Beanstalk Longshot test farm
Menu

Report rp_cc0f0b718fad5a2f fail

#1576 perf: less game work a frame in Castle Town (registers as locals, vector rewrites)

any on jabbaxipil/Blaine,

Machine

Commit
f862753024c268ad743975b50e53f30743d531e4
Baseline
430eef4f2cc94e398b06d0daeb585a679d3ebcda
App
nightly-166-e4316243, attempt 2
Game build
e0f3ccfa
os
linux
cpu
AMD Ryzen 7 5800X3D 8-Core Processor
gpu
NVIDIA GeForce RTX 4070
tags
linuxnvidianvidia-proprietarydgpu
driver
615.71.9
ram_gb
64
vram_mb
12,282
gpu_vendor
nvidia
os_version
Ubuntu 26.04.1 LTS (kernel 7.0.0)

Builds

main (baseline)

0 s warm

from 430eef4f2cc9

the base was built already

PR

751.6 s warm

lift
93.5 s
read
2.1 s
finish
21.1 s
compile
634.6 s

Program cache: warm

Timed builds

Main and the PR each built whole in a fresh folder, from the same starting caches, alternating; shader preparation not timed. lift: the lifted C differs

CacheSide#SecondsStages (s)Lifted C
liftmain 1595.1 lift 77 · read 2.1 · finish 16.2 · compile 499.6 80ba986c2f42
liftPR 1750.2 lift 97.9 · read 2 · finish 16 · compile 634.1 7997de70f64b
liftPR 2746.7 lift 93.1 · read 2 · finish 16.1 · compile 635.3 7997de70f64b
liftmain 2590.7 lift 72.2 · read 2.1 · finish 15.9 · compile 500.4 80ba986c2f42

Scenarios

market-revisit

timeout

a step ran out of time: timeout: not met in time

Metric (median)mainPRΔ
first-visit.fps_mean64.9562.5−3.8%
first-visit.draws_mean2,210.52,320.1+5%
first-visit.hitch_count24+100%
first-visit.vram_mean_mb7,8587,578−3.6%
first-visit.vram_peak_mb7,8587,594−3.4%
first-visit.stutter_count1054+440%
first-visit.frametime_max_ms84.3272.99−13.4%
first-visit.frametime_p50_ms14.2314.28+0.4%
first-visit.frametime_p99_ms27.0235.87+32.7%
first-visit.frametime_1pct_low_ms36.9748.22+30.4%
screenshot_diff—0.57—

Pictures against main: market-first diff 0.5733 (main against itself )

Runs (2)
Metricmain 0pr 1
hangnono
crashednono
exit_code——
first-visit.fps_mean64.9562.5
first-visit.draws_mean2,210.52,320.1
first-visit.hitch_count24
first-visit.vram_mean_mb7,8587,578
first-visit.vram_peak_mb7,8587,594
first-visit.stutter_count1054
first-visit.frametime_max_ms84.3272.99
first-visit.frametime_p50_ms14.2314.28
first-visit.frametime_p99_ms27.0235.87
first-visit.frametime_1pct_low_ms36.9748.22
screenshot_diff—0.57

Rules

ScenarioCellRuleLimitmainPRΔ %Status
market-revisitcrashFalse00—pass
market-revisitfirst-visit.fps_mean.min_delta_pct1064.9562.5−3.8%pass
market-revisitreturn-visit.fps_mean.min_delta_pct10———skippedthe metric was not measured
market-revisitscreenshot.max_diff0.05—0.57—fail
market-revisitfirst-visit.frametime_p99_ms.max_delta_pct1527.0235.87+32.7%fail
market-revisitreturn-visit.frametime_p99_ms.max_delta_pct15———skippedthe metric was not measured

Pictures of the failure

The last frame of a step that got stuck, or main's and the PR's picture where they differed more than the plan allows; shrunk, and kept a day: until .

market-revisit-market-first-main-failed.png
market-revisit-market-first-main-failed.png
market-revisit-market-first-pr-failed.png
market-revisit-market-first-pr-failed.png

Log (PR side, end)

