Repository navigation
Conversation
github-actions
Bot
added
perf
|
This pull request is labeled |
adriadam10
requested changes
Oct 28, 2026
adriadam10
left a comment
adriadam10
left a comment
Collaborator
There was a problem hiding this comment.
The code and the test look right (guest_thread_host_affinity fails without the Thread.cpp change), and keeping it opt-in is reasonable. One thing is missing before merging:
- The
perflabel needs the evidence the Performance report form asks for (see the bot comment above): theaps5-summary.jsonlline of a capture onmainand one with the change, from the same title, scene and machine, both traces attached, and the output oftools/trace_digest.pyfor each. The table in### Testedhas the numbers but not the captures behind them. Please paste those and the PR can go on.
MarcZX10
suggested changes
Oct 28, 2026
MarcZX10
left a comment
MarcZX10
left a comment
Contributor
There was a problem hiding this comment.
Reviewed — changes requested
The affinity change and its focused test build successfully, but this perf PR is missing the paired trace evidence required by the repository's performance-report rules.
What I ran
Reviewed exact head d74683cb2b088c1d2c596e47b92a437cd4dd994e against current main 319621ba2e578bd053b0baba2877fe70520e0737; the merge tree is clean. Environment: Debian 12 x86_64, GCC 12.2.0, CMake, Ninja.
git diff --check refs/review/current-main...refs/review/12544
git merge-tree --write-tree refs/review/current-main refs/review/12544
cmake -S /tmp/AnyPS5-review-12544 -B /tmp/AnyPS5-review-12544/build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON
cmake --build /tmp/AnyPS5-review-12544/build --target guest_thread_host_affinity_tests -j4
ctest --test-dir /tmp/AnyPS5-review-12544/build -R '^guest_thread_host_affinity$' --output-on-failure
gh pr checks 12544 --repo boykopovar/AnyPS5Results
| Check | Result |
|---|---|
Merge with current main |
Clean |
| Diff check | Passed; no output |
| CMake configure | Passed |
guest_thread_host_affinity_tests build |
Passed, 196/196 build steps |
guest_thread_host_affinity |
Did not reach assertions: the host CPU guard aborts on this Xeon E5-1620 v2, which lacks required AVX2/FMA and other extensions |
| Conventions, eligibility, intake, labels, report | Passed |
| Hosted Windows | Job passed; changed test selection and CTest were skipped, and changed host-only test executables linked |
| Hosted Linux, register-context, relinker, ASan, write-watch | Skipped by workflow |
Raw output
3f32fd9561cfb012905bcb98478f740733bfa66e
[195/196] Building CXX object core/libs/prx/libkernel/CMakeFiles/libkernel.dir/Time/Time.cpp.o
[196/196] Linking CXX executable tests/guest_thread_host_affinity_tests
1/1 Test #2311: guest_thread_host_affinity .......Subprocess aborted***Exception: 0.01 sec
[src/HostCpu.cpp:193 operator()] host CPU Intel(R) Xeon(R) CPU E5-1620 v2 @ 3.70GHz: has SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CMPXCHG16B, LAHF/SAHF, AVX, F16C, AES, PCLMULQDQ, RDRAND, XSAVEOPT, OS AVX state (XCR0); lacks LZCNT, BMI1, BMI2, MOVBE, AVX2, FMA, RDSEED, ADX, SHA, CLFLUSHOPT, CLWB, RDPID, XSAVEC, SSE4A, PREFETCHW, CLZERO
[libc] unhandled exception 'St13runtime_error' thrown from 0x7f86db880e4d
terminate called after throwing an instance of 'std::runtime_error'
what(): The host CPU (Intel(R) Xeon(R) CPU E5-1620 v2 @ 3.70GHz) lacks LZCNT, BMI1, BMI2, MOVBE, AVX2, FMA. The PS5's Zen 2 CPU has these extensions and titles run their instructions natively, so this CPU cannot run them.
labels pass 11s
windows pass 9m35s
linux skipping 0
register-context skipping 0
relinker skipping 0
relinker-asan skipping 0
write-watch skipping 0
conventions pass 15s
eligibility pass 5s
intake pass 6s
report pass 40s
Negative control
| Assertion | Change made to code under test | Did the new test fail? |
|---|---|---|
| A guest thread receives the requested host CPU mask while the creating thread keeps its mask. | No mutation; the test aborted in the project CPU guard before assertions. | Not exercised. |
Findings
Blocking: The PR is labeled perf, but it does not include the two aps5-summary.jsonl lines for main and the change, the matching trace attachments, or tools/trace_digest.py output. The report form requires those captures from the same machine, title, scene, and settings; the current presents-per-10-second table does not provide the requested capture data. Please attach both traces, paste both summary lines, and include each digest output.
Non-blocking: This host cannot run the focused test because it fails the project's CPU-feature guard before reaching test assertions. The hosted Windows job linked changed host-only test targets but skipped CTest; the Linux and specialized jobs were skipped. The PR description reports separate Linux test runs and title measurements, which I did not independently reproduce.
Sources
- Affinity implementation at the reviewed head
- Focused affinity test at the reviewed head
- Environment-variable documentation at the reviewed head
- Performance-report requirements at the reviewed base
- PR and performance rules at the reviewed base
Not verified
I could not run the test assertions on this CPU, independently verify the hybrid-core behavior on supported hardware, reproduce the title captures, or verify the reported performance delta. A QEMU rerun was unavailable because the temporary worktree and executable were no longer present after the first CTest attempt.
To get this to a good state
Attach the paired base/change traces, paste their aps5-summary.jsonl lines, and add the output of python3 tools/trace_digest.py for each trace. Please also run guest_thread_host_affinity on a CPU that passes the project's feature guard. I will rerun the same review commands against the updated head.
|
Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.
|
Contributor
Author
|
Paired captures are below: same build, title, scene and machine. The only difference between them is They are not from a build of Both captures use commit 4fd12fcaa9d4 of that stack ("dirty" is only the
The 10 s summary intervals around the captures agree: 35 to 37 presents without the variable, 38 to 41 with it. Summary line without: {"commit":"4fd12fcaa9d4","dirty":true,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"115:15","APS5_TRACE_SUMMARY_EVERY":"10"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted™: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 115:15","first_present":2035,"ms":15010.576},"presents":54,"frame_ms":{"p50":274.747,"p90":311.123,"p99":355.030,"max":355.030},"gpu_busy_ms_per_present":103.516,"worker_busy_pct":{"queue worker 0x0":73.501,"queue worker 0x43":77.755,"queue worker 0x21":10.866,"queue worker 0x52":0.195,"queue worker 0x50":0.175,"queue worker 0x29":14.694,"queue worker 0x56":99.897,"queue worker 0x39":42.680,"queue worker 0x49":34.319,"queue worker 0x31":32.003,"queue worker 0x41":32.799},"top_phases_ms_per_frame":{"Driver.Packet":1159.480,"Driver.Packet.memory_wait":746.334,"Driver.Packet.dispatch":258.444,"Driver.Dispatch":223.868,"Driver.Packet.draw":121.516,"Driver.Draw":120.990,"Graphics.Draw":97.047,"Graphics.Draw.draw_and_resource_release":92.772},"trace_dropped":true,"trace_dropped_threads":{"queue worker 0x0":{"events_full_after_s":11.470,"events_lost":80478,"totals_full":false}},"trace":"aps5-trace-4fd12fcaa9d4-1208546.json.gz"}Summary line with: {"commit":"4fd12fcaa9d4","dirty":true,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_GUEST_AFFINITY":"1","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"115:15","APS5_TRACE_SUMMARY_EVERY":"10"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted™: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 115:15","first_present":2148,"ms":15011.492},"presents":60,"frame_ms":{"p50":248.756,"p90":274.771,"p99":294.253,"max":294.253},"gpu_busy_ms_per_present":98.558,"worker_busy_pct":{"queue worker 0x0":73.454,"queue worker 0x43":77.199,"queue worker 0x21":11.619,"queue worker 0x50":0.250,"queue worker 0x52":0.191,"queue worker 0x29":17.039,"queue worker 0x56":99.980,"queue worker 0x41":38.626,"queue worker 0x39":38.527,"queue worker 0x49":33.558,"queue worker 0x31":37.132},"top_phases_ms_per_frame":{"Driver.Packet":1065.943,"Driver.Packet.memory_wait":690.066,"Driver.Packet.dispatch":241.375,"Driver.Dispatch":206.140,"Driver.Packet.draw":106.747,"Driver.Draw":106.308,"Graphics.Draw":88.036,"Graphics.Draw.draw_and_resource_release":84.402},"trace_dropped":true,"trace_dropped_threads":{"queue worker 0x0":{"events_full_after_s":11.060,"events_lost":95808,"totals_full":false}},"trace":"aps5-trace-4fd12fcaa9d4-1217548.json.gz"}
