Fr0zkaandClaude Opus 5.5 7d92e148a9 wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun ...

Luna: on a marker miss the deferred warmup no longer gates the initial view
(starts after the full CPU view is presented). The forced-cold DXGI 0x887A0005 was
ours: GetDeviceRemovedReason = DEVICE_HUNG after the audit readback of tile
(-7,-7,0,lod8), slot 1433 -- the readback copied the theoretical 61,440-vertex maximum
from a 19,456-vertex pooled slot, 41,984 vertices past the end of the last chunk's
buffer. Readback now clamps to slot capacity (audit-only path; normal launches never
read back). Auto and forced-cold survival both pass (1,434 submissions, ~44.6 s, <8 GB).
Added a no-write option for the warmup marker. Direction answer in LUNA_DIALOGUE.md.

My runs outside the sandbox (Step26.exe, marker removed first, interleaved, -Timeline):
  CPU 1.299 / 1.291 / 1.266 s
  GPU marker miss 1.319 s (all-CPU, = CPU again; step 25 was 2.015 s)
  GPU cached 1.151 / 1.122 / 1.138 s, all 1,434 GPU tiles (timestamps 243 / 307 / 241 ms,
      max dispatch 6.3 / 66.7 / 5.2 ms)
Audits: hashes w1/w8, walk 5,642/5,642, near/far 0 sign/topology diffs vs CPU and
WARP on protected/thin, ten captures pass.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MmSoF64BbRFmk9C3p1VsBJ

2026-10-08 03:22:38 +02:00

..

precompiled

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

AnalyzeStep6NearAudit.ps1

wip(standalone): add missing GPU skirt shader binaries and step 6-8 scripts/report

2026-10-07 06:14:04 +02:00

AnalyzeStep15Timeline.py

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

BuildGpuFieldPort.ps1

wip(standalone): GPU step 3 -- GPU decides far rock/air, sparse CPU exceptions, hybrid viewer (H)

2026-10-06 09:14:22 +02:00

BuildGpuFieldStep6NearRunner.ps1

wip(standalone): add missing GPU skirt shader binaries and step 6-8 scripts/report

2026-10-07 06:14:04 +02:00

BuildGpuFieldViewer.ps1

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

BuildGpuFirstSubmitRepro.ps1

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

CompareFarCpuHashes.ps1

wip(standalone): GPU field step 2 -- far-field shapes/strata/relief in shared C++/HLSL, GPU field+mesh

2026-10-06 08:02:02 +02:00

CompareStep6Captures.py

wip(standalone): add missing GPU skirt shader binaries and step 6-8 scripts/report

2026-10-07 06:14:04 +02:00

CompareStep7Captures.py

wip(standalone): add missing GPU skirt shader binaries and step 6-8 scripts/report

2026-10-07 06:14:04 +02:00

CompareStep8Captures.py

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

CompareStep8LodIsolations.py

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

CompareStep9Captures.py

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

CompareStep10Captures.py

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

CompareStep11Captures.py

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

CompareStep12Captures.py

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

CompareStep13Captures.py

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

CompareStep14Captures.py

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

CompareStep15Captures.py

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

CompareStep17TileControls.py

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

CompileStep6GpuShaders.ps1

wip(standalone): add missing GPU skirt shader binaries and step 6-8 scripts/report

2026-10-07 06:14:04 +02:00

CompileStep14Shader.ps1

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

GpuFarViewer.h

wip(standalone): GPU step 24 -- warmup policy off the critical path; first launch = CPU

2026-10-08 02:06:49 +02:00

GpuFieldPort.cpp

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

GpuFirstSubmitRepro.cpp

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

LUNA_DIALOGUE.md

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

REPORT.md

wip(standalone): GPU step 3 -- GPU decides far rock/air, sparse CPU exceptions, hybrid viewer (H)

2026-10-06 09:14:22 +02:00

RunGpuFarAgreement.ps1

wip(standalone): GPU step 24 -- warmup policy off the critical path; first launch = CPU

2026-10-08 02:06:49 +02:00

RunGpuFarViewer.ps1

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

RunGpuFieldPort.ps1

wip(standalone): GPU field step 2 -- far-field shapes/strata/relief in shared C++/HLSL, GPU field+mesh

2026-10-06 08:02:02 +02:00

RUNNERS.md

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

RunStep6CaptureMatrix.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6ColdInterleaved.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6FarAgreement.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6FarPacked.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6GpuCrossing.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6NearAgreement.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6NearDiag.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6NearDiagFull.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6NearMismatchProbe.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep6NearPacked.ps1