[hos host] vk: a frame's replay time (ms): submit 9.34, present 0.18, draw 5.70, descriptors 2.31, texture lookups 0.06, uploads 0.00, pipelines 0.23, programs 0.89, commands 1.84, dispatch 0.83, memory sync 0.49, queue submit 0.27, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.09 MB synced in 1194.7 copies, 21.39 MB written back
[hos host] vk: 7440 presents, 73.9 fps; a frame: 1939 draws (0 skipped), 240 dispatches, 150 passes, 142 barriers, 6 submits, 0.1 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2749 textures (5227 MB)
[hos host] vk: host time in GPU resource waits: 3.01 ms a frame (fences 0.01, timelines 2.99, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 521.3 full barriers recorded, 193.5 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.8; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 8.9 ms a frame (the longest submission 2.4 ms); 0.1 textures made a frame
[hos host] vk: a frame: 0.00 MB unswizzled on the CPU in 0.00 ms
[hos BaseProcExecutor2] mem: a copy of the same bytes left out: 4 KB to 0x1051459ad8 from 0x1051458548 (38000 so far, 324 MB)
[hos ModuleSystemWorker4] nvn: a texture initialized over storage that holds its packaged data already: 768x768 format 0x7b, 528 of 528 KB not written again (6800 such textures so far, 2484 MB)
[hos sead::MainThread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 555 ranges, 2448 MB), 414 pages written, 414 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.025; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.009, bytes copied 0.000; 25 kept (164 MB), 7 spare buffers, 2157 texture pages copied (8 MB)
[hos host] vk: replay reads of guest memory as submitted: 7040000 (4296037 pages from the captures)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1713 dirty pages from their submits' captures, 70738 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1403 dispatches, 42126 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9071 runs, 747.0 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24974 in captures, 5 only in the collection, 30 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7060000 (4308302 pages from the captures)
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 1536 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: render targets drawn over from another start, let go: 1311 so far (3589 MB), 1101 of them (2819 MB) between the passes
[hos host] vk: program picks: 9685340 kept (3414404 by the last picks' textures, a uniform buffer bound since), 2545 new, 3512114 with other textures (of 13200000)
[hos host] vk: replay reads of guest memory as submitted: 7080000 (4320569 pages from the captures)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.02, writes collected 0.04
[hos sead::MainThread] nvn: EventSignal (CPU) #15000: event 0x1006e38004 mode 0 value 1 (lr 0x8348c78c)
[hos host] vk: the recording streams: 48 of 2 MB, the most one held 301 KB; 0 made on demand (0 MB kept now); c1 stages: at most 75 a recording; 0 descriptor pools made when a recording's own ran out
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1603 dirty pages from their submits' captures, 69124 from earlier ones', 16 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1367 dispatches, 40740 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8846 runs, 727.8 MB compared; 69 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24230 in captures, 3 only in the collection, 34 only by the replay itself
[hos host] vk:   read live, by pool: 0x1065800000 (8 MB, captured, state 2): 16;
[hos ModuleSystemWorker2] nvn: CommandBufferWaitEvent #155000: event 0x100d6c8020 mode 0 value 114645 (now 114626) (lr 0x80664f90)
[hos host] vk: replay reads of guest memory as submitted: 7100000 (4332839 pages from the captures)
[hos host] vk: readback pages refreshed from guest memory: 2771787 pages so far (0 held: may be written ahead in their submission)
[hos ModuleSystemWorker5] nvn: CommandBufferSignalEvent #115000: event 0x100d6c8004 mode 0 value 114999 (lr 0x80291504)
[hos host] vk: replay reads of guest memory as submitted: 7120000 (4345030 pages from the captures)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1697 dirty pages from their submits' captures, 69956 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1385 dispatches, 41360 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8823 runs, 726.7 MB compared; 68 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24258 in captures, 1 only in the collection, 34 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos Presentation Thread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 555 ranges, 2436 MB), 400 pages written, 400 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.021; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.005, bytes copied 0.000; 25 kept (168 MB), 9 spare buffers, 2974 texture pages copied (11 MB)
[hos host] vk: replay reads of guest memory as submitted: 7140000 (4357319 pages from the captures)
[hos ModuleSystemWorker3] vk: readbacks of GPU results: 1345 of 3000 waited for their write-back, 0.20 ms each; 0 looked at the stopped queues first, 0 waited for their submission's last part (guest memory read in place after theirs), 0 waited for their whole submission (their pages written in place, no mirror)
[hos ModuleSystemWorker3] vk: invalidates since the last readbacks line: 2690 kept their virtual core (nothing there to wait for), 2942 gave it up (counted by 256 a thread; RX_VK_INVALIDATE_QUICK=0: none kept)
[hos ModuleSystemWorker1] nvn: DrawMeshTasksIndirect: 260000 so far (x1 0x11791f0000 x2 0 x3 0x11791f0000 x4 0x654040)
[hos host] vk: a frame: 323 draws kept their neighbour's descriptors (0 of them through other slots' registrations)
[hos host] vk: a frame: 0.0 draws left out (not ready), 0.00 waits for shaders or pipelines (0.0 ms); translation 0 ms in all, 0 jobs queued
[hos host] vk: a frame's replay time (ms): submit 8.51, present 0.16, draw 5.20, descriptors 2.11, texture lookups 0.06, uploads 0.00, pipelines 0.22, programs 0.81, commands 1.67, dispatch 0.77, memory sync 0.40, queue submit 0.25, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.11 MB synced in 1191.3 copies, 21.38 MB written back
[hos host] vk: 7560 presents, 76.5 fps; a frame: 1936 draws (0 skipped), 239 dispatches, 150 passes, 142 barriers, 6 submits, 0.1 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2749 textures (5227 MB)