|
| thread | busy | wait | untraced | in slices < 20 us | largest |
|---|---|---|---|---|---|
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.0 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.0 |
| title | 0.0 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.0 |
| title | 0.0 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.0 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.0 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.0 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| queue worker 0x0 | 112.6 (41 %) | 76.8 (28 %) | 88.5 | 122.0 | Driver.Packet 134.2, Driver.Rewind 0.2 |
| title | 0.1 (0 %) | 0.0 (0 %) | 277.9 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| draw committer | 44.5 (16 %) | 1.0 (0 %) | 232.5 | 27.1 | Graphics.Draw 33.4, Recorder.Sync 10.3 |
| presenter | 0.4 (0 %) | 0.1 (0 %) | 277.5 | 0.1 | VideoOut.Flip 0.5 |
| queue worker 0x43 | 68.0 (24 %) | 208.5 (75 %) | 1.5 | 3.4 | Driver.Packet 215.1, Driver.EndSubmission 0.1 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 278.0 | 0.0 | Vulkan.ComputePipeline 0.0 |
| queue worker 0x21 | 13.4 (5 %) | 259.4 (93 %) | 5.1 | 6.5 | Driver.Packet 29.2, Driver.EndSubmission 0.6 |
| queue worker 0x52 | 0.3 (0 %) | 272.4 (98 %) | 5.2 | 0.2 | Driver.Packet 0.4, Driver.EndSubmission 0.1 |
| queue worker 0x50 | 0.4 (0 %) | 272.4 (98 %) | 5.2 | 0.3 | Driver.Packet 0.4, Driver.EndSubmission 0.0 |
| queue worker 0x29 | 14.1 (5 %) | 262.6 (94 %) | 1.3 | 6.4 | Driver.Packet 39.1, Driver.EndSubmission 0.6 |
| queue worker 0x56 | 1.3 (0 %) | 272.9 (98 %) | 3.8 | 0.9 | Driver.Packet 274.2, Driver.EndSubmission 0.0 |
| queue worker 0x39 | 14.0 (5 %) | 258.8 (93 %) | 5.1 | 6.4 | Driver.Packet 117.6, Driver.EndSubmission 0.6 |
| queue worker 0x49 | 10.8 (4 %) | 262.0 (94 %) | 5.2 | 4.5 | Driver.Packet 94.6, Driver.EndSubmission 0.5 |
| queue worker 0x31 | 14.9 (5 %) | 258.0 (93 %) | 5.2 | 6.4 | Driver.Packet 87.5, Driver.EndSubmission 1.1 |
| queue worker 0x41 | 13.4 (5 %) | 263.3 (95 %) | 1.2 | 5.7 | Driver.Packet 89.4, Driver.EndSubmission 0.6 |
| gpu | 103.8 (37 %) | 0.0 (0 %) | 174.2 | 0.0 | batch 103.8 |
Largest slices (ms per frame, inclusive, all threads): Driver.WorkerWait 1843.3, Driver.WorkerWait.dequeue_order 1802.1, Driver.Packet 1081.6, Driver.Packet.memory_wait 734.4, Driver.Packet.dispatch 249.8, Driver.Dispatch 214.1, batch 103.8, Driver.Packet.draw 81.3, Driver.Draw 81.0, Graphics.Draw 72.7
Longest waits
- Driver.WorkerWait (queue worker 0x52) 369.9 ms at 5980 ms, present 2056
- Driver.WorkerWait (queue worker 0x52) 354.5 ms at 11790 ms, present 2078
- Driver.WorkerWait (queue worker 0x50) 354.5 ms at 11790 ms, present 2078
- Driver.WorkerWait (queue worker 0x31) 343.2 ms at 11801 ms, present 2078
- Driver.Packet.memory_wait (queue worker 0x56) 343.1 ms at 13329 ms, present 2083
- Driver.WorkerWait (queue worker 0x50) 341.9 ms at 4637 ms, present 2051
- Driver.WorkerWait (queue worker 0x52) 341.5 ms at 7357 ms, present 2061
- Driver.WorkerWait (queue worker 0x52) 340.4 ms at 14095 ms, present 2085
- Driver.WorkerWait (queue worker 0x52) 332.9 ms at 4353 ms, present 2050
- Driver.WorkerWait (queue worker 0x52) 330.5 ms at 3544 ms, present 2047
Largest GPU batches (GPU ranges inside, and the CPU slices that recorded each, inclusive)
- batch 81789: 6.2 ms at 12795 ms, present 2081: GPU no ranges; recorded by no CPU slices
- batch 81662: 6.2 ms at 12527 ms, present 2080: GPU no ranges; recorded by no CPU slices
- batch 78197: 6.2 ms at 4795 ms, present 2052: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.9, Driver.Packet.dispatch (queue worker 0x0) 5.2, Driver.Dispatch (queue worker 0x0) 5.2, Vulkan.DispatchRecipe (queue worker 0x0) 5.1
- batch 79691: 6.2 ms at 8100 ms, present 2064: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.9, Driver.Packet.dispatch (queue worker 0x0) 5.3, Driver.Dispatch (queue worker 0x0) 5.3, Vulkan.DispatchRecipe (queue worker 0x0) 5.2
- batch 77954: 6.2 ms at 4225 ms, present 2050: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.6, Driver.Packet.dispatch (queue worker 0x0) 5.4, Driver.Dispatch (queue worker 0x0) 5.3, Vulkan.DispatchRecipe (queue worker 0x0) 5.2
tools/trace_digest.py, with
seconds 115:15 from present 2148: 60 presents, frame p50 248.8 / p90 274.8 / p99 294.3 / max 294.3 ms, GPU busy 98.6 ms per present
NVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit 4fd12fcaa9d4 (dirty), dropped: yes
{"commit":"4fd12fcaa9d4","dirty":true,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_GUEST_AFFINITY":"1","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"115:15","APS5_TRACE_SUMMARY_EVERY":"10"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted\u2122: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 115:15","first_present":2148,"ms":15011.492},"presents":60,"frame_ms":{"p50":248.756,"p90":274.771,"p99":294.253,"max":294.253},"gpu_busy_ms_per_present":98.558,"worker_busy_pct":{"queue worker 0x0":73.454,"queue worker 0x43":77.199,"queue worker 0x21":11.619,"queue worker 0x50":0.25,"queue worker 0x52":0.191,"queue worker 0x29":17.039,"queue worker 0x56":99.98,"queue worker 0x41":38.626,"queue worker 0x39":38.527,"queue worker 0x49":33.558,"queue worker 0x31":37.132},"top_phases_ms_per_frame":{"Driver.Packet":1065.943,"Driver.Packet.memory_wait":690.066,"Driver.Packet.dispatch":241.375,"Driver.Dispatch":206.14,"Driver.Packet.draw":106.747,"Driver.Draw":106.308,"Graphics.Draw":88.036,"Graphics.Draw.draw_and_resource_release":84.402},"trace_dropped":true,"trace_dropped_threads":{"queue worker 0x0":{"events_full_after_s":11.06,"events_lost":95808,"totals_full":false}},"trace":"aps5-trace-4fd12fcaa9d4-1217548.json.gz"}Slowest frames (largest slices inside each, inclusive)
- present 2158: 294.3 ms at 2515 ms: Driver.Packet (queue worker 0x56) 294.1, Driver.WorkerWait (queue worker 0x52) 293.5, Driver.WorkerWait (queue worker 0x50) 293.4, Driver.Packet (queue worker 0x39) 293.2
- present 2181: 283.4 ms at 8256 ms: Driver.Packet (queue worker 0x56) 283.3, Driver.WorkerWait (queue worker 0x52) 282.9, Driver.WorkerWait (queue worker 0x50) 282.9, Driver.Packet (queue worker 0x39) 282.7
- present 2159: 282.5 ms at 2809 ms: Driver.Packet (queue worker 0x56) 282.3, Driver.Packet (queue worker 0x29) 281.7, Driver.WorkerWait (queue worker 0x50) 281.6, Driver.WorkerWait (queue worker 0x52) 281.5
- present 2199: 277.7 ms at 12751 ms: Driver.WorkerWait (queue worker 0x52) 277.7, Driver.Packet (queue worker 0x56) 277.5, Driver.WorkerWait (queue worker 0x50) 277.2, Driver.Packet (queue worker 0x39) 277.0
- present 2167: 276.5 ms at 4761 ms: Driver.Packet (queue worker 0x56) 276.4, Driver.WorkerWait (queue worker 0x50) 275.8, Driver.WorkerWait (queue worker 0x52) 275.6, Driver.Packet.memory_wait (queue worker 0x56) 274.2