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

RunStep7CaptureMatrix.ps1

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

RunStep7ColdInterleaved.ps1

wip(standalone): GPU step 7 partial -- parity check deferred off cold path (stopped early)

2026-10-06 20:17:37 +02:00

RunStep7Process.ps1

wip(standalone): GPU step 7 partial -- parity check deferred off cold path (stopped early)

2026-10-06 20:17:37 +02:00

RunStep8CaptureMatrix.ps1

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

RunStep8ColdInterleaved.ps1

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

RunStep8PresentControls.ps1

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

RunStep8PresentVisibilityControls.ps1

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

RunStep9NormalWindowInterleaved.ps1

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

RunStep10CaptureMatrix.ps1

wip(standalone): GPU step 10 -- first-submit stall bisected to cold shader use under 8 workers

2026-10-07 05:38:14 +02:00

RunStep10CpuReference.ps1

wip(standalone): GPU step 10 -- first-submit stall bisected to cold shader use under 8 workers

2026-10-07 05:38:14 +02:00

RunStep10FinalInterleaved.ps1

wip(standalone): GPU step 10 -- first-submit stall bisected to cold shader use under 8 workers

2026-10-07 05:38:14 +02:00

RunStep10GpuSurvival.ps1

wip(standalone): GPU step 10 -- first-submit stall bisected to cold shader use under 8 workers

2026-10-07 05:38:14 +02:00

RunStep10WarmupBisect.ps1

wip(standalone): GPU step 10 -- first-submit stall bisected to cold shader use under 8 workers

2026-10-07 05:38:14 +02:00

RunStep11AllCpuAffinity.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep11CaptureMatrix.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep11CpuReference.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep11FinalInterleaved.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep11GpuSurvival.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep11ShaderFirstUse.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep11WarmupControls.ps1

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

RunStep12CaptureMatrix.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12ConnectorLoopMeasurement.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12CpuReference.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12FinalInterleaved.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12FinalShaderFirstUse.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12GpuCrossing.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12GpuSurvival.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12PostPoolInterleaved.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12ShaderBaseline.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12ShaderFirstUse.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12SkirtLoopMeasurement.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep12WholeNearAgreement.ps1

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

RunStep13CaptureMatrix.ps1

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

RunStep13CpuReference.ps1

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

RunStep13FinalInterleaved.ps1

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

RunStep13GpuSurvival.ps1

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

RunStep13ShaderFirstUse.ps1

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

RunStep14CaptureMatrix.ps1

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

RunStep14CpuReference.ps1

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

RunStep14FinalInterleaved.ps1

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

RunStep14GpuSurvival.ps1

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

RunStep14ShaderFirstUse.ps1

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

RunStep14WarmupProfile.ps1

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

RunStep15CaptureMatrix.ps1

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

RunStep15CpuReference.ps1

wip(standalone): GPU step 16 -- CPU cold drain; hybrid beats CPU once shaders are driver-cached

2026-10-07 09:34:22 +02:00

RunStep15GpuSurvival.ps1

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

RunStep15Timeline.ps1

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

RunStep16CpuDrainControl.ps1

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

RunStep16CpuPresentControl.ps1

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

RunStep16FairRace.ps1

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

RunStep16GpuWarmupControl.ps1

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

RunStep17TileControls.ps1

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

StartGpuFarViewer.ps1

wip(standalone): GPU step 4 -- batched far generation, connector panels on GPU; far survival passes

2026-10-06 11:01:03 +02:00

step3_cpu_far.png

wip(standalone): GPU step 3 -- GPU decides far rock/air, sparse CPU exceptions, hybrid viewer (H)

2026-10-06 09:14:22 +02:00

step3_gpu_far.png

wip(standalone): GPU step 3 -- GPU decides far rock/air, sparse CPU exceptions, hybrid viewer (H)

2026-10-06 09:14:22 +02:00

step6_capture_comparison.md

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_chamber.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_far.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_passage_floor_closeup.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_passage.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_remote_yaw0.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_remote_yaw90.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_remote_yaw180.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_remote_yaw270.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_tunnel_bend.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_cpu_tunnel.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_chamber.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_far.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_passage_floor_closeup.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_passage_normalfix.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_passage.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_remote_yaw0.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_remote_yaw90.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_remote_yaw180.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_remote_yaw270.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_tunnel_bend.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step6_gpu_tunnel.png