[hos host] vk: host time in GPU resource waits: 2.89 ms a frame (fences 0.01, timelines 2.88, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 520.0 full barriers recorded, 192.8 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.4; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 9.1 ms a frame (the longest submission 2.7 ms); 0.1 textures made a frame
[hos host] vk: a frame: 0.00 MB unswizzled on the CPU in 0.00 ms
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 2353 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: replay reads of guest memory as submitted: 7160000 (4369534 pages from the captures)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1619 dirty pages from their submits' captures, 70059 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1381 dispatches, 41509 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9065 runs, 747.0 MB compared; 69 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25009 in captures, 5 only in the collection, 32 only by the replay itself
[hos host] vk:   read live, by pool:
[hos ModuleSystemWorker2] vcores: 171169 takes, 24001 waits (mean 40 us, longest 5.0 ms, 0 over HOS_VCORE_LOG_MS), 11194 kicks, 11497 yields, 0 starved grants, 0 escapes in 5 s
[hos host] vk: program picks: 9833711 kept (3464371 by the last picks' textures, a uniform buffer bound since), 2545 new, 3563743 with other textures (of 13400000)
[hos host] vk: replay reads of guest memory as submitted: 7180000 (4381765 pages from the captures)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.02, writes collected 0.04
[hos host] vk: replay reads of guest memory as submitted: 7200000 (4393988 pages from the captures)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1782 dirty pages from their submits' captures, 68993 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1367 dispatches, 40930 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8821 runs, 727.8 MB compared; 68 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24331 in captures, 7 only in the collection, 35 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7220000 (4406228 pages from the captures)
[hos Presentation Thread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 555 ranges, 2436 MB), 405 pages written, 405 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.021; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.004, bytes copied 0.000; 25 kept (144 MB), 12 spare buffers, 2974 texture pages copied (11 MB)
[hos host] vk: replay reads of guest memory as submitted: 7240000 (4418452 pages from the captures)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1706 dirty pages from their submits' captures, 70975 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1402 dispatches, 42136 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9081 runs, 747.1 MB compared; 68 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25028 in captures, 4 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 2017 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos Presentation Thread] vi: 7653 frames presented (76.2 fps over the last 5.0 s, interval 1)
[hos Presentation Thread] vi: present intervals: <15.5 ms 361, 15.5-17.8 12, 17.8-22 7, 22-30 1, 30-40 0, over 40 0
[hos host] vk: replay reads of guest memory as submitted: 7260000 (4430715 pages from the captures)
[hos sead::MainThread] nvn: GPU waits in 5 s: 337 sync waits (421 ms), 0 polls timed out (0 ms), 0 queue finishes (0 ms)
[hos host] vk: replay reads of guest memory as submitted: 7280000 (4442985 pages from the captures)
[hos host] vk: the recording streams: 48 of 2 MB, the most one held 301 KB; 0 made on demand (0 MB kept now); c1 stages: at most 75 a recording; 0 descriptor pools made when a recording's own ran out
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1677 dirty pages from their submits' captures, 68888 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1367 dispatches, 40947 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8813 runs, 727.7 MB compared; 58 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24305 in captures, 6 only in the collection, 34 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: program picks: 9981924 kept (3514030 by the last picks' textures, a uniform buffer bound since), 2545 new, 3615530 with other textures (of 13600000)
[hos host] vk: a frame: 321 draws kept their neighbour's descriptors (0 of them through other slots' registrations)
[hos host] vk: a frame: 0.0 draws left out (not ready), 0.00 waits for shaders or pipelines (0.0 ms); translation 0 ms in all, 0 jobs queued
[hos host] vk: a frame's replay time (ms): submit 8.47, present 0.18, draw 5.18, descriptors 2.10, texture lookups 0.05, uploads 0.00, pipelines 0.21, programs 0.82, commands 1.65, dispatch 0.76, memory sync 0.41, queue submit 0.27, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.12 MB synced in 1193.6 copies, 21.38 MB written back
[hos host] vk: 7680 presents, 77.2 fps; a frame: 1933 draws (0 skipped), 239 dispatches, 150 passes, 142 barriers, 6 submits, 0.0 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2749 textures (5227 MB)
[hos host] vk: host time in GPU resource waits: 2.81 ms a frame (fences 0.01, timelines 2.80, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 519.6 full barriers recorded, 192.7 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.3; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 9.0 ms a frame (the longest submission 2.9 ms); 0.0 textures made a frame
[hos host] vk: a frame: 0.00 MB unswizzled on the CPU in 0.00 ms
[hos ModuleSystemWorker2] nvn: a texture initialized over storage that holds its packaged data already: 1536x1536 format 0x7b, 2030 of 2030 KB not written again (7000 such textures so far, 2546 MB)
[hos host] vk: readback pages refreshed from guest memory: 2844884 pages so far (0 held: may be written ahead in their submission)
[hos host] vk: replay reads of guest memory as submitted: 7300000 (4455236 pages from the captures)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.03, writes collected 0.04
[hos ModuleSystemWorker3] nvn: CommandBufferWaitEvent #160000: event 0x100d6c8020 mode 0 value 118445 (now 118426) (lr 0x80664f90)
[hos ModuleSystemWorker4] vk: readbacks of GPU results: 1359 of 3000 waited for their write-back, 0.22 ms each; 0 looked at the stopped queues first, 0 waited for their submission's last part (guest memory read in place after theirs), 0 waited for their whole submission (their pages written in place, no mirror)
