dreamwell-matter 1.0.0

DreamMatter benchmark — GPU physics materialization demo and profiler
Documentation
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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
# DREAMWELL 3D Physics Benchmark


**Path:** `docs/architecture/DREAMWELL_3D_PHYSICS_BENCHMARK.md`

| Field               | Value                                                                             |
| ------------------- | --------------------------------------------------------------------------------- |
| Title               | Dreamwell 3D Physics Benchmark — DreamMatter Timelapse                            |
| Version             | 1.0.0                                                                             |
| Status              | Internal Benchmark Specification                                                  |
| Audience            | Engine, rendering, runtime, benchmark, QA, and release engineers                  |
| Purpose             | Validate DreamMatter dynamic-to-static materialization under a bounded GPU budget |
| Duration            | 30 seconds                                                                        |
| Output              | Real-time benchmark run + KPI log + optional capture/export artifact              |
| Default Resolution  | 1920 × 1080                                                                       |
| Default VRAM Budget | 2 GiB total benchmark budget                                                      |
| Scene Type          | 3D                                                                                |
| Assets Required     | None                                                                              |
| Determinism         | Authority-side deterministic; GPU presentation benchmark is non-deterministic     |

---

## 1. Purpose


This document defines the first canonical visual and performance benchmark for Dreamwell’s 3D DreamMatter pipeline.

The benchmark demonstrates one complete DreamMatter lifecycle in a controlled scene:

* a bounded white Dreamlet swarm begins on the **left** side of a black room
* the swarm initially occupies a **sphere-shaped dynamic state**
* the swarm travels smoothly toward the **right** side of the scene
* the swarm converges into a **cube target**
* the swarm settles and materializes into a stable render representation
* the materialized form remains visible as a static/baked-style object
* the object can optionally dissolve back into DreamMatter at the end of the run

This benchmark is not a content showcase. It is a technical benchmark designed to validate:

* DreamMatter runtime states
* Dreamlet-to-DreamFabric transitions
* Quantum Culling
* PBR scene-shell rendering
* shadow and post-processing interaction
* observer-scoped rendering behavior
* GPU-only simulation and render preparation
* bounded VRAM operation
* stable frame pacing and performance metrics

---

## 2. Benchmark goals


The benchmark must prove the following in one 30-second run:

1. DreamMatter can render a high-density dynamic swarm without CPU particle updates
2. Dreamlets can maintain a coherent sphere-like swarm state
3. Dreamlets can translate across the scene under GPU simulation
4. Dreamlets can converge to a new target topology
5. Dreamlets can enter settled and materialized states
6. The materialized object can render as a stable PBR scene object
7. The pipeline remains inside a fixed VRAM budget, default 2 GiB
8. The frame pipeline remains stable and measurable throughout all runtime states
9. All major KPIs are emitted in a structured benchmark log
10. The scene can run with no external assets, textures, or models

---

## 3. Benchmark summary


### 3.1 Visual description


The benchmark scene is a black enclosed room with:

* black floor
* black walls
* black ceiling
* one white Dreamlet swarm sphere on the left
* one white cube target on the right
* one directional sun light
* one basic PBR-compatible room material setup
* one post-process stack
* one global illumination baseline

The Dreamlet swarm is rendered as white matter-like particles or micro-geometry units. The room is dark, uncluttered, and asset-free to ensure that the benchmark is focused on DreamMatter behavior and pipeline cost.

### 3.2 Scene intent


This benchmark is intentionally minimal.

It exists to isolate:

* simulation cost
* culling cost
* assembly cost
* materialization cost
* static representation cost
* total frame cost across state transitions

---

## 4. Benchmark scene specification


### 4.1 World layout


A simple room centered at origin.

Recommended dimensions:

* room width: 20.0 meters
* room depth: 12.0 meters
* room height: 8.0 meters

Recommended coordinates:

* Dreamlet swarm origin center: `(-5.0, 1.5, 0.0)`
* target cube center: `(5.0, 1.5, 0.0)`
* camera origin: `(0.0, 3.0, 11.0)` looking toward `(0.0, 2.0, 0.0)`

### 4.2 Geometry


Only primitive geometry is allowed.