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

step7_capture_comparison.md

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_chamber.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_far.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_passage_floor_closeup.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_passage.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_remote_yaw0.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_remote_yaw90.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_remote_yaw180.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_remote_yaw270.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_tunnel_bend.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_cpu_tunnel.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_chamber.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_far.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_passage_floor_closeup.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_passage.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_remote_yaw0.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_remote_yaw90.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_remote_yaw180.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_remote_yaw270.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_tunnel_bend.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_gpu_tunnel.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_probe_cpu_remote_yaw0.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_probe_gpu_remote_yaw0.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_seam_cpu_remote_yaw0.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step7_seam_gpu_remote_yaw0.png

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

step8_capture_comparison.md

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_clamped_capture_comparison.md

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_chamber.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_far.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_passage_floor_closeup.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_passage.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_remote_yaw0.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_remote_yaw90.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_remote_yaw180.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_remote_yaw270.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_tunnel_bend.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_cpu_tunnel.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_chamber.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_far.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_passage_floor_closeup.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_passage.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_protected_remote_yaw0.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_protected_remote_yaw90.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw0.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw90_direct_audit.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw90_drawargs_full.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw90_drawargs.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw90.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw180.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_remote_yaw270.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_tunnel_bend.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_gpu_tunnel.png

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_lod_isolation_comparison.md

wip(standalone): GPU step 8 partial -- GPU skirts, first-present hold; remote gap open

2026-10-07 02:19:38 +02:00

step8_report.md

wip(standalone): add missing GPU skirt shader binaries and step 6-8 scripts/report

2026-10-07 06:14:04 +02:00

step9final_capture_comparison.md

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_cpu_remote_yaw0.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_cpu_remote_yaw90.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_cpu_remote_yaw180.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_cpu_remote_yaw270.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_gpu_remote_yaw0.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_gpu_remote_yaw90.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_gpu_remote_yaw180.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_gpu_remote_yaw270.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_lod3_remote_cpu.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_lod3_remote_gpu.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_lod4_remote_cpu.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_lod4_remote_gpu.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_remote_cpu.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step9final_remote_gpu.png

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

step10_capture_comparison.md

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

step11_capture_comparison.md

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

step12_capture_comparison.md

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

step13_capture_comparison.md

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

step14_capture_comparison.md

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

step15_capture_comparison.md

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step15_timeline_fast_cpu_tiles.csv

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step15_timeline_fast_frames.csv

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step15_timeline_fast_gpu_batches.csv

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step15_timeline_paced_cpu_tiles.csv

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step15_timeline_paced_frames.csv

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step15_timeline_paced_gpu_batches.csv

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

step16_capture_comparison.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

step17_final_capture_comparison.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

step17_final_r29_capture_comparison.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

step17_final_r29_tilecontrol_comparison.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

step17_r26_tilecontrol_comparison.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

step17_tilecontrol_comparison.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

step18_final_capture_comparison.md

wip(standalone): GPU step 18 -- warm-up marker: hybrid now wins cached and ties first launch

2026-10-07 21:07:05 +02:00

step18_tilecontrol_comparison.md

wip(standalone): GPU step 18 -- warm-up marker: hybrid now wins cached and ties first launch

2026-10-07 21:07:05 +02:00

step19_capture_capture_comparison.md

wip(standalone): GPU step 19 -- streamed GPU input export; RUNNERS.md

2026-10-07 22:09:29 +02:00

step19_final_capture_comparison.md

wip(standalone): GPU step 19 -- streamed GPU input export; RUNNERS.md

2026-10-07 22:09:29 +02:00

step19_warm_capture_comparison.md

wip(standalone): GPU step 19 -- streamed GPU input export; RUNNERS.md

2026-10-07 22:09:29 +02:00

step20final_capture_comparison.md

wip(standalone): GPU step 20 -- detail band batched by 8 columns; gap instrumented

2026-10-07 22:35:22 +02:00

step21_capture_comparison.md

wip(standalone): GPU step 21 -- GPU input export shares the 8 workers (3:1 priority); no wall gain

2026-10-07 23:20:42 +02:00

step22_capture_comparison.md

wip(standalone): GPU step 22 -- startup multi-batch pump, pool before handoff: cached 1.12-1.20 s

2026-10-08 00:00:06 +02:00

step22_final_capture_comparison.md

wip(standalone): GPU step 22 -- startup multi-batch pump, pool before handoff: cached 1.12-1.20 s

2026-10-08 00:00:06 +02:00

step23_final_capture_comparison.md

wip(standalone): GPU step 23 -- skip empty connector-ring dispatch; cached 1.07-1.17 s

2026-10-08 00:32:13 +02:00

step23_ring_skip_capture_comparison.md

wip(standalone): GPU step 23 -- skip empty connector-ring dispatch; cached 1.07-1.17 s