[hos ModuleSystemWorker4] vk: invalidates since the last readbacks line: 2783 kept their virtual core (nothing there to wait for), 3105 gave it up (counted by 256 a thread; RX_VK_INVALIDATE_QUICK=0: none kept)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1753 dirty pages from their submits' captures, 71123 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1401 dispatches, 42150 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9073 runs, 747.0 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25040 in captures, 7 only in the collection, 34 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: replay reads of guest memory as submitted: 7320000 (4467506 pages from the captures)
[hos sead::MainThread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 556 ranges, 2432 MB), 409 pages written, 409 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.022; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.005, bytes copied 0.000; 25 kept (170 MB), 6 spare buffers, 1770 texture pages copied (6 MB)
[hos BaseProcExecutor3] mem: a copy of the same bytes left out: 4 KB to 0x1050890400 from 0x105088ee70 (40000 so far, 331 MB)
[hos host] vk: replay reads of guest memory as submitted: 7340000 (4479757 pages from the captures)
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 2128 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1762 dirty pages from their submits' captures, 68769 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1368 dispatches, 40943 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8826 runs, 727.8 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24378 in captures, 7 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7360000 (4491990 pages from the captures)
[hos ModuleSystemWorker5] nvn: 1150000 recordings made, 48071 live copies (18.0 MB), 799 texture copies; 6 command buffers' segments (4.3 MB)
[hos host] vk: replay reads of guest memory as submitted: 7380000 (4504250 pages from the captures)
[hos host] vk: program picks: 10130461 kept (3563910 by the last picks' textures, a uniform buffer bound since), 2545 new, 3666993 with other textures (of 13800000)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1713 dirty pages from their submits' captures, 70120 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1400 dispatches, 41492 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8821 runs, 726.8 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25074 in captures, 7 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: replay reads of guest memory as submitted: 7400000 (4516595 pages from the captures)
[hos ModuleSystemWorker3] nvn: CommandBufferSignalEvent #120000: event 0x100d6c8014 mode 0 value 120000 (lr 0x80660bcc)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.02, writes collected 0.04
[hos host] vk: replay reads of guest memory as submitted: 7420000 (4528757 pages from the captures)
[hos host] vk: memory: working set 12663 MB (peak 13662), private 4137 MB, staging 32 MB in 2 arenas, 49 made so far (17 ms)
[hos host] vk: memory heaps: heap 0 (device): 228 blocks 7421 MB, 15514 allocations 6568 MB; used 7554 of 10280 MB; heap 1 (system): 15 blocks 816 MB, 40 allocations 770 MB; used 2906 of 48166 MB;
[hos host] vk: image blocks: 1 blocks 32 MB, 5 images 16 MB; 1 free ranges, the largest 15 MB
[hos host] vk: image blocks (render targets): 14 blocks 448 MB, 96 images 250 MB; 46 free ranges, the largest 31 MB
[hos host] vk: image blocks (textures, compacted): 122 blocks 3904 MB, 2646 images 3754 MB; 1478 free ranges, the largest 12 MB
[hos host] vk: render target images: 548 reused from the pool, 865 made new, 762 destroyed from it; it holds 0 (0 MB)
[hos host] vk: VRAM: 7554 MB used of 10280 MB allowed (853 MB of it free in the blocks); textures 5231 MB (budget 7164 MB), 2753 textures
[hos host] vk: textures: decoded ASTC 3214 MB, rendered 940 MB (62 drawn in one frame, kept: 33438 KB), other 731 MB; used in the last 8 presents 2280 MB
[hos host] vk: decoded ASTC used again: 485 textures (1149 MB) copied from ones the game finalized, not decoded; 485 of those retired, 379 listed; 1118 MB of blocks hashed in 149 ms
[hos host] vk:   288 MB: 0x1147c16000 12288x12288x1 x1 levels 1 NVN 0x32 rendered, last used 1 presents ago
[hos host] vk:   214 MB: 0x10a5b22c00 1365x1365x1 x90 levels 11 NVN 0x80 ASTC, last used 1 presents ago
[hos host] vk:   214 MB: 0x10a22aa600 1365x1365x1 x90 levels 11 NVN 0x8e ASTC, last used 1 presents ago
[hos host] vk:   64 MB: 0x113cba9000 512x512x64 x1 levels 1 NVN 0x3d rendered, last used 1 presents ago
[hos host] vk:   64 MB: 0x113755c000 4096x4096x1 x1 levels 1 NVN 0x34 rendered, last used 1 presents ago
[hos host] vk:   54 MB: 0x100032c0400 3072x3072x1 x3 levels 1 NVN 0x32 rendered, last used 1 presents ago
[hos host] vk:   54 MB: 0x10002d97a00 3072x3072x1 x3 levels 1 NVN 0x32 rendered, last used 6162 presents ago
[hos host] vk:   54 MB: 0x100030a4200 3072x3072x1 x3 levels 1 NVN 0x32 rendered, last used 4289 presents ago
[hos host] vk: finalized textures the GPU drew once, kept for being made again: 23 (2416 KB): 242x40x1 NVN 0x25: 4 (151 KB); 260x260x1 NVN 0xd: 9 (1583 KB); 306x104x1 NVN 0x1: 4 (124 KB); 592x49x1 NVN 0x25: 3 (339 KB); 656x113x1 NVN 0x1: 3 (217 KB);
[hos Presentation Thread] vk: presents after the CPU's signal (600): 524 waited for it, 1.20 ms each, 0 let go after RX_VK_FRAME_GATE_MS
[hos host] vk: status: 600 frames, the slowest 21.7 ms; VRAM 7.4 of 10.0 GB
[hos host] vk: program census (play): created 0 (0 with stages; 0 with no present for 100 ms before, 0 in a frame of 100 ms or more; 0.0 ms of the game's threads), freed 0 (0 never used); first uses 0: program ready 0, loaded 0, translated 0, failed 0; pipeline ready 0, from a cache 0, compiled 0, made 0, left out 0, none 0; 0 with no create call seen, 0 whose stages are not the create call's; after the create call: under 16.7 ms 0, under 100 ms 0, under 1 s 0, longer 0; needing work 0 (0 of them 100 ms or more after the create call, 0 created in a load or a long frame); in play: 123 program records loaded (80.6 ms), 2212 new graphics pipelines, pipelines from the driver's cache 726, from the start's cache 0, compiled 1624; 223253 draws and dispatches left out, 585 waits (83320.6 ms)
[hos sead::MainThread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 556 ranges, 2442 MB), 405 pages written, 405 copied (1.6 MB)
