Fr0zkaandClaude Opus 5.5 666dfb811c wip(standalone): GPU step 16 -- CPU cold drain; hybrid beats CPU once shaders are driver-cached ...

CPU uploads now bypass the 4-batch / 2 ms cap until the first resident view.
Luna's race (inside the Codex sandbox): CPU 1.354 s vs GPU 3.152 s, warm-up
queue 1.16-1.21 s on every launch. My runs outside the sandbox, same binary
(Step16FairRace, RunGpuFarViewer, interleaved): CPU 1.497 / 1.359 / 1.334 s;
GPU 3.088 (first launch of this build, warm-up 1,287 ms) / 1.262 / 1.260 s
(warm-up 3.2 / 3.1 ms). AMD DxCache was written at my GPU launches (08:46:57,
09:33:37) and never during Luna's runs: the sandbox cannot persist the driver
shader cache, which is why every one of Luna's launches paid the compile.
Audits: hashes, walk 5,642/5,642, 0 diffs vs CPU/WARP, survival 5.10 GB.

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

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

..

precompiled

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

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 16 -- CPU cold drain; hybrid beats CPU once shaders are driver-cached

2026-10-07 09:34:22 +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

GpuFarViewer.h

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

GpuFieldPort.cpp

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

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

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 16 -- CPU cold drain; hybrid beats CPU once shaders are driver-cached

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

RunGpuFarViewer.ps1

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

2026-10-06 11:01:03 +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

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 16 -- CPU cold drain; hybrid beats CPU once shaders are driver-cached

2026-10-07 09:34:22 +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 16 -- CPU cold drain; hybrid beats CPU once shaders are driver-cached

2026-10-07 09:34:22 +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

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

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