Repository navigation
Expand file tree
/
Copy pathindex.html
More file actions
175 lines (168 loc) · 9.01 KB
/
Copy pathindex.html
File metadata and controls
175 lines (168 loc) · 9.01 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<title>Completion Receipt — WebGPU Host Observation</title>
<link rel="stylesheet" href="styles.css" />
</head>
<body>
<main class="card">
<header class="header">
<div class="header-title">
<p class="eyebrow">COMPLETION RECEIPT</p>
<h1 class="headline">GPU finished. Firefox waited.</h1>
<p class="subline">WEBGPU / HOST OBSERVATION</p>
</div>
<dl class="meta">
<div class="meta-row"><dt>host</dt><dd>Linux · Intel Iris Xe</dd></div>
<div class="meta-row"><dt>chrome</dt><dd>Chrome 152.0.7977.75</dd></div>
<div class="meta-row"><dt>firefox</dt><dd>Firefox 155.0.1 (WebGPU manually enabled)</dd></div>
<div class="meta-row"><dt>runs</dt><dd><span data-metric="samples"></span> runs after <span data-metric="warmups"></span> warmups</dd></div>
<div class="meta-row"><dt>workload</dt><dd>65536 items · 64 loops</dd></div>
</dl>
</header>
<section class="cta-panel" aria-label="Run the receipt on this device">
<div class="cta-action">
<p class="cta-status" data-cta-status>MANUAL · NOTHING RUNS UNTIL YOU PRESS RUN CAPTURE</p>
<button id="cta-run-trigger" class="cta-button" type="button" aria-controls="repro-dialog" aria-haspopup="dialog">RUN ON THIS DEVICE</button>
</div>
<p class="cta-copy">Same WebGPU workload, local only. Nothing is measured until you press <strong>Run capture</strong> inside the panel; nothing is uploaded.</p>
<div class="cta-chips" aria-label="Local-run guarantees">
<span class="chip">NO TELEMETRY</span>
<span class="chip">NO AUTO-RUN</span>
<span class="chip">CLOSE PANEL TO STOP</span>
</div>
</section>
<section class="row row-primary">
<article class="tile tile-idle" data-run="firefox155-idle">
<p class="tile-label">IDLE POLL</p>
<p class="tile-hero" data-metric="host-median">— ms</p>
<p class="tile-caption">Firefox 155.0.1 host completion, idle</p>
<div class="split">
<div class="split-cell">
<p class="split-label">GPU complete</p>
<p class="split-value" data-metric="gpu-median">— ms</p>
</div>
<div class="split-cell">
<p class="split-label">host notified</p>
<p class="split-value split-value--emph" data-metric="host-median-inline">— ms</p>
</div>
</div>
<p class="tile-note">GPU work finishes; the completion report does not. Consistent with Firefox Bug 1870699 (WebGPU polled on a timer).</p>
</article>
<article class="tile tile-timeline">
<p class="tile-label">COMPLETION TIMELINE</p>
<p class="tile-caption">Shared 0–100 ms scale · submit → GPU complete → host notified</p>
<div class="axis">
<span>0</span><span>20</span><span>40</span><span>60</span><span>80</span><span>100 ms</span>
</div>
<ol class="timeline" data-timeline>
<li class="lane" data-run="chrome152">
<span class="lane-label">Chrome 152</span>
<div class="lane-track">
<div class="lane-bar" data-bar></div>
<span class="lane-marker lane-marker--gpu" data-marker="gpu"></span>
<span class="lane-marker lane-marker--host" data-marker="host"></span>
</div>
<span class="lane-value" data-metric="host-median"></span>
</li>
<li class="lane" data-run="firefox155-idle">
<span class="lane-label">Firefox 155 idle</span>
<div class="lane-track">
<div class="lane-bar" data-bar></div>
<span class="lane-marker lane-marker--gpu" data-marker="gpu"></span>
<span class="lane-marker lane-marker--host" data-marker="host"></span>
</div>
<span class="lane-value" data-metric="host-median"></span>
</li>
<li class="lane" data-run="firefox155-kick">
<span class="lane-label">Firefox + fast kicks</span>
<div class="lane-track">
<div class="lane-bar" data-bar></div>
<span class="lane-marker lane-marker--gpu" data-marker="gpu"></span>
<span class="lane-marker lane-marker--host" data-marker="host"></span>
</div>
<span class="lane-value" data-metric="host-median"></span>
</li>
<li class="lane" data-run="firefox155-kick20">
<span class="lane-label">Firefox + 20 ms kick</span>
<div class="lane-track">
<div class="lane-bar" data-bar></div>