[hos host] vk: program loads at create (play): 5596 queued, 5596 loaded (median 0.40 ms, 99th percentile 3.86 ms, worst 13.25 ms, 4092 ms and 1245.7 MB read in all); 7798 programs (1346 of them compute variants without their code, 0 graphics programs no note names), 11 graphics pipelines (9996 not in a cache), 0 compute pipelines (619 not in a cache); a first use ran 0 itself and waited for 0 (0.0 ms); 3 pipelines made at create taken by draws; 3160 create calls with nothing kept for their stages, 0 before the index; units 8756 (5596 ready, 0 loading), 14878 programs alive; 0 noted graphics pipeline probes deferred to first use
[hos host] vk: module identifiers (play): 691 known, 691 learned in this run; 0 compute pipelines made by them at a create call, 0 asked that way were in no cache (their code read, as before)
[hos host] vk: a frame: 324 draws kept their neighbour's descriptors (0 of them through other slots' registrations)
[hos host] vk: a frame: 0.0 draws left out (not ready), 0.00 waits for shaders or pipelines (0.0 ms); translation 0 ms in all, 0 jobs queued
[hos host] vk: a frame's replay time (ms): submit 8.67, present 0.17, draw 5.33, descriptors 2.17, texture lookups 0.06, uploads 0.00, pipelines 0.22, programs 0.84, commands 1.70, dispatch 0.79, memory sync 0.41, queue submit 0.26, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.12 MB synced in 1192.1 copies, 21.38 MB written back
[hos host] vk: 7800 presents, 76.2 fps; a frame: 1939 draws (0 skipped), 239 dispatches, 150 passes, 142 barriers, 6 submits, 0.1 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2753 textures (5231 MB)
[hos host] vk: host time in GPU resource waits: 2.95 ms a frame (fences 0.01, timelines 2.94, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 520.4 full barriers recorded, 193.0 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.4; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 9.1 ms a frame (the longest submission 2.7 ms); 0.1 textures made a frame
[hos host] vk: a frame: 0.03 MB unswizzled on the CPU in 0.00 ms
[hos host] vk: the render thread's whole frames (600 presents): 0 jobs went on before their present after 1000 ms, 0 times a game's thread waited for the work handed over, 0 times for a fence of the frame in the making
[hos host] vk: the render thread (600 presents, 13.06 ms a frame): its jobs 9.15 ms a frame (submits 1.78, presents 0.18, others 7.20), idle 3.91; the game waited for room at 0 presents, 0.00 ms each (0.00 a frame; 153.51 jobs queued at a present); 0 catch-ups (a finish, a texture read back, a write-back not recorded yet: 0), 0.00 ms each; 0 presents held for the CPU's signal, 0.00 ms each (0 let go by RX_VK_HOLD_MS); 536 captures at a CPU signal; the submits' waits read in 0.047 ms a frame; readbacks: 3005 next to a texture's storage (no catch-up), 0 in one, 0 read exactly (after the work handed over: among the catch-ups); a present made 9.19 ms after its hand-over (21.71 at most)
[hos host] vk: frame overlap (600 presents): the GPU between recordings, ms a frame, idle + in the next one's uploads: up to a frame's first recording with a draw 2.13 + 0.35 (the window's copy is in the idle), after it 1.10 + 0.50; after a frame's last recording that reads or writes the game's memory come 1.01 recordings (0.03 ms of GPU time), after its last that reads it 1.01 (0.03 ms)
[hos host] vk: frame ahead (600 presents): 10.53 parts a frame recorded in the signal's job, 71.0% of them took nothing that came with the signal; of the others, 3.05 a frame took pages of the signal's capture into a mirror (47.7 pages a frame) and 0.00 pages read as guest memory had them (0.0 pages a frame); a frame's first part took nothing in 0 of 536 frames, its first three in 0
[hos host] vk: frame ahead, what came with the signal: 536 signals with a capture, each of 74.2 pages in 12.2 pools; between a frame's last submit and that signal 0.00 texture and sampler registrations, 0.00 releases and 0.00 other CPU signals a signal (0 of them with any)
[hos host] vk: frame overlap, by place: idle + uploads before a frame's 2nd to 12th recording with a draw, then all later ones: 0.06+0.05, 0.07+0.02, 0.70+0.11, 0.14+0.06, 0.01+0.15, 0.02+0.01, 0.02+0.02, 0.03+0.02, 0.04+0.02, 0.01+0.02, 0.01+0.01, 0.00+0.00
[hos host] vk: a capture taken (ms): all 0.021; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.004, bytes copied 0.000; 25 kept (158 MB), 7 spare buffers, 2282 texture pages copied (8 MB)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1690 dirty pages from their submits' captures, 70086 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1375 dispatches, 41413 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9059 runs, 747.0 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24311 in captures, 1 only in the collection, 30 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7440000 (4540908 pages from the captures)
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 1616 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: replay reads of guest memory as submitted: 7460000 (4552998 pages from the captures)
[hos host] vk: the recording streams: 48 of 2 MB, the most one held 301 KB; 0 made on demand (0 MB kept now); c1 stages: at most 75 a recording; 0 descriptor pools made when a recording's own ran out
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1676 dirty pages from their submits' captures, 68355 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1369 dispatches, 40708 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8813 runs, 727.7 MB compared; 69 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24328 in captures, 5 only in the collection, 35 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: replay reads of guest memory as submitted: 7480000 (4565152 pages from the captures)
[hos ModuleSystemWorker1] vk: readbacks of GPU results: 1327 of 3000 waited for their write-back, 0.21 ms each; 0 looked at the stopped queues first, 0 waited for their submission's last part (guest memory read in place after theirs), 0 waited for their whole submission (their pages written in place, no mirror)
[hos ModuleSystemWorker1] vk: invalidates since the last readbacks line: 2685 kept their virtual core (nothing there to wait for), 2947 gave it up (counted by 256 a thread; RX_VK_INVALIDATE_QUICK=0: none kept)
[hos host] vk: readback pages refreshed from guest memory: 2918354 pages so far (0 held: may be written ahead in their submission)
[hos host] vk: program picks: 10278817 kept (3613683 by the last picks' textures, a uniform buffer bound since), 2545 new, 3718637 with other textures (of 14000000)