2026-10-08 00:32:13 +02:00

step23_step22_control_capture_comparison.md

wip(standalone): GPU step 23 -- skip empty connector-ring dispatch; cached 1.07-1.17 s

2026-10-08 00:32:13 +02:00

step24_final384_capture_comparison.md

wip(standalone): GPU step 24 -- warmup policy off the critical path; first launch = CPU

2026-10-08 02:06:49 +02:00

step24_final_capture_comparison.md

wip(standalone): GPU step 24 -- warmup policy off the critical path; first launch = CPU

2026-10-08 02:06:49 +02:00

step25_final384_capture_comparison.md

wip(standalone): GPU step 25 -- shape profile, roughness endpoint reuse, export before CPU frame; first launch REGRESSED

2026-10-08 02:55:32 +02:00

step26_final384_capture_comparison.md

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00

STEP4_REPORT.md

wip(standalone): GPU step 4 -- batched far generation, connector panels on GPU; far survival passes

2026-10-06 11:01:03 +02:00

STEP5_REPORT.md

wip(standalone): GPU step 5 -- far GPU work 646 -> 112 ms; hybrid whole view equal to CPU

2026-10-06 13:50:07 +02:00

STEP6_REPORT.md

wip(standalone): GPU step 6 -- near field on GPU, exact; cold start still slower than CPU

2026-10-06 20:03:10 +02:00

STEP7_RESUME_REPORT.md

wip(standalone): GPU step 7 -- viewer cold start 2.45 -> 1.49 s on CPU; GPU skirts missing

2026-10-06 22:39:40 +02:00

STEP9_REPORT.md

wip(standalone): GPU step 9 -- remote band closed by CPU fallback; CPU still wins whole view

2026-10-07 04:36:34 +02:00

STEP10_REPORT.md

wip(standalone): GPU step 10 -- first-submit stall bisected to cold shader use under 8 workers

2026-10-07 05:38:14 +02:00

STEP11_REPORT.md

wip(standalone): GPU step 11 -- per-shader cold cost measured; far EmitCS split, still CPU wins

2026-10-07 06:13:59 +02:00

STEP12_REPORT.md

wip(standalone): GPU step 12 -- far evaluator cold use 1704 -> 3 ms; near detail now the floor

2026-10-07 06:51:53 +02:00

STEP13_REPORT.md

wip(standalone): GPU step 13 -- near base/skirts/connectors cheaper to compile; warm-up still ~2.0 s

2026-10-07 07:41:31 +02:00

STEP14_REPORT.md

wip(standalone): GPU step 14 -- warm-up 2.04 -> 1.05 s; cold path now waits on streaming, not shaders

2026-10-07 08:03:00 +02:00

STEP15_REPORT.md

wip(standalone): GPU step 15 -- cold-start drain: hybrid whole view 3.2 -> 1.25-1.68 s

2026-10-07 08:47:36 +02:00

STEP16_REPORT.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

STEP17_REPORT.md

wip(standalone): GPU step 17 -- far skirt-depth fix (LOD3/4 back on GPU); adaptive start trades cached speed for first-launch

2026-10-07 20:36:40 +02:00

STEP18_REPORT.md

wip(standalone): GPU step 19 -- streamed GPU input export; RUNNERS.md

2026-10-07 22:09:29 +02:00

STEP19_REPORT.md

wip(standalone): GPU step 19 -- streamed GPU input export; RUNNERS.md

2026-10-07 22:09:29 +02:00

STEP20_REPORT.md

wip(standalone): GPU step 20 -- detail band batched by 8 columns; gap instrumented

2026-10-07 22:35:22 +02:00

STEP21_REPORT.md

wip(standalone): GPU step 21 -- GPU input export shares the 8 workers (3:1 priority); no wall gain

2026-10-07 23:20:42 +02:00

STEP22_REPORT.md

wip(standalone): GPU step 22 -- startup multi-batch pump, pool before handoff: cached 1.12-1.20 s

2026-10-08 00:00:06 +02:00

STEP23_REPORT.md

wip(standalone): GPU step 23 -- skip empty connector-ring dispatch; cached 1.07-1.17 s

2026-10-08 00:32:13 +02:00

STEP24_REPORT.md

wip(standalone): GPU step 24 -- warmup policy off the critical path; first launch = CPU

2026-10-08 02:06:49 +02:00

STEP25_REPORT.md

wip(standalone): GPU step 25 -- shape profile, roughness endpoint reuse, export before CPU frame; first launch REGRESSED

2026-10-08 02:55:32 +02:00

STEP26_REPORT.md

wip(standalone): GPU step 26 -- first launch fixed again; device-removed was our readback overrun

2026-10-08 03:22:38 +02:00