Repository navigation
Conversation
Complete the publication checklist and series cross-links for boykopovar#7469. No source or test changes.
|
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.
|
|
I compiled the old and new One addition to the fragmentation note in the description:
The old cost is O(pages/len) per probe; the new cost is O(blocks before the first fit). Whether this is a net win depends on the title's allocation pattern. A SuperTuxKart A-B-B-A result of +1.95% doesn't cover page-granular allocators. AI-assisted: yes (Claude Code); results above were run on this machine. |
|
@wachi35676 Thanks for testing the allocation semantics and measuring the page-granular case. Fixed in The three corrections (#7467, #7469 and #7470) were also checked together against the previous series in a same-scene SuperTuxKart A-B-B-A repeat: 32.680 -> 32.722 FPS (1920x1080 internal rendering, maximized 2560x1369 client, warm caches, trace seconds 20-28, RTX 3070 Laptop/NVIDIA 617.42). Focus and loaded-module hashes were verified. This combined check does not assign a separate FPS gain to this PR. #7473 is being validated separately. |
github-actions
Bot
added
perf
|
This pull request is labeled |
|
Additional current-main Linux check of the corrected PR ( Control: Ryzen 9 7950X, Radeon RX 6650 XT, RADV Mesa 26.2.4-arch1.1, Linux 7.2.9-arch1-1. Same scripted inputs, identical initial warm-cache clones (SHA-256 manifests attached), pipelined draws and DMA enabled in both. Official
This single ordered pair shows no gain in this scene (throughput −0.87%); it is not replicated evidence of a regression or a general conclusion about other titles. Both runs exited zero, restored configuration and the kernel limit, and left no native process. The 240-second screenshots show the same scene, HP 100/100 and readable tutorial text with no obvious visual regression.
Raw traces, trace digests, exact commands, cache hashes, screenshots and test logs. Official trace summary lines{"commit":"585075cb7ed3","dirty":false,"env":{"APS5_DRAW_CACHE_ENTRIES":"4096","APS5_HOST_IMPORT_MIB":"16384","APS5_PAD_INPUT_FILE":"<path>","APS5_PIPELINED_DRAWS":"1","APS5_PIPELINE_DMA":"1","APS5_SPIRV_OPT":"full","APS5_TRACE_SECONDS":"200:30"},"gpu":"AMD Radeon RX 6650 XT (RADV NAVI23)","driver":"radv Mesa 26.2.4-arch1.1","os":"Linux 7.2.9-arch1-1","resolution":"3840x2160","window":{"spec":"seconds 200:30","first_present":9729,"ms":30031.926},"presents":824,"frame_ms":{"p50":35.644,"p90":42.066,"p99":48.886,"max":55.291},"gpu_busy_ms_per_present":2.162,"worker_busy_pct":{"queue worker 0x0":26.296},"top_phases_ms_per_frame":{"Driver.Packet":9.573,"Graphics.Draw":4.414,"Graphics.Draw.validate":2.852,"VideoOut.Flip":1.744,"VideoOut.Flip.present":1.711,"Driver.Present":1.711,"Driver.Present.render_fence_wait":1.553,"Vulkan.Retire":1.553},"trace_dropped":false,"trace":"aps5-trace-585075cb7ed3-3566494.json.gz"}
{"commit":"bb114f70d356","dirty":false,"env":{"APS5_DRAW_CACHE_ENTRIES":"4096","APS5_HOST_IMPORT_MIB":"16384","APS5_PAD_INPUT_FILE":"<path>","APS5_PIPELINED_DRAWS":"1","APS5_PIPELINE_DMA":"1","APS5_SPIRV_OPT":"full","APS5_TRACE_SECONDS":"200:30"},"gpu":"AMD Radeon RX 6650 XT (RADV NAVI23)","driver":"radv Mesa 26.2.4-arch1.1","os":"Linux 7.2.9-arch1-1","resolution":"3840x2160","window":{"spec":"seconds 200:30","first_present":9824,"ms":30000.814},"presents":816,"frame_ms":{"p50":35.629,"p90":43.156,"p99":53.196,"max":65.631},"gpu_busy_ms_per_present":2.185,"worker_busy_pct":{"queue worker 0x0":28.054},"top_phases_ms_per_frame":{"Driver.Packet":10.304,"Graphics.Draw":4.798,"Graphics.Draw.validate":3.28,"VideoOut.Flip":1.74,"Driver.Present":1.711,"VideoOut.Flip.present":1.711,"Recorder.Sync":1.659,"Vulkan.Retire":1.552},"trace_dropped":false,"trace":"aps5-trace-bb114f70d356-3673174.json.gz"}AI-assisted: yes. |
github-actions
Bot
removed
the
perf numbers missing
|
@JJenkx, thanks for the Lodoss run. It matches what I see: no FPS claim for this PR. To show where it does work, I counted the allocator in SuperTuxKart's menu scene on current main: about 49 AI-assisted: yes (Claude Code) |
This branch has not been deployed
4 participants
What
This PR is part of a series of graphics and memory-management performance improvements for AnyPS5, with SuperTuxKart as the main profiling workload. This is item 06/10; the UI Lab startup/cache fix follows as a separate eleventh PR. Each topic's measurements are distinguished from the combined result.