[hos host] vk: replay reads of guest memory as submitted: 7500000 (4577299 pages from the captures)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1724 dirty pages from their submits' captures, 69728 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1391 dispatches, 41437 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8828 runs, 726.8 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25059 in captures, 3 only in the collection, 34 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos sead::MainThread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 555 ranges, 2444 MB), 404 pages written, 404 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.022; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.005, bytes copied 0.000; 25 kept (164 MB), 6 spare buffers, 3055 texture pages copied (11 MB)
[hos host] vk: replay reads of guest memory as submitted: 7520000 (4589468 pages from the captures)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.02, writes collected 0.04
[hos ModuleSystemWorker1] nvn: CommandBufferWaitEvent #165000: event 0x100d6c8020 mode 0 value 122245 (now 122226) (lr 0x80664f90)
[hos host] vk: queues ordered by their events: 54000 submissions stopped at a wait, 53983 went on once it was reached, 14 waits gone past; made to go by presents 0, releases 0, waits for the replay 0, 50 ms 1 (stopped 51.59 ms at most); waits met 165029, not met 0; captures at a CPU event's wait 4869, 0 of them at a wait met on arrival
[hos host] vk: replay reads of guest memory as submitted: 7540000 (4601602 pages from the captures)
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 2389 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos rmbl::VibrationThread] hid: 40960 vibration values, strongest 0.00 since the last count
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1735 dirty pages from their submits' captures, 70103 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1376 dispatches, 41472 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9057 runs, 747.0 MB compared; 68 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24342 in captures, 2 only in the collection, 32 only by the replay itself
[hos host] vk:   read live, by pool:
[hos ModuleSystemWorker3] nvn: a texture initialized over storage that holds its packaged data already: 256x512 format 0x4b, 692 of 692 KB not written again (7200 such textures so far, 2619 MB)
[hos host] vk: a frame: 324 draws kept their neighbour's descriptors (0 of them through other slots' registrations)
[hos host] vk: a frame: 0.0 draws left out (not ready), 0.00 waits for shaders or pipelines (0.0 ms); translation 0 ms in all, 0 jobs queued
[hos host] vk: a frame's replay time (ms): submit 8.60, present 0.18, draw 5.31, descriptors 2.18, texture lookups 0.09, uploads 0.01, pipelines 0.22, programs 0.83, commands 1.68, dispatch 0.78, memory sync 0.43, queue submit 0.25, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.09 MB synced in 1190.5 copies, 21.38 MB written back
[hos host] vk: 7920 presents, 77.2 fps; a frame: 1937 draws (0 skipped), 239 dispatches, 150 passes, 142 barriers, 6 submits, 0.2 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2761 textures (5232 MB)
[hos host] vk: host time in GPU resource waits: 2.95 ms a frame (fences 0.01, timelines 2.93, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 520.4 full barriers recorded, 192.8 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.3; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 9.0 ms a frame (the longest submission 2.8 ms); 0.2 textures made a frame
[hos host] vk: a frame: 0.01 MB unswizzled on the CPU in 0.00 ms
[hos host] vk: replay reads of guest memory as submitted: 7560000 (4613758 pages from the captures)
[hos host] vk: replay reads of guest memory as submitted: 7580000 (4625936 pages from the captures)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1717 dirty pages from their submits' captures, 68509 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1368 dispatches, 40807 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8832 runs, 727.8 MB compared; 68 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24333 in captures, 4 only in the collection, 34 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7600000 (4638078 pages from the captures)
[hos host] vk: program picks: 10427191 kept (3663568 by the last picks' textures, a uniform buffer bound since), 2545 new, 3770263 with other textures (of 14200000)
[hos aal::LowPrioWorkerThread] vcores: 177841 takes, 24339 waits (mean 37 us, longest 3.6 ms, 0 over HOS_VCORE_LOG_MS), 11371 kicks, 11622 yields, 0 starved grants, 0 escapes in 5 s
[hos sead::MainThread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 555 ranges, 2446 MB), 407 pages written, 407 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.024; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.007, bytes copied 0.000; 25 kept (164 MB), 6 spare buffers, 3698 texture pages copied (14 MB)
[hos host] vk: replay reads of guest memory as submitted: 7620000 (4650268 pages from the captures)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1735 dirty pages from their submits' captures, 69947 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1396 dispatches, 41492 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8924 runs, 744.2 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25083 in captures, 4 only in the collection, 32 only by the replay itself
[hos host] vk:   read live, by pool:
[hos BaseProcExecutor2] mem: a copy of the same bytes left out: 4 KB to 0x1057494970 from 0x10574933e0 (42000 so far, 339 MB)
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 5808 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.03, writes collected 0.04
[hos host] vk: replay reads of guest memory as submitted: 7640000 (4662448 pages from the captures)
[hos ModuleSystemWorker3] nvn: DrawMeshTasksIndirect: 280000 so far (x1 0x117db40030 x2 0 x3 0x117db40000 x4 0x44b040)
[hos ModuleSystemWorker4] vk: readbacks of GPU results: 1333 of 3000 waited for their write-back, 0.21 ms each; 0 looked at the stopped queues first, 0 waited for their submission's last part (guest memory read in place after theirs), 0 waited for their whole submission (their pages written in place, no mirror)
[hos ModuleSystemWorker4] vk: invalidates since the last readbacks line: 2720 kept their virtual core (nothing there to wait for), 2912 gave it up (counted by 256 a thread; RX_VK_INVALIDATE_QUICK=0: none kept)
[hos host] vk: replay reads of guest memory as submitted: 7660000 (4674547 pages from the captures)