Threads (ms per frame)
| thread | busy | wait | untraced | in slices < 20 us | largest |
|---|---|---|---|---|---|
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| queue worker 0x0 | 96.8 (39 %) | 67.5 (27 %) | 85.8 | 110.2 | Driver.Packet 115.4, Driver.Rewind 0.2 |
| title | 0.1 (0 %) | 0.0 (0 %) | 250.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |
| presenter | 0.4 (0 %) | 0.2 (0 %) | 249.6 | 0.1 | VideoOut.Flip 0.6 |
| draw committer | 41.7 (17 %) | 0.9 (0 %) | 207.7 | 25.5 | Graphics.Draw 30.9, Recorder.Sync 10.1 |
| queue worker 0x43 | 62.4 (25 %) | 186.4 (75 %) | 1.3 | 3.2 | Driver.Packet 192.2, Driver.EndSubmission 0.1 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 250.2 | 0.0 | Vulkan.ComputePipeline 0.0 |
| queue worker 0x21 | 13.1 (5 %) | 233.0 (93 %) | 4.1 | 6.2 | Driver.Packet 28.2, Driver.EndSubmission 0.5 |
| queue worker 0x50 | 0.4 (0 %) | 245.6 (98 %) | 4.2 | 0.3 | Driver.Packet 0.5, Driver.EndSubmission 0.1 |
| queue worker 0x52 | 0.4 (0 %) | 245.7 (98 %) | 4.2 | 0.3 | Driver.Packet 0.4, Driver.EndSubmission 0.0 |
| queue worker 0x29 | 13.2 (5 %) | 232.9 (93 %) | 4.1 | 6.2 | Driver.Packet 41.7, Driver.EndSubmission 0.5 |
| queue worker 0x56 | 1.1 (0 %) | 246.0 (98 %) | 3.1 | 0.8 | Driver.Packet 247.1, Driver.EndSubmission 0.0 |
| queue worker 0x41 | 12.0 (5 %) | 234.0 (94 %) | 4.1 | 5.6 | Driver.Packet 95.8, Driver.EndSubmission 0.5 |
| queue worker 0x39 | 13.1 (5 %) | 233.0 (93 %) | 4.1 | 6.3 | Driver.Packet 91.9, Driver.EndSubmission 0.6 |
| queue worker 0x49 | 10.2 (4 %) | 236.0 (94 %) | 4.0 | 4.3 | Driver.Packet 83.3, Driver.EndSubmission 0.4 |
| queue worker 0x31 | 13.7 (5 %) | 232.4 (93 %) | 4.1 | 6.2 | Driver.Packet 92.0, Driver.EndSubmission 0.5 |
| gpu | 98.6 (39 %) | 0.0 (0 %) | 151.6 | 0.0 | batch 98.6 |
Largest slices (ms per frame, inclusive, all threads): Driver.WorkerWait 1636.7, Driver.WorkerWait.dequeue_order 1596.8, Driver.Packet 988.6, Driver.Packet.memory_wait 675.7, Driver.Packet.dispatch 232.1, Driver.Dispatch 196.4, batch 98.6, Driver.Packet.draw 68.1, Driver.Draw 67.9, GpuMutex.wait.dispatch 66.1
Longest waits
- Driver.WorkerWait (queue worker 0x52) 323.5 ms at 12705 ms, present 2198
- Driver.WorkerWait (queue worker 0x50) 312.3 ms at 14451 ms, present 2205
- Driver.WorkerWait (queue worker 0x50) 309.8 ms at 11440 ms, present 2193
- Driver.Packet.memory_wait (queue worker 0x56) 300.3 ms at 2617 ms, present 2158
- Driver.Packet.memory_wait (queue worker 0x56) 300.1 ms at 8319 ms, present 2181
- Driver.WorkerWait (queue worker 0x50) 298.9 ms at 1718 ms, present 2154
- Driver.Packet.memory_wait (queue worker 0x56) 296.6 ms at 12815 ms, present 2199
- Driver.WorkerWait (queue worker 0x52) 295.7 ms at 5984 ms, present 2171
- Driver.WorkerWait (queue worker 0x52) 294.8 ms at 9240 ms, present 2184
- Driver.WorkerWait (queue worker 0x50) 293.9 ms at 9732 ms, present 2186
Largest GPU batches (GPU ranges inside, and the CPU slices that recorded each, inclusive)
- batch 82529: 5.9 ms at 2935 ms, present 2159: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.0, Driver.Packet.dispatch (queue worker 0x0) 5.4, Driver.Dispatch (queue worker 0x0) 5.4, Vulkan.DispatchRecipe (queue worker 0x0) 5.3
- batch 82644: 5.8 ms at 3153 ms, present 2160: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.0, Driver.Packet.dispatch (queue worker 0x0) 5.4, Driver.Dispatch (queue worker 0x0) 5.4, Vulkan.DispatchRecipe (queue worker 0x0) 5.3
- batch 85347: 5.8 ms at 8636 ms, present 2182: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 6.9, Driver.Packet.dispatch (queue worker 0x0) 5.5, Driver.Dispatch (queue worker 0x0) 5.5, Vulkan.DispatchRecipe (queue worker 0x0) 5.4
- batch 82406: 5.8 ms at 2633 ms, present 2158: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.1, Driver.Packet.dispatch (queue worker 0x0) 5.0, Driver.Dispatch (queue worker 0x0) 5.0, Vulkan.DispatchRecipe (queue worker 0x0) 4.8
- batch 82286: 5.8 ms at 2359 ms, present 2157: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 6.8, Driver.Packet.dispatch (queue worker 0x0) 5.4, Driver.Dispatch (queue worker 0x0) 5.4, Vulkan.DispatchRecipe (queue worker 0x0) 5.3
@MarcZX10 on the test: guest_thread_host_affinity runs on the i9-14900KF above. It passes, and it fails with main's Thread.cpp.
AI-assisted: yes
Contributor
Re-checked — changes still requestedThe paired captures, traces, and What I checked
Results
Negative controlYour comment reports FindingsBlocking
Non-blocking
Sources
Not verifiedI did not run the game, rebuild either capture, or independently run To get this to a good statePlease provide a capture and summary line from a clean/current |
APS5_GUEST_AFFINITY=1 pins every guest thread to the efficiency cores of a hybrid CPU (APS5_GUEST_AFFINITY_MASK=<hex> picks the cores), leaving the performance cores the queue workers are pinned to free of the title's spinning job workers.
XairRrR
force-pushed
the
perf/libkernel-guest-thread-affinity
branch
from
October 29, 2026 01:20
d74683c to
fda3f6c
Compare
Contributor
Author
|
@MarcZX10 @adriadam10 the description now holds captures from clean builds, as asked:
Both are from the same machine and the same prison yard fight, with The windows differ (195:15 and 115:15) because
AI-assisted: yes |
github-actions
Bot
removed
the
perf numbers missing
MarcZX10
suggested changes
Oct 29, 2026
MarcZX10
left a comment
MarcZX10
left a comment
Contributor
There was a problem hiding this comment.
Re-checked — changes requested for the performance report
The code branch is now based on current main, and the new captures match their summary lines, but both captures use extra APS5_TRACE_* switches excluded by the required performance protocol.
What I ran
- Reviewed exact head
fda3f6ca6c0cbf3370ecd5fb580312f707ecc082againstmain3a2b91981651569163768854a1fcdc7193b6667b. git diff --check upstream/main...refs/remotes/review/12544.git merge-tree --write-tree upstream/main refs/remotes/review/12544.python3 tools/check_conventions.py --base upstream/main --head refs/remotes/review/12544.- Downloaded the two attached trace assets and ran
python3 tools/trace_digest.py <trace> --frames 2 --top 3, matching each provided summary line to its trace.
Results
| Check | Result |
|---|---|
| Current-main merge preview | Clean; tree f0e052678e77fe66adfbceccab77b3f6ba4ce16c |
| Diff check | Passed; no output |
| Conventions | Passed |
| Main trace digest | 54 presents; p50 278.5 ms; summary matches the attached trace; trace reports dropped events |
| Change trace digest | 62 presents; p50 243.6 ms; summary matches the attached trace; trace reports dropped events |
| Updated-head focused tests | 56/56 for guest_thread_host_affinity and guest_pthread*, author-reported; not rerun locally |
| Hosted checks at review time | Windows build pending; Linux product and specialized jobs skipped; conventions, labels, eligibility, and report passed |
Raw trace-digest output excerpts
**seconds 195:15** from present 2391: 54 presents, frame p50 278.5 / p90 313.4 / p99 342.6 / max 342.6 ms, GPU busy 93.2 ms per present
NVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit 3a2b91981651, dropped: yes
**seconds 115:15** from present 2238: 62 presents, frame p50 243.6 / p90 264.0 / p99 335.2 / max 335.2 ms, GPU busy 89.7 ms per present
NVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit fda3f6ca6c0c, dropped: yes
Negative control
No independent negative control was run on this rebased head. My prior local run on the earlier head built the test executable but stopped at the project CPU-feature guard before its assertions.
