as notes (not changed here):
Detecting...; after a WebGL context loss it still showsWebGL2. Loading and error states belong to No loading, error, or degraded states: failed data loads leave an empty map and SYS NOMINAL #10; noted in README Known limitations.topLevelCss()test helper ignores@supportsand is copied in three test files; the MapView comment says the callback "must" be in the constructor (it is the non-racy choice).Docs reconciled
WEBGPU COMPUTE, the GPU name fallback and truncation, and the device-failure and context-loss limits.Closes #8
🤖 Generated with Claude Code
https://claude.ai/code/session_01AELE7jB1i3uqqrHu3SjJjf