[hos host] vk: the recording streams: 48 of 2 MB, the most one held 301 KB; 0 made on demand (0 MB kept now); c1 stages: at most 75 a recording; 0 descriptor pools made when a recording's own ran out
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1636 dirty pages from their submits' captures, 69915 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1370 dispatches, 41280 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8974 runs, 729.6 MB compared; 66 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24319 in captures, 3 only in the collection, 32 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: readback pages refreshed from guest memory: 2991402 pages so far (0 held: may be written ahead in their submission)
[hos host] vk: replay reads of guest memory as submitted: 7680000 (4686777 pages from the captures)
[hos Presentation Thread] vi: 8036 frames presented (76.5 fps over the last 5.0 s, interval 1)
[hos Presentation Thread] vi: present intervals: <15.5 ms 374, 15.5-17.8 6, 17.8-22 2, 22-30 1, 30-40 0, over 40 0
[hos host] vk: a frame: 323 draws kept their neighbour's descriptors (0 of them through other slots' registrations)
[hos host] vk: a frame: 0.0 draws left out (not ready), 0.00 waits for shaders or pipelines (0.0 ms); translation 0 ms in all, 0 jobs queued
[hos host] vk: a frame's replay time (ms): submit 8.67, present 0.16, draw 5.32, descriptors 2.16, texture lookups 0.05, uploads 0.00, pipelines 0.22, programs 0.82, commands 1.72, dispatch 0.78, memory sync 0.42, queue submit 0.27, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.10 MB synced in 1190.6 copies, 21.38 MB written back
[hos host] vk: 8040 presents, 75.9 fps; a frame: 1938 draws (0 skipped), 239 dispatches, 150 passes, 142 barriers, 6 submits, 0.0 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2761 textures (5232 MB)
[hos host] vk: host time in GPU resource waits: 2.71 ms a frame (fences 0.01, timelines 2.70, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 520.2 full barriers recorded, 193.1 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.5; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 8.9 ms a frame (the longest submission 2.5 ms); 0.0 textures made a frame
[hos host] vk: a frame: 0.00 MB unswizzled on the CPU in 0.00 ms
[hos sead::MainThread] nvn: GPU waits in 5 s: 336 sync waits (391 ms), 0 polls timed out (0 ms), 0 queue finishes (0 ms)
[hos sead::MainThread] nvn: CommandBufferSignalEvent #125000: event 0x100d6c8020 mode 0 value 125000 (lr 0x80664f14)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1751 dirty pages from their submits' captures, 70321 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1400 dispatches, 42288 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9041 runs, 747.6 MB compared; 65 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25182 in captures, 11 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: replay reads of guest memory as submitted: 7700000 (4698879 pages from the captures)
[hos host] vk: program picks: 10575694 kept (3713424 by the last picks' textures, a uniform buffer bound since), 2545 new, 3821760 with other textures (of 14400000)
[hos Presentation Thread] vk: capture: a submit 0.16 ms (lock 0.00, scan 0.03 over 555 ranges, 2446 MB), 417 pages written, 417 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.025; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.008, bytes copied 0.000; 25 kept (168 MB), 12 spare buffers, 3724 texture pages copied (14 MB)
[hos host] vk: ASTC texture #5571 0x1132f63a00 512x512 x1 levels 10: the image of #4892 0x10f1b66600 (the same blocks, finalized by the game) copied, not decoded again (500 so far, 1181 MB)
[hos host] vk: replay reads of guest memory as submitted: 7720000 (4711072 pages from the captures)
[hos host] vk: render targets drawn over from another start, let go: 1311 so far (3589 MB), 1101 of them (2819 MB) between the passes
[hos host] vk: present 8077 written to /<user>/<user>/Beanstalk/jobs/jb_aa9567eb25764a3f/runs/market-revisit-g0-01-pr/shots/market-revisit-pr-01-last-frame.png (the control bridge)
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 3498 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1873 dirty pages from their submits' captures, 70183 from earlier ones', 16 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1397 dispatches, 41884 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8863 runs, 728.3 MB compared; 104 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24654 in captures, 24 only in the collection, 31 only by the replay itself
[hos host] vk:   read live, by pool: 0x1065800000 (8 MB, captured, state 2): 16;
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7740000 (4723462 pages from the captures)
[hos host] vk: recorder: 10474752 packets run in 10.0 s (busy 2.00 s, spun idle 4.80 s, slept 2816 times, 988 of them under 200 us), 0 calls made directly (no room); the replay waited for it 36074 times, 128.47 ms in all, the longest 1.27 ms; vkEndCommandBuffer 13193/72.15 ms (max 0.26), replay_mu unlock 9738/28.49 ms (max 0.34), vkResetDescriptorPool 13143/27.83 ms (max 1.27)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.03, writes collected 0.05
[hos sead::MainThread] nvn: CommandBufferWaitEvent #170000: event 0x100d6c8020 mode 0 value 126045 (now 126026) (lr 0x80664f90)
[hos host] vk: replay reads of guest memory as submitted: 7760000 (4735765 pages from the captures)
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1956 dirty pages from their submits' captures, 70114 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1381 dispatches, 41760 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9067 runs, 748.1 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24409 in captures, 11 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: replay reads of guest memory as submitted: 7780000 (4747937 pages from the captures)
[hos ModuleSystemWorker1] nvn: 1200000 recordings made, 48107 live copies (18.5 MB), 799 texture copies; 5 command buffers' segments (4.3 MB)
[hos ModuleSystemWorker1] nvn: a texture initialized over storage that holds its packaged data already: 1024x1024 format 0x7b, 912 of 912 KB not written again (7400 such textures so far, 2699 MB)
[hos host] vk: replay reads of guest memory as submitted: 7800000 (4760199 pages from the captures)
[hos sead::MainThread] vk: capture: a submit 0.16 ms (lock 0.00, scan 0.03 over 555 ranges, 2452 MB), 419 pages written, 419 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.025; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.007, bytes copied 0.000; 25 kept (170 MB), 6 spare buffers, 4446 texture pages copied (17 MB)