Findings
Blocking
- Both summary lines set
APS5_TRACE_SUMMARY_EVERY=10andAPS5_TRACE_DIR=<path>. The repository’s Performance report form requires no additionalAPS5_TRACE_*orAPS5_PROFILE_*switches. Please recapturemainand this head with only the required capture setting, then replace the summaries, traces, and digests. Both current digests also report dropped trace events; please check whether the prescribed recapture avoids that loss.
Non-blocking
- The rebase brings the branch directly onto current
mainand preserves the single-topic diff. I independently checked the trace files and summary records, but did not rerun the affinity CTests on this host. - The reported single-run p50 improves from 278.5 ms to 243.6 ms; I am not treating that number as independently reproduced.
Sources
- Exact PR head diff
- Performance report capture protocol at current main
- Contribution rules at current main
- Main trace
- Change trace
Not verified
- Affinity assertions on this host or on a hybrid CPU.
- PS5 behavior.
- The updated hosted Windows build, which was pending when checked.
- The reported title run independently.
To get this to a good state
Recapture both builds with the documented capture settings and no extra trace/profile switches, include the refreshed summary lines, traces, and digests, and confirm the digests no longer report dropped events if possible. I will recheck the updated report and hosted build on the new head.
Contributor
Author
|
@MarcZX10 recaptured as the form describes, and the description holds the new pair:
The summary lines, traces and digests in the description are replaced. One point I can't meet literally: the launch script writes the log to a file in both runs. The logger reports 0.2 to 4.4 KiB per 10 s there. AI-assisted: yes |
MarcZX10
suggested changes
Oct 29, 2026
MarcZX10
left a comment
MarcZX10
left a comment
Contributor
There was a problem hiding this comment.
Re-checked — changes requested
I reviewed the performance report at exact head fda3f6ca6c0cbf3370ecd5fb580312f707ecc082. The new main-versus-change pair fixes the commit-provenance problem from my prior review, and both published digest sections reproduce exactly. Two capture-form requirements remain unmet.
What I ran
On Linux, I downloaded the two trace assets linked by the PR and ran the trace_digest.py version from current main (3a2b91981651569163768854a1fcdc7193b6667b) with each matching summary line extracted from the PR description:
python3 tools/trace_digest.py aps5-trace-3a2b91981651-1495087.json.gz --summary main-summary.jsonl
python3 tools/trace_digest.py aps5-trace-fda3f6ca6c0c-1527350.json.gz --summary pr-summary.jsonlI also checked the current performance-report form and hosted checks. I did not rebuild the project or rerun the game or unit tests.
Results
| Check | Result |
|---|---|
| Baseline/candidate provenance | Summary lines identify clean main 3a2b91981651 and clean PR commit fda3f6ca6c0c |
| Capture match | Both summaries report the same title, scene, machine, GPU, driver, resolution, and capture window; the reported environment difference is APS5_GUEST_AFFINITY |
| Trace integrity/digest | Both traces report no dropped events; both independently generated digest outputs exactly match the PR's posted outputs |
| Trace attachment | The two files are linked from a separate fork release; they are not attached to this PR as the form requests |
| Capture logging | The PR says the launch script writes the log to a file, contrary to the form's instruction not to redirect it |
| Hosted checks | Windows, conventions, labels, eligibility, and report pass; Linux and specialized jobs are skipped |
| Focused tests | The description reports 56/56 guest_thread_host_affinity and guest_pthread* tests on the rebased head; I did not rerun them |
Negative control
The main capture is the requested baseline: 28 presents and 282.4 ms frame-time p50. The candidate reports 32 presents and 250.1 ms p50. This review verifies the supplied trace and summary artifacts, not an independently repeated performance run.
Findings
Blocking
- The traces are not attached to the PR. The current form asks for each
aps5-trace-<commit>-<pid>.json.gzto be attached in its Traces field. The description instead links the assets from a separate fork release. Please attach both files to this PR. - The capture log was redirected. The form explicitly says not to redirect the log because it costs time, and the report says the launcher wrote it to a file for both captures. Please recapture without redirecting the log, then replace the affected summary lines, trace attachments, and digest output. If maintainers want to accept an exception, that decision should be recorded before treating this as a compliant performance report.
Non-blocking
- The actual paired data is internally consistent: the exact trace digests match the posted output, and neither trace reports dropped events.
- The hosted Linux and specialized checks were skipped; the Windows workflow passed. The reported focused test result is useful but was not independently rerun here.
Sources
- Performance capture form at current
main - PR report and summary lines at the reviewed head
- Main trace asset
- Candidate trace asset
trace_digest.pyat the baseline commit
Not verified
I did not build the PR, rerun its CTest targets, or reproduce the game captures. Linux and specialized hosted jobs were skipped.
To get this to a good state
Attach both traces directly to the PR and provide a fresh pair captured without log redirection, with matching summary lines and newly generated digests. Then I can recheck the report against the same requirements.
Contributor
|