Required geometry:

* floor plane
* back wall
* left wall
* right wall
* ceiling plane
* target cube
* optional hidden room front cutout for camera framing

No imported models.
No textures.
No authored assets.
No animation clips.

### 4.3 Materials


All materials use engine-provided defaults only.

Room material:

* base color: near-black
* roughness: high
* metalness: low
* emissive: none

Dreamlet dynamic material:

* base color: white
* roughness: low-to-mid
* metalness: 0 unless explicitly testing metallic mode
* emissive: optional low white emissive, disabled by default

Target/materialized cube material:

* base color: white
* roughness: mid
* metalness: 0 by default

### 4.4 Lighting


Minimum required lighting:

* one directional “sun” light
* one ambient / IBL / flat GI baseline
* one post-process stack

Recommended default:

* directional light yaw: 35°
* directional light pitch: -45°
* white light color
* stable shadow map enabled
* IBL fallback allowed if no HDR environment is loaded

### 4.5 Post-processing


Required:

* HDR render target
* bloom enabled at conservative settings
* tonemap enabled

Optional:

* vignette disabled
* chromatic aberration disabled
* film grain disabled

The benchmark should measure DreamMatter, not stylization noise.

---

## 5. Benchmark runtime states


The benchmark explicitly covers all five DreamMatter runtime states.

### 5.1 Swarm state


The Dreamlets behave like a loose sphere-shaped cloud.

Visual target:

* a stable sphere silhouette with local motion
* visible internal turbulence or low-amplitude motion
* no convergence toward cube yet

### 5.2 Convergence state


The Dreamlets begin moving to the right and gradually align to cube target topology.

Visual target:

* smooth lateral translation
* narrowing swarm spread
* increasing shape coherence
* visibly changing from sphere distribution to cube-compatible distribution

### 5.3 Settled state


The Dreamlets stop behaving like a loose cloud and occupy a near-complete cube volume.

Visual target:

* low-motion clustered matter
* cube silhouette mostly complete
* small residual motion allowed

### 5.4 Materialized state


The cube becomes a stable render representation.

Visual target:

* stable white cube
* rendered as ordinary scene shell geometry or stable DreamFabric proxy
* no visible particle turbulence except where explicitly testing hybrid mode

### 5.5 Dissolution state


Optional final phase.

Visual target:

* cube breaks apart back into DreamMatter
* Dreamlets return to loose state
* benchmark ends after confirming reverse transition

---

## 6. 30-second timelapse specification


The benchmark run is fixed at 30 seconds.

### 6.1 Timeline


| Time Window   | State                                        | Description                                             |
| ------------- | -------------------------------------------- | ------------------------------------------------------- |
| 0.0s – 4.0s   | Initialization / steady swarm                | Warm-up, stabilize scene, display sphere swarm          |
| 4.0s – 10.0s  | Swarm state                                  | Loose sphere maintained on left side                    |
| 10.0s – 18.0s | Convergence state                            | Swarm travels right and converges toward cube           |
| 18.0s – 22.0s | Settled state                                | Cube silhouette completes, low-motion residual settling |
| 22.0s – 27.0s | Materialized state                           | Stable static/baked-style cube render representation    |
| 27.0s – 30.0s | Dissolution state (optional but recommended) | Cube dissolves back to swarm                            |

### 6.2 Warm-up requirement


The benchmark must include an initial warm-up phase.

Warm-up purpose:

* stabilize shader pipelines
* stabilize GPU clocks where possible
* fill persistent caches
* avoid including one-time init transients in steady-state metrics

Warm-up capture policy:

* collect warm-up metrics separately
* do not merge warm-up into steady-state averages unless explicitly labeled

---

## 7. Camera and observer setup


### 7.1 Camera policy


The benchmark camera is fixed.

Required:

* no free-fly movement during canonical benchmark run
* no shake
* no motion blur
* no user input influence

Recommended framing:

* both sphere origin and target cube visible simultaneously
* room boundaries visible enough to validate lighting/shadows
* camera positioned to show lateral travel clearly

### 7.2 Observer context


The benchmark must still run through full observer-scoped logic.