[hos host] vk: program picks: 10723951 kept (3763413 by the last picks' textures, a uniform buffer bound since), 2545 new, 3873503 with other textures (of 14600000)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1743 dirty pages from their submits' captures, 69565 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1387 dispatches, 41016 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8829 runs, 726.8 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 25063 in captures, 3 only in the collection, 31 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos ModuleSystemWorker1] vk: readbacks of GPU results: 1320 of 3000 waited for their write-back, 0.18 ms each; 0 looked at the stopped queues first, 0 waited for their submission's last part (guest memory read in place after theirs), 0 waited for their whole submission (their pages written in place, no mirror)
[hos ModuleSystemWorker1] vk: invalidates since the last readbacks line: 2659 kept their virtual core (nothing there to wait for), 2973 gave it up (counted by 256 a thread; RX_VK_INVALIDATE_QUICK=0: none kept)
[hos host] vk: replay reads of guest memory as submitted: 7820000 (4772459 pages from the captures)
[hos host] vk: a frame: 322 draws kept their neighbour's descriptors (0 of them through other slots' registrations)
[hos host] vk: a frame: 0.0 draws left out (not ready), 0.00 waits for shaders or pipelines (0.0 ms); translation 0 ms in all, 0 jobs queued
[hos host] vk: a frame's replay time (ms): submit 8.98, present 0.32, draw 5.49, descriptors 2.24, texture lookups 0.07, uploads 0.00, pipelines 0.23, programs 0.86, commands 1.74, dispatch 0.81, memory sync 0.45, queue submit 0.26, slot wait 0.03
[hos host] vk: a frame: mirrors 226 (785 MB): 8.13 MB synced in 1197.0 copies, 21.38 MB written back
[hos host] vk: 8160 presents, 74.3 fps; a frame: 1941 draws (0 skipped), 239 dispatches, 150 passes, 142 barriers, 6 submits, 0.1 uploads (0.0 MB), 80 ub updates (56 overlays), GPU-bound wait 0.0 ms (recording slots only); 35016 pipelines (8779587 ms), 0 programs (0 failed), 2761 textures (5232 MB)
[hos host] vk: host time in GPU resource waits: 2.54 ms a frame (fences 0.01, timelines 2.53, idle 0.00, queries 0.00; sum across threads, not frame time; excludes pacing and driver submit/present blocking)
[hos host] vk: a frame: 520.2 full barriers recorded, 193.1 left out (nothing recorded since the one before), 0.0 pass starts ordered in their fragment stages only (RX_VK_PASS_BARRIER=game)
[hos host] vk: a frame: the game's barriers by their NVN flags: 0x1: 2.0; 0x8: 3.0; 0x9: 1.0; 0x11: 38.7; 0x12: 71.5; 0x13: 2.0; 0x21: 4.0; 0x28: 16.0; 0x92: 3.0; 0x97: 1.0;
[hos host] vk: GPU time 9.0 ms a frame (the longest submission 2.5 ms); 0.1 textures made a frame
[hos host] vk: a frame: 0.00 MB unswizzled on the CPU in 0.00 ms
[hos host] vk: shadow staging (600 takes): 0 shadow copies from staging, 3529 from the capture buffer; capture buffers: 600 host-cached, 0 not (write-combined: staged)
[hos host] vk: replay reads of guest memory as submitted: 7840000 (4784728 pages from the captures)
[hos host] vk: the recording streams: 48 of 2 MB, the most one held 301 KB; 0 made on demand (0 MB kept now); c1 stages: at most 75 a recording; 0 descriptor pools made when a recording's own ran out
[hos host] vk: write-backs left out (600 submits): 735, 3605.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1670 dirty pages from their submits' captures, 70488 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1381 dispatches, 41599 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 9076 runs, 747.0 MB compared; 67 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24310 in captures, 3 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: replay reads of guest memory as submitted: 7860000 (4796961 pages from the captures)
[hos host] vk: a submit's memory work (ms): maps 0.00, capture taken 0.03, writes collected 0.05
[hos ModuleSystemWorker2] tex: the game's texture heap: 123 compaction rounds asked for in 7200 frames without one running
[hos host] vk: readback pages refreshed from guest memory: 3064558 pages so far (0 held: may be written ahead in their submission)
[hos host] vk: replay reads of guest memory as submitted: 7880000 (4809207 pages from the captures)
[hos host] vk: sparse mirrors: 90, 785 MB with memory of 1244 MB (458 MB spared); blocks given memory later 9556, given back 11700; bindings served from guest memory meanwhile 0 (0 reaching blocks with memory too), from the mirror past every game pool 0; 500 binds (85.5 ms); 0 made whole for want of memory
[hos host] vk: sparse mirrors' buffers: 75 taken from the spares, 151 made new; 136 kept as spares, 0 destroyed; 61 spares now, 984 MB of address space (RX_VK_SPARSE_SPARE_MB=4096)
[hos host] vk: write-backs left out (600 submits): 714, 3502.3 MB; 28 bindings followed, 40 ranges the CPU read, 0 resets; CPU reads or writes of one left out: 0 so far
[hos host] vk: mirror syncs (600 submits): 1795 dirty pages from their submits' captures, 69349 from earlier ones', 0 from guest memory as it was at the replay; 25 captures kept; 0 pages the GPU wrote copied over (0 live); 0 kept for the GPU (RX_VK_PROTECT_BARE)
[hos host] vk: mirror syncs' pass (600 submits): 1367 dispatches, 40880 chunks; 0 syncs as copies (no table room)
[hos host] vk: write-back diff (600 submits): 8817 runs, 727.8 MB compared; 65 pages read into the shadows; 336 MB of shadows; pages the shadows knew that the CPU wrote: 24321 in captures, 6 only in the collection, 33 only by the replay itself
[hos host] vk:   read live, by pool:
[hos host] vk: imports: 167 live (2068 MB), 0 retired still imported (0 MB), 44 retired given back (616 MB)
[hos host] vk: replay reads of guest memory as submitted: 7900000 (4821442 pages from the captures)
[hos sead::MainThread] vk: capture: a submit 0.15 ms (lock 0.00, scan 0.03 over 554 ranges, 2450 MB), 409 pages written, 409 copied (1.6 MB)
[hos host] vk: a capture taken (ms): all 0.022; GPU wait for late pages 0.000 (0 of 600 takes), pools lock 0.000, retires and stamps 0.005, bytes copied 0.000; 25 kept (168 MB), 6 spare buffers, 4446 texture pages copied (17 MB)
[hos host] vk: program picks: 10872218 kept (3813310 by the last picks' textures, a uniform buffer bound since), 2545 new, 3925236 with other textures (of 14800000)
[hos host] vk: replay reads of guest memory as submitted: 7920000 (4833707 pages from the captures)