{"body": "# Re-checked \u2014 new head, changes requested\n\nThis evidence addendum is for exact PR head Raw trace-digest output\n\n~~~text\nseconds 195:8 from present 2410: 28 presents, frame p50 282.4 / p90 308.3 / p99 359.7 / max 359.7 ms, GPU busy 104.7 ms per present\nNVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit 3a2b919, dropped: no\n\njson\n{\"commit\":\"3a2b91981651\",\"dirty\":false,\"env\":{\"APS5_DISPATCH_CACHE_ENTRIES\":\"4096\",\"APS5_DUMP_FRAMES\":\"0\",\"APS5_DUMP_FRAMES_EVERY\":\"99999\",\"APS5_HOST_IMPORT_MIB\":\"16384\",\"APS5_IMAGE_TABLE_KEY_SCAN\":\"262144\",\"APS5_LAZY_SHADER_PREPARE\":\"1\",\"APS5_LOCAL_NULL_INVALID_BUFFERS\":\"1\",\"APS5_LOCAL_REBASE_HUGE_ARRAYS\":\"1\",\"APS5_NO_LIBRARY_PREBUILD\":\"1\",\"APS5_NO_PIPELINE_LTO\":\"1\",\"APS5_NO_SNAPSHOT_CHECK\":\"1\",\"APS5_PAD_INPUT_FILE\":\"<path>\",\"APS5_TRACE_SECONDS\":\"195:8\"},\"gpu\":\"NVIDIA GeForce RTX 4070 Ti SUPER\",\"driver\":\"NVIDIA 615.78.08\",\"os\":\"Linux 7.2.11-zen1-1-zen\",\"cpu\":\"Intel(R) Core(TM) i9-14900KF\",\"cpu_threads\":32,\"memory_mib\":31906,\"title_id\":\"PPSA05684\",\"title\":\"Uncharted\\u2122: Legacy of Thieves Collection\",\"content_version\":\"01.002.000\",\"resolution\":\"2560x1440\",\"window\":{\"spec\":\"seconds 195:8\",\"first_present\":2410,\"ms\":8012.792},\"presents\":28,\"frame_ms\":{\"p50\":282.42,\"p90\":308.327,\"p99\":359.737,\"max\":359.737},\"gpu_busy_ms_per_present\":104.734,\"worker_busy_pct\":{\"queue worker 0x0\":72.598,\"queue worker 0x43\":77.285,\"queue worker 0x21\":12.541,\"queue worker 0x50\":0.192,\"queue worker 0x52\":0.244,\"queue worker 0x29\":20.896,\"queue worker 0x56\":96.338,\"queue worker 0x31\":36.244,\"queue worker 0x39\":45.045,\"queue worker 0x49\":19.195,\"queue worker 0x41\":35.466},\"top_phases_ms_per_frame\":{\"Driver.Packet\":1186.075,\"Driver.Packet.memory_wait\":758.571,\"Driver.Packet.dispatch\":276.247,\"Driver.Dispatch\":244.039,\"Driver.Packet.draw\":118.164,\"Driver.Draw\":117.662,\"Graphics.Draw\":96.348,\"Graphics.Draw.draw_and_resource_release\":92.135},\"trace_dropped\":false,\"trace\":\"aps5-trace-3a2b91981651-1495087.json.gz\"}\n\n\nSlowest frames (largest slices inside each, inclusive)\n\n- present 2412: 359.7 ms at 561 ms: Driver.Packet (queue worker 0x56) 359.5, Driver.WorkerWait (queue worker 0x52) 359.2, Driver.WorkerWait (queue worker 0x50) 358.7, Driver.Packet.memory_wait (queue worker 0x56) 357.5\n- present 2431: 342.6 ms at 5988 ms: Driver.WorkerWait (queue worker 0x52) 342.6, Driver.Packet (queue worker 0x56) 342.4, Driver.Packet (queue worker 0x31) 342.0, Driver.WorkerWait (queue worker 0x50) 341.9\n- present 2417: 308.3 ms at 2009 ms: Driver.Packet (queue worker 0x56) 308.2, Driver.WorkerWait (queue worker 0x52) 307.7, Driver.WorkerWait (queue worker 0x50) 307.6, Driver.Packet (queue worker 0x39) 307.4\n- present 2426: 305.1 ms at 4539 ms: Driver.Packet (queue worker 0x56) 304.9, Driver.WorkerWait (queue worker 0x52) 304.6, Driver.WorkerWait (queue worker 0x50) 304.5, Driver.Packet (queue worker 0x41) 304.4\n- present 2432: 300.0 ms at 6330 ms: Driver.Packet (queue worker 0x56) 299.8, Driver.WorkerWait (queue worker 0x52) 299.3, Driver.Packet (queue worker 0x39) 299.2, Driver.WorkerWait (queue worker 0x50) 298.5\n\nThreads (ms per frame)\n\n| thread | busy | wait | untraced | in slices < 20 us | largest |\n|---|---|---|---|---|---|\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.0 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.0 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| queue worker 0x0 | 140.2 (49 %) | 114.5 (40 %) | 31.4 | 158.5 | Driver.Packet 179.0, Driver.Rewind 0.3 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 286.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| draw committer | 45.4 (16 %) | 0.7 (0 %) | 240.1 | 25.8 | Graphics.Draw 35.0, Recorder.Sync 9.6 |\n| presenter | 0.7 (0 %) | 0.5 (0 %) | 284.9 | 0.1 | VideoOut.Flip 1.1, Driver.Present 0.2 |\n| queue worker 0x43 | 79.8 (28 %) | 203.3 (71 %) | 3.1 | 3.3 | Driver.Packet 220.0, Driver.EndSubmission 0.3 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 286.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| queue worker 0x21 | 14.7 (5 %) | 268.9 (94 %) | 2.5 | 6.2 | Driver.Packet 35.0, Driver.EndSubmission 0.5 |\n| queue worker 0x50 | 0.4 (0 %) | 275.9 (96 %) | 9.9 | 0.2 | Driver.Packet 0.4, Driver.EndSubmission 0.0 |\n| queue worker 0x52 | 0.4 (0 %) | 275.9 (96 %) | 9.9 | 0.2 | Driver.Packet 0.6 |\n| queue worker 0x29 | 15.0 (5 %) | 261.2 (91 %) | 10.0 | 6.2 | Driver.Packet 58.9, Driver.EndSubmission 0.5 |\n| queue worker 0x56 | 1.2 (0 %) | 274.3 (96 %) | 10.6 | 0.9 | Driver.Packet 275.5, Driver.EndSubmission 0.0 |\n| queue worker 0x31 | 15.2 (5 %) | 261.3 (91 %) | 9.7 | 6.1 | Driver.Packet 102.5, Driver.EndSubmission 0.8 |\n| queue worker 0x39 | 14.7 (5 %) | 262.2 (92 %) | 9.3 | 6.1 | Driver.Packet 127.9, Driver.EndSubmission 0.7 |\n| queue worker 0x49 | 11.2 (4 %) | 265.5 (93 %) | 9.5 | 4.4 | Driver.Packet 54.1, Driver.EndSubmission 0.5 |\n| queue worker 0x41 | 14.4 (5 %) | 269.3 (94 %) | 2.5 | 6.0 | Driver.Packet 100.6, Driver.EndSubmission 0.5 |\n| gpu | 104.5 (37 %) | 0.0 (0 %) | 181.7 | 0.0 | batch 104.5 |\n\nLargest slices (ms per frame, inclusive, all threads): Driver.WorkerWait 1880.2, Driver.WorkerWait.dequeue_order 1848.5, Driver.Packet 1154.7, Driver.Packet.memory_wait 758.2, Driver.Packet.dispatch 274.9, Driver.Dispatch 237.3, batch 104.5, Driver.Packet.draw 102.6, Driver.Draw 102.3, Graphics.Draw 84.3\n\nLongest waits\n\n- Driver.WorkerWait (queue worker 0x52) 396.9 ms at 5933 ms, present 2430\n- Driver.WorkerWait (queue worker 0x52) 359.3 ms at 561 ms, present 2412\n- Driver.Packet.memory_wait (queue worker 0x56) 355.7 ms at 363 ms, present 2411\n- Driver.WorkerWait (queue worker 0x50) 341.9 ms at 5988 ms, present 2431\n- Driver.WorkerWait (queue worker 0x49) 334.7 ms at 586 ms, present 2412\n- Driver.Packet.memory_wait (queue worker 0x56) 330.8 ms at 6073 ms, present 2431\n- Driver.WorkerWait (queue worker 0x50) 330.5 ms at 874 ms, present 2412\n- Driver.WorkerWait (queue worker 0x50) 325.4 ms at 1429 ms, present 2414\n- Driver.WorkerWait (queue worker 0x29) 322.7 ms at 598 ms, present 2412\n- Driver.Packet.memory_wait (queue worker 0x41) 317.4 ms at 6026 ms, present 2431\n\nLargest GPU batches (GPU ranges inside, and the CPU slices that recorded each, inclusive)\n\n- batch 662: 6.4 ms at 1521 ms, present 2415: GPU no ranges; recorded by Driver.WorkerWait (queue worker 0x0) 29.6, Driver.WorkerWait.dequeue_order (queue worker 0x0) 29.6, Driver.Packet (queue worker 0x0) 2.6, Driver.Packet.pipeline_drain (queue worker 0x0) 1.3\n- batch 358: 6.1 ms at 863 ms, present 2412: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.4, Driver.Packet.draw (queue worker 0x0) 6.4, Driver.Draw (queue worker 0x0) 6.4, Driver.Draw.shader_compile_and_link (queue worker 0x0) 3.4\n- batch 2184: 6.1 ms at 4894 ms, present 2427: GPU no ranges; recorded by Recorder.Submit (draw committer) 2.0, Recorder.Reap (draw committer) 0.1, Graphics.Draw (draw committer) 0.0, Graphics.Draw.draw_and_resource_release (draw committer) 0.0\n- batch 3183: 6.0 ms at 7212 ms, present 2435: GPU no ranges; recorded by Driver.WorkerWait (queue worker 0x0) 44.8, Driver.WorkerWait.dequeue_order (queue worker 0x0) 44.8, Driver.Packet (queue worker 0x0) 2.9, Driver.Packet.draw (queue worker 0x0) 1.2\n- batch 2435: 6.0 ms at 5445 ms, present 2429: GPU no ranges; recorded by Driver.WorkerWait (queue worker 0x0) 46.0, Driver.WorkerWait.dequeue_order (queue worker 0x0) 46.0, Driver.Packet (queue worker 0x0) 1.9, Driver.Packet.pipeline_drain (queue worker 0x0) 0.8\n\n\nseconds 195:8 from present 2503: 32 presents, frame p50 250.1 / p90 278.1 / p99 302.5 / max 302.5 ms, GPU busy 93.9 ms per present\nNVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit fda3f6c, dropped: no\n\njson\n{\"commit\":\"fda3f6ca6c0c\",\"dirty\":false,\"env\":{\"APS5_DISPATCH_CACHE_ENTRIES\":\"4096\",\"APS5_DUMP_FRAMES\":\"0\",\"APS5_DUMP_FRAMES_EVERY\":\"99999\",\"APS5_GUEST_AFFINITY\":\"1\",\"APS5_HOST_IMPORT_MIB\":\"16384\",\"APS5_IMAGE_TABLE_KEY_SCAN\":\"262144\",\"APS5_LAZY_SHADER_PREPARE\":\"1\",\"APS5_LOCAL_NULL_INVALID_BUFFERS\":\"1\",\"APS5_LOCAL_REBASE_HUGE_ARRAYS\":\"1\",\"APS5_NO_LIBRARY_PREBUILD\":\"1\",\"APS5_NO_PIPELINE_LTO\":\"1\",\"APS5_NO_SNAPSHOT_CHECK\":\"1\",\"APS5_PAD_INPUT_FILE\":\"<path>\",\"APS5_TRACE_SECONDS\":\"195:8\"},\"gpu\":\"NVIDIA GeForce RTX 4070 Ti SUPER\",\"driver\":\"NVIDIA 615.78.08\",\"os\":\"Linux 7.2.11-zen1-1-zen\",\"cpu\":\"Intel(R) Core(TM) i9-14900KF\",\"cpu_threads\":32,\"memory_mib\":31906,\"title_id\":\"PPSA05684\",\"title\":\"Uncharted\\u2122: Legacy of Thieves Collection\",\"content_version\":\"01.002.000\",\"resolution\":\"2560x1440\",\"window\":{\"spec\":\"seconds 195:8\",\"first_present\":2503,\"ms\":8103.759},\"presents\":32,\"frame_ms\":{\"p50\":250.076,\"p90\":278.09,\"p99\":302.491,\"max\":302.491},\"gpu_busy_ms_per_present\":93.863,\"worker_busy_pct\":{\"queue worker 0x0\":73.316,\"queue worker 0x43\":77.34,\"queue worker 0x21\":10.509,\"queue worker 0x52\":0.25,\"queue worker 0x50\":0.187,\"queue worker 0x29\":9.595,\"queue worker 0x56\":96.463,\"queue worker 0x49\":35.609,\"queue worker 0x41\":32.881,\"queue worker 0x31\":14.157,\"queue worker 0x39\":43.454},\"top_phases_ms_per_frame\":{\"Driver.Packet\":993.112,\"Driver.Packet.memory_wait\":620.976,\"Driver.Packet.dispatch\":235.123,\"Driver.Dispatch\":205.303,\"Driver.Packet.draw\":107.265,\"Driver.Draw\":106.82,\"Graphics.Draw\":86.683,\"Graphics.Draw.draw_and_resource_release\":82.868},\"trace_dropped\":false,\"trace\":\"aps5-trace-fda3f6ca6c0c-1527350.json.gz\"}\n\n\nSlowest frames (largest slices inside each, inclusive)\n\n- present 2534: 302.5 ms at 7801 ms: Driver.Packet (queue worker 0x39) 230.9, Driver.Packet (queue worker 0x43) 224.8, Driver.Packet (queue worker 0x0) 191.6, Driver.WorkerWait (queue worker 0x21) 191.3\n- present 2503: 291.2 ms at 0 ms: Driver.WorkerWait (queue worker 0x52) 290.5, Driver.WorkerWait (queue worker 0x50) 290.5, Driver.WorkerWait.dequeue_order (queue worker 0x50) 281.6, Driver.WorkerWait.dequeue_order (queue worker 0x52) 281.4\n- present 2520: 282.3 ms at 4240 ms: Driver.Packet (queue worker 0x56) 282.1, Driver.Packet (queue worker 0x49) 281.8, Driver.WorkerWait (queue worker 0x52) 281.8, Driver.WorkerWait (queue worker 0x50) 281.8\n- present 2532: 278.1 ms at 7263 ms: Driver.Packet (queue worker 0x56) 278.0, Driver.Packet (queue worker 0x49) 277.9, Driver.WorkerWait (queue worker 0x52) 277.5, Driver.WorkerWait (queue worker 0x50) 277.5\n- present 2531: 273.8 ms at 6989 ms: Driver.Packet (queue worker 0x56) 273.6, Driver.WorkerWait (queue worker 0x50) 273.3, Driver.WorkerWait (queue worker 0x52) 273.2, Driver.Packet.memory_wait (queue worker 0x56) 271.8\n\nThreads (ms per frame)\n\n| thread | busy | wait | untraced | in slices < 20 us | largest |\n|---|---|---|---|---|---|\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.1 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| queue worker 0x0 | 127.3 (50 %) | 97.0 (38 %) | 28.9 | 152.3 | Driver.Packet 159.1, Driver.Rewind 0.3 |\n| title | 0.1 (0 %) | 0.0 (0 %) | 253.2 | 0.1 | Shader.ResolveGraphicsStagesAbi 0.1 |\n| presenter | 0.4 (0 %) | 0.2 (0 %) | 252.7 | 0.1 | VideoOut.Flip 0.6, Driver.Present 0.0 |\n| draw committer | 39.8 (16 %) | 0.6 (0 %) | 212.9 | 25.1 | Graphics.Draw 30.4, Recorder.Sync 8.6 |\n| queue worker 0x43 | 68.1 (27 %) | 182.1 (72 %) | 3.1 | 3.3 | Driver.Packet 194.9, Driver.EndSubmission 0.1 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| shader prepare | 0.0 (0 %) | 0.0 (0 %) | 253.2 | 0.0 | Vulkan.ComputePipeline 0.0 |\n| queue worker 0x21 | 12.8 (5 %) | 238.0 (94 %) | 2.4 | 6.1 | Driver.Packet 25.5, Driver.EndSubmission 0.7 |\n| queue worker 0x52 | 0.4 (0 %) | 244.5 (97 %) | 8.3 | 0.2 | Driver.Packet 0.6, Driver.EndSubmission 0.0 |\n| queue worker 0x50 | 0.3 (0 %) | 244.6 (97 %) | 8.3 | 0.3 | Driver.Packet 0.3, Driver.EndSubmission 0.1 |\n| queue worker 0x29 | 12.5 (5 %) | 232.1 (92 %) | 8.6 | 6.3 | Driver.Packet 23.4, Driver.EndSubmission 0.5 |\n| queue worker 0x56 | 1.1 (0 %) | 243.1 (96 %) | 9.1 | 0.8 | Driver.Packet 244.1, Driver.EndSubmission 0.0 |\n| queue worker 0x49 | 10.5 (4 %) | 234.3 (93 %) | 8.5 | 4.6 | Driver.Packet 89.4, Driver.EndSubmission 0.5 |\n| queue worker 0x41 | 12.0 (5 %) | 232.7 (92 %) | 8.6 | 5.9 | Driver.Packet 82.6, Driver.EndSubmission 0.3 |\n| queue worker 0x31 | 12.7 (5 %) | 232.1 (92 %) | 8.4 | 6.3 | Driver.Packet 34.9, Driver.EndSubmission 0.6 |\n| queue worker 0x39 | 13.5 (5 %) | 237.2 (94 %) | 2.6 | 6.3 | Driver.Packet 109.1, Driver.EndSubmission 0.6 |\n| gpu | 93.9 (37 %) | 0.0 (0 %) | 159.4 | 0.0 | batch 93.9 |\n\nLargest slices (ms per frame, inclusive, all threads): Driver.WorkerWait 1718.2, Driver.WorkerWait.dequeue_order 1684.3, Driver.Packet 963.9, Driver.Packet.memory_wait 620.6, Driver.Packet.dispatch 233.7, Driver.Dispatch 198.5, batch 93.9, Driver.Packet.draw 92.8, Driver.Draw 92.6, Graphics.Draw 74.4\n\nLongest waits\n\n- Driver.WorkerWait (queue worker 0x50) 315.4 ms at 3193 ms, present 2515\n- Driver.WorkerWait (queue worker 0x52) 293.4 ms at 994 ms, present 2506\n- Driver.Packet.memory_wait (queue worker 0x56) 291.0 ms at 4307 ms, present 2520\n- Driver.WorkerWait (queue worker 0x52) 290.5 ms at 1 ms, present 2503\n- Driver.WorkerWait (queue worker 0x50) 290.5 ms at 1 ms, present 2503\n- Driver.Packet.memory_wait (queue worker 0x56) 288.1 ms at 7062 ms, present 2531\n- Driver.WorkerWait (queue worker 0x52) 281.8 ms at 4241 ms, present 2520\n- Driver.WorkerWait (queue worker 0x50) 281.8 ms at 4241 ms, present 2520\n- Driver.Packet.memory_wait (queue worker 0x56) 279.9 ms at 78 ms, present 2503\n- Driver.WorkerWait (queue worker 0x52) 277.5 ms at 7263 ms, present 2532\n\nLargest GPU batches (GPU ranges inside, and the CPU slices that recorded each, inclusive)\n\n- batch 218: 5.5 ms at 482 ms, present 