Required observer values:

* observer position derived from camera or benchmark observer anchor
* observer layer fixed to active scene layer
* observer radius sufficient to contain the benchmark scene
* Quantum Culling still active and measurable

Why:

* this validates the real production path rather than a bypass path

---

## 8. Benchmark technical configuration


### 8.1 Required runtime configuration


* resolution: `1920x1080`
* fullscreen: disabled by default, windowed benchmark mode preferred
* VSync: off by default for raw performance measurement; optional secondary run with VSync on
* frame pacing capture: enabled
* debug overlays: off for canonical run
* profiler markers: on
* post-process: on
* PBR shell: on
* shadows: on
* IBL/GI baseline: on
* Quantum Culling: on
* DreamMatter: on
* mesh shader path: on if supported, else vertex fallback

### 8.2 Default DreamMatter config


These are recommended defaults for the first benchmark and may be revised after profiling.

* Dreamlet count target: 250,000 to 1,000,000 depending on hardware tier
* default reference count: 500,000 Dreamlets
* primitive class: white sphere-like or shard-like micro-geometry
* dynamic radius of initial sphere: 1.25 meters
* target cube edge length: 2.5 meters
* emission disabled after initialization; benchmark focuses on state transition rather than continuous spawning
* promotion/materialization threshold explicit and logged

### 8.3 VRAM budget


Default benchmark VRAM budget:

* **2 GiB total**

This includes:

* DreamMatter buffers
* meshlet buffers
* PBR buffers
* shadow maps
* HDR target
* bloom mips
* IBL resources
* indirect draw buffers
* readback buffers
* fallback/default resources

The benchmark must log:

* total VRAM reserved
* estimated VRAM active by subsystem
* peak VRAM during run

### 8.4 Hardware tiers


The benchmark should support at minimum:

* Tier A: modern desktop GPU with mesh shader support
* Tier B: modern desktop GPU without mesh shader support but with compute capability sufficient for fallback path

Benchmark reports must identify which path was used.

---

## 9. Formal benchmark pass graph


The benchmark uses the standard Dreamwell render graph.

### 9.1 Per-frame sequence


1. update observer context
2. upload observer uniform
3. run Quantum Culling
4. run DreamMatter simulate
5. run DreamMatter convergence/materialization logic
6. run DreamMatter cull/compact
7. update DreamFabric promotion cache if thresholds are met
8. render shadow pass
9. render main PBR scene
10. render DreamMatter dynamic geometry or DreamFabric static geometry as applicable
11. run post-process bloom + tonemap
12. present
13. record KPIs

### 9.2 Phase ownership


| Phase                       | Owner                                            |
| --------------------------- | ------------------------------------------------ |
| Observer update             | `ObserverChannel`                                |
| Culling                     | `QuantumCullPass`                                |
| Simulation                  | `DreamMatterSystem`                              |
| Materialization / promotion | `DreamMatterSystem` + `DreamFabric` + `Tapestry` |
| Main scene render           | `GpuScene` / `DreamFabric`                       |
| Post-process                | `PostProcessStack`                               |
| KPI capture                 | benchmark harness                                |

---

## 10. Algorithms used and where


### 10.1 Algorithm A — Observer context upload


**Used in:** every frame before GPU compute begins

**Why:** ensures the benchmark uses the same observer-scoped path as production scenes

**Benchmark expectation:**

* one observer uniform write per frame
* no redundant uploads

### 10.2 Algorithm B — Quantum Culling


**Used in:** every frame, before DreamMatter render preparation and meshlet rendering

**Why:** validates observer-scoped visible-work reduction even in a simple demo room

**Benchmark expectation:**

* visible cluster count is logged
* culled cluster count is logged
* culling cost is logged in GPU timings

### 10.3 Algorithm C — DreamMatter simulation


**Used in:** every frame in dynamic, convergence, and dissolution states

**Why:** validates GPU-only matter simulation and stable frame cost under load

**Benchmark expectation:**

* active Dreamlet count logged
* simulation pass timing logged
* no CPU-side per-Dreamlet update