<span class="lane-marker lane-marker--gpu" data-marker="gpu"></span>
<span class="lane-marker lane-marker--host" data-marker="host"></span>
</div>
<span class="lane-value" data-metric="host-median"></span>
</li>
</ol>
<p class="tile-note">Solid bar = elapsed host-completion wait. Marker = when the GPU itself actually finished. The gap between them is the reported joint, not GPU slowness.</p>
</article>
<article class="tile tile-prediction" data-run="firefox155-kick20">
<p class="tile-label">PREDICTION RECEIPT</p>
<p class="tile-caption">Diagnostic probe: does forcing observation cadence move the reported floor?</p>
<div class="predict-row">
<span class="predict-key">predicted</span>
<span class="predict-val">20 ms kick cadence → ~20 ms completion</span>
</div>
<div class="predict-row">
<span class="predict-key">observed</span>
<span class="predict-val predict-val--emph" data-metric="host-median-inline">— ms</span>
</div>
<div class="predict-row">
<span class="predict-key">kicks / run</span>
<span class="predict-val" data-metric="kicks-median"></span>
</div>
<p class="tile-note">The kick is a diagnostic probe of the polling mechanism only — not a fix to run in a real app. It does not make the GPU workload faster.</p>
</article>
</section>
<section class="row row-secondary">
<article class="tile tile-distribution">
<p class="tile-label">DISTRIBUTION · <span data-metric="samples"></span> RUNS EACH</p>
<div class="dist-scroll">
<table class="dist-table">
<thead>
<tr><th>run</th><th>gpu median</th><th>host min</th><th>host median</th><th>host max</th></tr>
</thead>
<tbody data-distribution-body></tbody>
</table>
</div>
<p class="tile-note">GPU column shows median only. Kick-mode captures include empty diagnostic submissions in their raw min/mean — those are not workload timings and are not displayed here.</p>
</article>
<article class="tile tile-limits">
<p class="tile-label">LIMITS / PROVENANCE</p>
<ul class="limits-list">
<li>One Linux host, Intel Iris Xe integrated GPU. No cross-platform or cross-vendor claim.</li>
<li>Firefox 155.0.1 WebGPU was manually enabled via flag; not default release support.</li>
<li>Canonical evidence is headless, on the real Iris Xe GPU. Ordinary headed-Firefox parity on that real GPU remains untested (see PROVENANCE.md for a separate, non-canonical software-rendered headed smoke that is not a substitute).</li>
<li>Consistent with Mozilla Bug 1870699 (polling on a timer). No source-level confirmation was performed — this is black-box timing evidence offered to an already-open, already-triaged issue.</li>
<li>Not claimed as a first-discovery finding, a product, a performance benchmark, or evidence of demand. No cross-browser prevalence claim.</li>
</ul>
<div class="actions">
<a class="action" href="fixtures/chrome152-small.json">RAW FIXTURES — Chrome 152</a>
<a class="action" href="fixtures/firefox155-small.json">RAW FIXTURES — Firefox 155 idle</a>
<a class="action" href="fixtures/firefox155-kick-small.json">RAW FIXTURES — Firefox fast kick</a>
<a class="action" href="fixtures/firefox155-kick20-fixed.json">RAW FIXTURES — Firefox 20 ms kick</a>
<a class="action" href="repro/index.html">REPRODUCE — run this workload locally</a>
<a class="action" href="https://bugzilla.mozilla.org/show_bug.cgi?id=1870699" rel="noopener noreferrer">UPSTREAM BUG — Mozilla 1870699</a>
</div>
</article>
</section>
</main>
<dialog id="repro-dialog" aria-labelledby="repro-dialog-title">
<div class="repro-dialog-shell">
<header class="repro-dialog-bar">
<div>
<p class="repro-dialog-kicker">LOCAL REPRODUCER · NOTHING AUTO-RUNS</p>
<h2 id="repro-dialog-title">Run the same workload on this device</h2>
</div>
<div class="repro-dialog-actions">
<a class="repro-direct" href="repro/index.html" target="_blank" rel="noopener noreferrer">OPEN DIRECT</a>
<button id="repro-close" class="repro-close" type="button">CLOSE</button>
</div>
</header>
<iframe id="repro-frame" class="repro-frame" data-src="repro/index.html" title="WebGPU completion reproducer"></iframe>
</div>
</dialog>
<script type="module" src="app.mjs"></script>
</body>
</html>