2504: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 5.7, Driver.Packet.draw (queue worker 0x0) 5.1, Driver.Draw (queue worker 0x0) 5.0, Driver.Draw.shader_compile_and_link (queue worker 0x0) 2.4\n- batch 515: 5.4 ms at 1091 ms, present 2507: GPU no ranges; recorded by no CPU slices\n- batch 1408: 5.2 ms at 2940 ms, present 2514: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 5.7, Driver.Packet.draw (queue worker 0x0) 5.1, Driver.Draw (queue worker 0x0) 5.1, Driver.Draw.shader_compile_and_link (queue worker 0x0) 2.6\n- batch 266: 5.1 ms at 569 ms, present 2505: GPU no ranges; recorded by no CPU slices\n- batch 1904: 5.1 ms at 3990 ms, present 2519: GPU no ranges; recorded by Driver.Packet (queue worker 0x39) 9.5, Driver.Packet.dispatch (queue worker 0x39) 9.5, Driver.Dispatch (queue worker 0x39) 9.0, GpuMutex.wait.dispatch (queue worker 0x39) 7.2\n~~~\n\nRaw hosted-check output\n\n~~~text\nconventions\tpass\t14s\thttps://github.com/boykopovar/AnyPS5/actions/runs/40806002604/job/123224999334\nlabels\tpass\t7s\thttps://github.com/boykopovar/AnyPS5/actions/runs/40806002950/job/123224999579\nwindows\tpass\t18m48s\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204892251\nlinux\tskipping\t0\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204894774\nregister-context\tskipping\t0\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204894024\nrelinker\tskipping\t0\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204893949\nrelinker-asan\tskipping\t0\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204894048\nwrite-watch\tskipping\t0\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204894216\neligibility\tpass\t4s\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786629/job/123204856468\nreport\tpass\t49s\thttps://github.com/boykopovar/AnyPS5/actions/runs/40799786707/job/123204856542\n~~~\n\nRaw attachment/log inspection output\n\n~~~text\nattachment URLs: 0\nlog-to-file statement in PR body: True\n~~~\n\n3a2b91981651569163768854a1fcdc7193b6667b\n- Trace-digest code at the baseline commit\n- PR report at reviewed head fda3f6ca6c0cbf3370ecd5fb580312f707ecc082\n- Main trace asset\n- Candidate trace asset\n- Author\u2019s note about log output\n\n## Not verified\n\nI did not build the PR, run its CTest targets, reproduce the game captures, or independently verify the reported 56 test executions. Linux and specialized hosted jobs are skipped.\n\n## To get this to a good state\n\nAttach both trace files to the PR, recapture without log redirection, and update the summary lines and digest output. I will rerun the same digest commands and recheck hosted status on the new report.\n"}
|
12 tasks done
Merged
12 tasks done
Contributor
|
Note 🤖 Claude Opus 5.5 responding on behalf of Mouhand-Kaddo Another title with the same pattern, measured on Linux: moving the spinning job workers to the efficiency cores makes Demon's Souls (PPSA01342) about 15 % faster in the world. This is not a run of this branch. It is the same idea applied locally on Setup: Intel Core i7-13700K (8 performance cores with 2 threads each, 8 efficiency cores), AMD Radeon RX 6900 XT, RADV Mesa 26.2.6, Linux 7.2.14, Release build. The title was relinked with Why it helps here: on
The live check: in a The Windows comment on |
Contributor
|
Note 🤖 Claude Opus 5.5 responding on behalf of Mouhand-Kaddo Follow-up with this branch itself: fda3f6c cherry-picked onto
|
# Conflicts: # docs/dev/ENVIRONMENT.md
Collaborator
|
Hi adriadam10, Your change request from 10-28 (at
|
Contributor
✅ AMD GPU test: passedPR head
Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (13, not counted against this PR)Validation IDs already present on main (4 IDs in 767 tests)Skipped (60)ReproduceAutomated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
12 tasks done
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
On a hybrid CPU the queue workers and the presenter run on the performance cores by default on Linux (#626). Guest threads have no host affinity: the affinity a title sets is kept for its getters but never applied to the host thread. A title whose job workers spin at full load on every core they own therefore puts those spinners on the same performance cores, often on the second hardware thread of a core a queue worker runs on.
APS5_GUEST_AFFINITY=1now pins every guest thread to the efficiency cores of a hybrid CPU when it starts, on Linux and Windows.APS5_GUEST_AFFINITY_MASK=<hex>picks the cores instead. Both are off by default, likeAPS5_JOB_AFFINITY. That variable selects threads by one title's job worker name and is Windows only; the new one does not look at thread names.Tested
Arch Linux 7.2.11-zen, GCC 16.2.1, Ninja, Release, on
main3a2b919, Intel Core i9-14900KF (8 performance cores with 2 threads each, 16 efficiency cores):New
guest_thread_host_affinity(Linux): setsAPS5_GUEST_AFFINITY_MASKto one core of the process and checks that two guest threads run on exactly that core, and that the creating thread keeps its mask. WithThread.cppfrommainit fails:a guest thread does not run on the cores APS5_GUEST_AFFINITY_MASK names.Full
ctest -j8on the first base (main3b362ba): all 2741 tests pass (60 skipped). Rebased on 3a2b919:guest_thread_host_affinityand theguest_pthread*tests pass (56/56).tools/check_conventions.py --base origin/mainpasses.PPSA05684 (Uncharted: Legacy of Thieves Collection): its 13 job workers (
NdJobWorkerThread 0..12) spin withpauseat 100 % each. WithAPS5_GUEST_AFFINITY=1they run on the efficiency cores.The captures below follow the Performance report form. They come from clean builds of
main3a2b919 and of this branch, fda3f6c ("dirty": falsein both), captured on the same machine in the same session, with warm caches. Both capture the Uncharted 4 prison yard fight, reached through the developer menu by the same scripted inputs, withAPS5_TRACE_SECONDS=195:8and no otherAPS5_TRACE_*orAPS5_PROFILE_*switch. The environments differ only inAPS5_GUEST_AFFINITY. TheAPS5_LOCAL_*names come from my launch script and do not exist onmain.The 8 s window keeps queue 0 inside its default event buffer, so neither trace drops events. The launch script writes the log to a file in both runs; the logger reports 0.2 to 4.4 KiB per 10 s there.
main3a2b919APS5_GUEST_AFFINITY=1queue worker 0x0busyClean
maindoes not reach this scene every time. Over 7 attempts in this session on the two builds, 5 ended early:std::bad_alloc.vkWaitForFencesreturnedVK_ERROR_DEVICE_LOST.The oversized views and the device loss do not happen on the local stack.
The captures are from the runs that reached the fight.