### 10.4 Algorithm D — DreamMatter convergence


**Used in:** convergence state

**Why:** validates topology transition from sphere swarm to cube target

**Benchmark expectation:**

* convergence completion ratio logged
* assembly coherence metric logged

### 10.5 Algorithm E — Settling classifier


**Used in:** convergence-to-settled transition

**Why:** validates transition from loose matter to low-motion cluster

**Benchmark expectation:**

* settled cluster count logged
* threshold crossing timestamp logged

### 10.6 Algorithm F — Materialization promotion


**Used in:** settled-to-materialized transition

**Why:** validates conversion from expensive dynamic matter to stable scene-shell representation

**Benchmark expectation:**

* materialization promotion event logged
* static representation generation timing logged
* dynamic Dreamlet cost reduction logged

### 10.7 Algorithm G — Dissolution


**Used in:** final optional phase

**Why:** validates reverse transition and confirms reversible DreamMatter lifecycle

**Benchmark expectation:**

* dissolution start/end timestamps logged
* restored dynamic count logged

---

## 11. KPI specification


The benchmark must emit structured KPIs per phase and for the full run.

### 11.1 Global KPIs


* benchmark name
* build hash
* runtime version
* GPU adapter name
* backend used
* mesh shader path or fallback path
* resolution
* total run duration
* average FPS
* p50 frame time
* p90 frame time
* p95 frame time
* p99 frame time
* max frame time
* average GPU frame time
* peak VRAM usage
* total VRAM reserved

### 11.2 DreamMatter KPIs


* configured Dreamlet capacity
* active Dreamlet count by frame
* visible Dreamlet count by frame
* compacted Dreamlet count by frame
* settled Dreamlet count by frame
* materialized DreamFabric count by frame
* dissolution count if used
* dropped Dreamlet events due to overflow

### 11.3 Quantum Culling KPIs


* total clusters considered
* visible clusters
* culled clusters
* indirect draw count
* average cull pass time
* p95 cull pass time

### 11.4 State transition KPIs


* time to stable swarm
* time to convergence start
* time to convergence completion
* time to settled threshold
* time to materialization
* time to dissolution start
* time to dissolution completion

### 11.5 Render pass KPIs


* observer upload time
* Quantum Culling time
* DreamMatter simulate time
* DreamMatter convergence/materialization time
* shadow pass time
* main PBR pass time
* post-process time
* total GPU frame time

---

## 12. Acceptance criteria


### 12.1 Functional acceptance


The benchmark passes functionally if:

* the swarm is visible on the left at start
* the swarm travels smoothly to the right
* the swarm converges into a cube-like shape
* the cube settles visibly
* the cube materializes into a stable render representation
* the optional dissolution state completes successfully if enabled
* no external content assets are required

### 12.2 Performance acceptance


The benchmark passes performance if all declared target hardware tiers meet their minimum thresholds.

Recommended baseline release thresholds for the default 2 GiB / 1920×1080 test:

#### Tier A — mesh shader capable desktop GPU


* average FPS >= 60
* p95 frame time <= 20 ms
* no catastrophic frame spikes during materialization transition
* no VRAM overflow

#### Tier B — fallback path desktop GPU


* average FPS >= 30
* p95 frame time <= 33 ms
* no VRAM overflow
* transitions remain visually correct under fallback path

### 12.3 Stability acceptance


The benchmark fails if:

* device lost is triggered by benchmark load
* buffers overflow incorrectly
* indirect args become invalid
* materialization causes sustained corruption or instability
* promotion/demotion oscillates uncontrollably
* frame pacing becomes pathological during state changes

---

## 13. Benchmark implementation checklist


### 13.1 Scene setup


* [ ] Create black room from primitive geometry only
* [ ] Create left-side DreamMatter sphere initial condition
* [ ] Create right-side cube target representation
* [ ] Configure fixed camera
* [ ] Configure sun light
* [ ] Configure PBR scene shell
* [ ] Configure HDR + bloom + tonemap
* [ ] Configure GI/IBL baseline

### 13.2 DreamMatter setup


* [ ] Initialize Dreamlet pool
* [ ] Set bounded capacity
* [ ] Set sphere swarm initial distribution
* [ ] Set cube target topology
* [ ] Set convergence policy
* [ ] Set settled threshold
* [ ] Set materialization promotion threshold
* [ ] Set optional dissolution policy

