Skip to content

main-route-selected-hour-current-head.md

Latest commit

 

History

History
56 lines (34 loc) · 4.01 KB

main-route-selected-hour-current-head.md

File metadata and controls

56 lines (34 loc) · 4.01 KB
Raw

Main Route Selected-Hour Current-HEAD Baseline

Date: 2026-05-15

Scope

This artifact measures the main route /, not /debug. It captures the current selected-hour WebGPU path without .bin, Python, or debug comparison data in the timing baseline.

Included Analyses

  • Ben-Gurion/20250815_grid_2m_fullday
  • Ness-Tziona/exploded/nes_tziona_unblock_2

Excluded BG Variants

This baseline intentionally excludes other Ben-Gurion variants so the current-head timing pass stays limited to the BG 2m base case and Ness Tziona 2m.

Proof Boundary

  • Ben-Gurion/20250815_grid_2m_fullday: utciSurfaceSource=compute-buffer-selected-hour, baseRenderTransport=compute-buffer-selected-hour, dataTextureBuildCount=0, selectedHourRuntimeContract.strongVisibleGpuPath=true, no python/bin/debug comparison fields, no forbidden comparison requests.
  • Ness-Tziona/exploded/nes_tziona_unblock_2: utciSurfaceSource=compute-buffer-selected-hour, baseRenderTransport=compute-buffer-selected-hour, dataTextureBuildCount=0, selectedHourRuntimeContract.strongVisibleGpuPath=true, no python/bin/debug comparison fields, no forbidden comparison requests.

Memory is scoped only to tracked app-owned UTCI/WebGPU buffers. The GPU VRAM total mirrors the main-route runtime helper: persistentExposureBytes + allHoursOutputBytes + selectedHourOutputBytes + renderOwnedSelectedHourBytes. Selected-hour high-watermark is reported as diagnostic context, but the displayed total is not a browser, OS, or device VRAM measurement.

Unavailable Timing Fields

The fresh / capture currently exposes only the coarser main-route timing fields shown below. The JSON artifact preserves these fields as null for both cases instead of inventing values:

  • payloadPrepareMs
  • workerBvhMs
  • pipelineUploadMs
  • firstSelectedHourReadyMs

This means the main-route baseline is suitable for current route-level evidence around first-visible timing, exposure precompute, scene-sync delay, scene-sync total, selected-hour dispatch, and tracked app-owned GPU memory. It is not a full fine-grained cold-start sub-bucket breakdown by itself.

Timing Table

Project Analysis Points Grid m Month Hour Time index First visible ms Exposure precompute ms Scene sync start delay ms Scene sync total ms Dispatch ms
Ben-Gurion Ben-Gurion/20250815_grid_2m_fullday 104,445 2.0 7 0 168 1283.8 708.3 269.0 80.9 3.8
Ness-Tziona Ness-Tziona/exploded/nes_tziona_unblock_2 511,840 2.0 7 0 168 8465.4 6648.2 1140.1 340.1 5.1

Memory Table

Project Analysis GPU VRAM MiB Persistent exposure MiB Selected-hour current MiB Selected-hour HWM MiB Render-owned selected-hour MiB Scope
Ben-Gurion Ben-Gurion/20250815_grid_2m_fullday 14.34 3.98 0.40 0.40 9.96 utci-owned-webgpu-buffers
Ness-Tziona Ness-Tziona/exploded/nes_tziona_unblock_2 70.29 19.53 1.95 1.95 48.81 utci-owned-webgpu-buffers

Current Optimization Inference

Fresh main-route numbers point at exposure precompute / cold-start compute as the next bottleneck: it averages 3678.3 ms across BG and Ness Tziona, while selected-hour dispatch stays at 4.5 ms. That means the next optimization pass should stay focused on cold-start work before first visible publication, not on .bin comparison, selected-hour transport, or 0.5m claims yet.

The optimization inference above is intentionally conservative: it is based on the available main-route fields plus older debug-route context where finer-grained sub-bucket detail exists.

If these numbers disagree with the older 2026-05-09 strategy snapshot, this fresh main-route baseline wins for current route-level timing decisions, while the older debug-route breakdown remains supporting historical context for unavailable sub-buckets such as payloadPrepareMs, workerBvhMs, pipelineUploadMs, firstSelectedHourReadyMs.