Find first-fit direct-memory space by walking the existing ordered allocation map and skipping occupied ranges. Preserve bounds, alignment and first-fit results without a second free-space index.
Tested
Pinned benchmark base:
6447ed1ed7fd11c7eef60df336e8d66fac271188. Tested topic snapshot:6da6576d9ceca4122a164fccac59af4e416375b5. Windows 11 Pro build 26300, Ryzen 7 5800H, RTX 3070 Laptop, NVIDIA 617.42. Frozen runtime module hashes and raw traces were verified. No opt-in optimization flags are required.3/3 targeted cases passed on Windows and native Ubuntu 26.04: MemoryPoolTest, guest_memory and guest_mmap_fixed_anonymous. A 32,768-operation independent naive reference model checks allocation/free, exact first-fit addresses, non-power-of-two alignment, bounds and failures.
SuperTuxKart captures for the bot's before/after requirement: Windows 11 Pro build 26300, Ryzen 7 5800H, RTX 3070 Laptop GPU, NVIDIA 617.42, WinLibs GCC 15.2.0 r7, Release; SuperTuxKart (PPSA69421) tutorial prompt over the main menu, 1920x1080 internal rendering in a maximized 2560x1369 window, warm shader caches, audio muted, no input;
APS5_TRACE_SECONDS=20:8withAPS5_TRACE_EVENTS=1048576, console output not redirected, no profiling switches; 2026-10-18, one valid capture of this PR (its second capture lost focus and is excluded) against two of main, order main, PR, main.required1018-main-64476447ed1ed7fd6447ed1ed7fdrequired1018-main-6447-awake-r26447ed1ed7fd6447ed1ed7fdrequired1018-7469809c0009e3cfThe FPS difference is within the run-to-run spread of main; this PR claims no FPS change. What it saves is shown under Where it helps.
Summary lines (main first, then this PR)
{"commit":"6447ed1ed7fd","dirty":false,"env":{"APS5_TRACE_SECONDS":"20:8"},"gpu":"NVIDIA GeForce RTX 3070 Laptop GPU","driver":"NVIDIA 617.42","os":"Windows 10.0.26300","resolution":"1920x1080","window":{"spec":"seconds 20:8","first_present":42,"ms":8034.483},"presents":77,"frame_ms":{"p50":103.246,"p90":111.156,"p99":116.826,"max":116.826},"gpu_busy_ms_per_present":5.271,"worker_busy_pct":{"queue worker 0x0":3.148},"top_phases_ms_per_frame":{"Driver.Packet":3.234,"Driver.Packet.draw":2.618,"Driver.Draw":2.562,"Driver.Draw.shader_compile_and_link":1.512,"Driver.Present":0.815,"Shader.DrawStage":0.813,"Driver.Present.queue_present":0.547,"Vulkan.QueuePresent":0.545},"trace_dropped":false,"trace":"aps5-trace-6447ed1ed7fd-21748.json.gz"} {"commit":"6447ed1ed7fd","dirty":false,"env":{"APS5_TRACE_SECONDS":"20:8"},"gpu":"NVIDIA GeForce RTX 3070 Laptop GPU","driver":"NVIDIA 617.42","os":"Windows 10.0.26300","resolution":"1920x1080","window":{"spec":"seconds 20:8","first_present":45,"ms":8081.323},"presents":77,"frame_ms":{"p50":103.882,"p90":112.153,"p99":121.116,"max":121.116},"gpu_busy_ms_per_present":5.214,"worker_busy_pct":{"queue worker 0x0":3.193},"top_phases_ms_per_frame":{"Driver.Packet":3.301,"Driver.Packet.draw":2.672,"Driver.Draw":2.617,"Driver.Draw.shader_compile_and_link":1.540,"Driver.Present":0.839,"Shader.DrawStage":0.828,"Driver.Present.queue_present":0.571,"Vulkan.QueuePresent":0.566},"trace_dropped":false,"trace":"aps5-trace-6447ed1ed7fd-38428.json.gz"} {"commit":"809c0009e3cf","dirty":false,"env":{"APS5_TRACE_SECONDS":"20:8"},"gpu":"NVIDIA GeForce RTX 3070 Laptop GPU","driver":"NVIDIA 617.42","os":"Windows 10.0.26300","resolution":"1920x1080","window":{"spec":"seconds 20:8","first_present":37,"ms":8095.128},"presents":78,"frame_ms":{"p50":102.689,"p90":111.426,"p99":116.817,"max":116.817},"gpu_busy_ms_per_present":5.309,"worker_busy_pct":{"queue worker 0x0":3.221},"top_phases_ms_per_frame":{"Driver.Packet":3.290,"Driver.Packet.draw":2.679,"Driver.Draw":2.618,"Driver.Draw.shader_compile_and_link":1.539,"Driver.Present":0.891,"Shader.DrawStage":0.834,"Driver.Present.queue_present":0.568,"Vulkan.QueuePresent":0.566},"trace_dropped":false,"trace":"aps5-trace-809c0009e3cf-41712.json.gz"}tools/performance_compare.py(first capture of each side)Performance comparison