### 13.3 Benchmark harness


* [ ] Add fixed 30-second benchmark controller
* [ ] Add phase timestamps
* [ ] Add per-pass GPU timing queries
* [ ] Add per-frame KPI logging
* [ ] Add VRAM accounting log
* [ ] Add final summary emission
* [ ] Add optional video capture hook

### 13.4 Output


* [ ] Human-readable console summary
* [ ] machine-readable JSON or benchmark envelope
* [ ] optional CSV frame log
* [ ] optional capture/export artifact

---

## 14. Suggested benchmark output schema


```json id="3fz4pk"
{
  "benchmark": "dreamwell_3d_physics_benchmark_v1",
  "duration_seconds": 30,
  "resolution": { "width": 1920, "height": 1080 },
  "gpu": {
    "adapter": "<name>",
    "backend": "vulkan",
    "mesh_shader_path": true
  },
  "budgets": {
    "vram_budget_bytes": 2147483648,
    "vram_peak_bytes": 0
  },
  "dreammatter": {
    "capacity": 500000,
    "peak_active": 0,
    "peak_visible": 0,
    "peak_materialized_fabrics": 0,
    "dropped_spawns": 0,
    "dropped_promotions": 0
  },
  "timings": {
    "frame_ms_avg": 0.0,
    "frame_ms_p95": 0.0,
    "gpu_frame_ms_avg": 0.0,
    "quantum_cull_ms_avg": 0.0,
    "simulate_ms_avg": 0.0,
    "materialize_ms_avg": 0.0,
    "shadow_ms_avg": 0.0,
    "pbr_ms_avg": 0.0,
    "post_ms_avg": 0.0
  },
  "states": {
    "swarm_start_s": 0.0,
    "convergence_start_s": 10.0,
    "settled_start_s": 18.0,
    "materialized_start_s": 22.0,
    "dissolution_start_s": 27.0
  }
}
```

---

## 15. Hot-path requirements for this benchmark


The benchmark must exercise the production path, not a special-case fake path.

### Required hot-path rules


* no per-frame Dreamlet allocation
* no per-frame shader or pipeline creation
* no per-particle CPU updates
* no synchronous readback in steady-state path
* no asset loading on the frame path
* no per-frame rebuild of room or target geometry
* no debug overlay in canonical benchmark run

### Specific gotchas to avoid


* do not animate the sphere and cube with CPU transforms alone and call it DreamMatter
* do not swap a hidden static cube in without logging the true materialization transition
* do not let post-process dominate frame cost in this benchmark
* do not let shadow settings mask DreamMatter cost by overwhelming the GPU budget
* do not use unbounded emitter behavior
* do not exceed VRAM budget and silently rely on driver paging

---

## 16. Recommended visual polish constraints


To keep the benchmark technically honest:

* white matter only
* black room only
* no textures required
* no imported HDRI required
* no decals
* no animated cameras
* no stylized grading
* no cinematic tricks that hide instability

The benchmark should look clean, minimal, and technically legible.

---

## 17. Benchmark variants


After the canonical benchmark is stable, the following variants are recommended.

### Variant A — mesh shader path


Default high-end run.

### Variant B — vertex fallback path


Same visual benchmark, no mesh shaders.

### Variant C — no post-process


Measures raw DreamMatter + PBR + shadow cost.

### Variant D — no materialization


Measures sustained dynamic swarm cost.

### Variant E — repeated dissolve/materialize loop


Stress-tests state transition stability.

---

## 18. Final release rule


This benchmark is considered release-ready only if:

* the visual state progression is correct
* the KPI output is complete
* the benchmark is reproducible
* the benchmark remains inside the declared VRAM budget
* the benchmark uses the real production render graph
* both mesh-shader and fallback paths are validated where applicable
* the benchmark is simple enough that future regressions are attributable

---

## 19. Summary


`DREAMWELL_3D_PHYSICS_BENCHMARK.md` defines the first canonical DreamMatter visual benchmark: a 30-second, asset-free, 1920×1080 black-room scene in which a white Dreamlet sphere swarm transitions through all DreamMatter runtime states while traveling from left to right and materializing into a cube. The benchmark is explicitly designed to validate DreamMatter, DreamFabric, Quantum Culling, PBR integration, bounded VRAM behavior, and hot-path cleanliness under one controlled, reproducible workload.