Traces:
main, this branchSummary line,
main:{"commit":"3a2b91981651","dirty":false,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_SECONDS":"195:8"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted™: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 195:8","first_present":2410,"ms":8012.792},"presents":28,"frame_ms":{"p50":282.420,"p90":308.327,"p99":359.737,"max":359.737},"gpu_busy_ms_per_present":104.734,"worker_busy_pct":{"queue worker 0x0":72.598,"queue worker 0x43":77.285,"queue worker 0x21":12.541,"queue worker 0x50":0.192,"queue worker 0x52":0.244,"queue worker 0x29":20.896,"queue worker 0x56":96.338,"queue worker 0x31":36.244,"queue worker 0x39":45.045,"queue worker 0x49":19.195,"queue worker 0x41":35.466},"top_phases_ms_per_frame":{"Driver.Packet":1186.075,"Driver.Packet.memory_wait":758.571,"Driver.Packet.dispatch":276.247,"Driver.Dispatch":244.039,"Driver.Packet.draw":118.164,"Driver.Draw":117.662,"Graphics.Draw":96.348,"Graphics.Draw.draw_and_resource_release":92.135},"trace_dropped":false,"trace":"aps5-trace-3a2b91981651-1495087.json.gz"}Summary line, this branch:
{"commit":"fda3f6ca6c0c","dirty":false,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_GUEST_AFFINITY":"1","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_SECONDS":"195:8"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted™: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 195:8","first_present":2503,"ms":8103.759},"presents":32,"frame_ms":{"p50":250.076,"p90":278.090,"p99":302.491,"max":302.491},"gpu_busy_ms_per_present":93.863,"worker_busy_pct":{"queue worker 0x0":73.316,"queue worker 0x43":77.340,"queue worker 0x21":10.509,"queue worker 0x52":0.250,"queue worker 0x50":0.187,"queue worker 0x29":9.595,"queue worker 0x56":96.463,"queue worker 0x49":35.609,"queue worker 0x41":32.881,"queue worker 0x31":14.157,"queue worker 0x39":43.454},"top_phases_ms_per_frame":{"Driver.Packet":993.112,"Driver.Packet.memory_wait":620.976,"Driver.Packet.dispatch":235.123,"Driver.Dispatch":205.303,"Driver.Packet.draw":107.265,"Driver.Draw":106.820,"Graphics.Draw":86.683,"Graphics.Draw.draw_and_resource_release":82.868},"trace_dropped":false,"trace":"aps5-trace-fda3f6ca6c0c-1527350.json.gz"}tools/trace_digest.py,mainseconds 195:8 from present 2410: 28 presents, frame p50 282.4 / p90 308.3 / p99 359.7 / max 359.7 ms, GPU busy 104.7 ms per present
NVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit 3a2b919, dropped: no
{"commit":"3a2b91981651","dirty":false,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_SECONDS":"195:8"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted\u2122: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 195:8","first_present":2410,"ms":8012.792},"presents":28,"frame_ms":{"p50":282.42,"p90":308.327,"p99":359.737,"max":359.737},"gpu_busy_ms_per_present":104.734,"worker_busy_pct":{"queue worker 0x0":72.598,"queue worker 0x43":77.285,"queue worker 0x21":12.541,"queue worker 0x50":0.192,"queue worker 0x52":0.244,"queue worker 0x29":20.896,"queue worker 0x56":96.338,"queue worker 0x31":36.244,"queue worker 0x39":45.045,"queue worker 0x49":19.195,"queue worker 0x41":35.466},"top_phases_ms_per_frame":{"Driver.Packet":1186.075,"Driver.Packet.memory_wait":758.571,"Driver.Packet.dispatch":276.247,"Driver.Dispatch":244.039,"Driver.Packet.draw":118.164,"Driver.Draw":117.662,"Graphics.Draw":96.348,"Graphics.Draw.draw_and_resource_release":92.135},"trace_dropped":false,"trace":"aps5-trace-3a2b91981651-1495087.json.gz"}Slowest frames (largest slices inside each, inclusive)
Threads (ms per frame)
Largest slices (ms per frame, inclusive, all threads): Driver.WorkerWait 1880.2, Driver.WorkerWait.dequeue_order 1848.5, Driver.Packet 1154.7, Driver.Packet.memory_wait 758.2, Driver.Packet.dispatch 274.9, Driver.Dispatch 237.3, batch 104.5, Driver.Packet.draw 102.6, Driver.Draw 102.3, Graphics.Draw 84.3
Longest waits
Largest GPU batches (GPU ranges inside, and the CPU slices that recorded each, inclusive)
batch 662: 6.4 ms at 1521 ms, present 2415: GPU no ranges; recorded by Driver.WorkerWait (queue worker 0x0) 29.6, Driver.WorkerWait.dequeue_order (queue worker 0x0) 29.6, Driver.Packet (queue worker 0x0) 2.6, Driver.Packet.pipeline_drain (queue worker 0x0) 1.3
batch 358: 6.1 ms at 863 ms, present 2412: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 7.4, Driver.Packet.draw (queue worker 0x0) 6.4, Driver.Draw (queue worker 0x0) 6.4, Driver.Draw.shader_compile_and_link (queue worker 0x0) 3.4
batch 2184: 6.1 ms at 4894 ms, present 2427: GPU no ranges; recorded by Recorder.Submit (draw committer) 2.0, Recorder.Reap (draw committer) 0.1, Graphics.Draw (draw committer) 0.0, Graphics.Draw.draw_and_resource_release (draw committer) 0.0
batch 3183: 6.0 ms at 7212 ms, present 2435: GPU no ranges; recorded by Driver.WorkerWait (queue worker 0x0) 44.8, Driver.WorkerWait.dequeue_order (queue worker 0x0) 44.8, Driver.Packet (queue worker 0x0) 2.9, Driver.Packet.draw (queue worker 0x0) 1.2
batch 2435: 6.0 ms at 5445 ms, present 2429: GPU no ranges; recorded by Driver.WorkerWait (queue worker 0x0) 46.0, Driver.WorkerWait.dequeue_order (queue worker 0x0) 46.0, Driver.Packet (queue worker 0x0) 1.9, Driver.Packet.pipeline_drain (queue worker 0x0) 0.8
tools/trace_digest.py, this branchseconds 195:8 from present 2503: 32 presents, frame p50 250.1 / p90 278.1 / p99 302.5 / max 302.5 ms, GPU busy 93.9 ms per present
NVIDIA GeForce RTX 4070 Ti SUPER, NVIDIA 615.78.08, Linux 7.2.11-zen1-1-zen, 2560x1440, commit fda3f6c, dropped: no
{"commit":"fda3f6ca6c0c","dirty":false,"env":{"APS5_DISPATCH_CACHE_ENTRIES":"4096","APS5_DUMP_FRAMES":"0","APS5_DUMP_FRAMES_EVERY":"99999","APS5_GUEST_AFFINITY":"1","APS5_HOST_IMPORT_MIB":"16384","APS5_IMAGE_TABLE_KEY_SCAN":"262144","APS5_LAZY_SHADER_PREPARE":"1","APS5_LOCAL_NULL_INVALID_BUFFERS":"1","APS5_LOCAL_REBASE_HUGE_ARRAYS":"1","APS5_NO_LIBRARY_PREBUILD":"1","APS5_NO_PIPELINE_LTO":"1","APS5_NO_SNAPSHOT_CHECK":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_SECONDS":"195:8"},"gpu":"NVIDIA GeForce RTX 4070 Ti SUPER","driver":"NVIDIA 615.78.08","os":"Linux 7.2.11-zen1-1-zen","cpu":"Intel(R) Core(TM) i9-14900KF","cpu_threads":32,"memory_mib":31906,"title_id":"PPSA05684","title":"Uncharted\u2122: Legacy of Thieves Collection","content_version":"01.002.000","resolution":"2560x1440","window":{"spec":"seconds 195:8","first_present":2503,"ms":8103.759},"presents":32,"frame_ms":{"p50":250.076,"p90":278.09,"p99":302.491,"max":302.491},"gpu_busy_ms_per_present":93.863,"worker_busy_pct":{"queue worker 0x0":73.316,"queue worker 0x43":77.34,"queue worker 0x21":10.509,"queue worker 0x52":0.25,"queue worker 0x50":0.187,"queue worker 0x29":9.595,"queue worker 0x56":96.463,"queue worker 0x49":35.609,"queue worker 0x41":32.881,"queue worker 0x31":14.157,"queue worker 0x39":43.454},"top_phases_ms_per_frame":{"Driver.Packet":993.112,"Driver.Packet.memory_wait":620.976,"Driver.Packet.dispatch":235.123,"Driver.Dispatch":205.303,"Driver.Packet.draw":107.265,"Driver.Draw":106.82,"Graphics.Draw":86.683,"Graphics.Draw.draw_and_resource_release":82.868},"trace_dropped":false,"trace":"aps5-trace-fda3f6ca6c0c-1527350.json.gz"}Slowest frames (largest slices inside each, inclusive)
Threads (ms per frame)
Largest slices (ms per frame, inclusive, all threads): Driver.WorkerWait 1718.2, Driver.WorkerWait.dequeue_order 1684.3, Driver.Packet 963.9, Driver.Packet.memory_wait 620.6, Driver.Packet.dispatch 233.7, Driver.Dispatch 198.5, batch 93.9, Driver.Packet.draw 92.8, Driver.Draw 92.6, Graphics.Draw 74.4
Longest waits
Largest GPU batches (GPU ranges inside, and the CPU slices that recorded each, inclusive)
batch 218: 5.5 ms at 482 ms, present 2504: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 5.7, Driver.Packet.draw (queue worker 0x0) 5.1, Driver.Draw (queue worker 0x0) 5.0, Driver.Draw.shader_compile_and_link (queue worker 0x0) 2.4
batch 515: 5.4 ms at 1091 ms, present 2507: GPU no ranges; recorded by no CPU slices
batch 1408: 5.2 ms at 2940 ms, present 2514: GPU no ranges; recorded by Driver.Packet (queue worker 0x0) 5.7, Driver.Packet.draw (queue worker 0x0) 5.1, Driver.Draw (queue worker 0x0) 5.1, Driver.Draw.shader_compile_and_link (queue worker 0x0) 2.6
batch 266: 5.1 ms at 569 ms, present 2505: GPU no ranges; recorded by no CPU slices
batch 1904: 5.1 ms at 3990 ms, present 2519: GPU no ranges; recorded by Driver.Packet (queue worker 0x39) 9.5, Driver.Packet.dispatch (queue worker 0x39) 9.5, Driver.Dispatch (queue worker 0x39) 9.0, GpuMutex.wait.dispatch (queue worker 0x39) 7.2
Toggling the variable on one build of a local stack of open PRs gave the same direction: 36.2 vs 40.4 presents per 10 s in the prison, and 13.3 vs 14.8 in The Lost Legacy's Sri Lanka.
The variable stays opt-in because
APS5_JOB_AFFINITYmeasured 0 to -9 % at a video stage in the title it was written for.Checklist
main; no other open PR implements the same functionsNotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file.mdfiles or images added to the repository (attach them to this PR)