Configuration differences
Frame and GPU times (ms)
Top phases (ms per frame)
tools/trace_digest.pyoutput (first capture of each side; all digests are in the evidence archive)required1018-main-6447seconds 20:8 from present 42: 77 presents, frame p50 103.2 / p90 111.2 / p99 116.8 / max 116.8 ms, GPU busy 5.3 ms per present
NVIDIA GeForce RTX 3070 Laptop GPU, NVIDIA 617.42, Windows 10.0.26300, 1920x1080, commit 6447ed1, dropped: no
{"commit":"6447ed1ed7fd","dirty":false,"env":{"APS5_TRACE_SECONDS":"20:8"},"gpu":"NVIDIA GeForce RTX 3070 Laptop GPU","driver":"NVIDIA 617.42","os":"Windows 10.0.26300","resolution":"1920x1080","window":{"spec":"seconds 20:8","first_present":42,"ms":8034.483},"presents":77,"frame_ms":{"p50":103.246,"p90":111.156,"p99":116.826,"max":116.826},"gpu_busy_ms_per_present":5.271,"worker_busy_pct":{"queue worker 0x0":3.148},"top_phases_ms_per_frame":{"Driver.Packet":3.234,"Driver.Packet.draw":2.618,"Driver.Draw":2.562,"Driver.Draw.shader_compile_and_link":1.512,"Driver.Present":0.815,"Shader.DrawStage":0.813,"Driver.Present.queue_present":0.547,"Vulkan.QueuePresent":0.545},"trace_dropped":false,"trace":"aps5-trace-6447ed1ed7fd-21748.json.gz"}Slowest frames (largest slices inside each, inclusive)
Threads (ms per frame)
required1018-7469seconds 20:8 from present 37: 78 presents, frame p50 102.7 / p90 111.4 / p99 116.8 / max 116.8 ms, GPU busy 5.3 ms per present
NVIDIA GeForce RTX 3070 Laptop GPU, NVIDIA 617.42, Windows 10.0.26300, 1920x1080, commit 809c000, dropped: no
{"commit":"809c0009e3cf","dirty":false,"env":{"APS5_TRACE_SECONDS":"20:8"},"gpu":"NVIDIA GeForce RTX 3070 Laptop GPU","driver":"NVIDIA 617.42","os":"Windows 10.0.26300","resolution":"1920x1080","window":{"spec":"seconds 20:8","first_present":37,"ms":8095.128},"presents":78,"frame_ms":{"p50":102.689,"p90":111.426,"p99":116.817,"max":116.817},"gpu_busy_ms_per_present":5.309,"worker_busy_pct":{"queue worker 0x0":3.221},"top_phases_ms_per_frame":{"Driver.Packet":3.29,"Driver.Packet.draw":2.679,"Driver.Draw":2.618,"Driver.Draw.shader_compile_and_link":1.539,"Driver.Present":0.891,"Shader.DrawStage":0.834,"Driver.Present.queue_present":0.568,"Vulkan.QueuePresent":0.566},"trace_dropped":false,"trace":"aps5-trace-809c0009e3cf-41712.json.gz"}Slowest frames (largest slices inside each, inclusive)
Threads (ms per frame)
Metadata-only search probe, same C++ implementation with stub backing, ABBA order and medians of five trials per process (2,000 alloc/free cycles per trial): empty pool 0.0752 -> 0.0786 us/cycle; one 64-MiB allocation 1.8810 -> 0.0881; one 512-MiB allocation 14.5025 -> 0.0872; one 4-GiB allocation 115.8989 -> 0.0888. With 2,048 adjacent 64-KiB allocations, 3.7746 -> 8.3561 us/cycle (about 2.2x slower). This trades page scans for allocation-node scans; the contiguous-prefix win does not cover all fragmentation patterns and is not whole-frame timing.
Where it helps
About 48
sceKernelAllocateDirectMemorycalls per frame search the physical pool for a free range.2185dbf92ePhysicalMemoryPool::Alloccalls per framePhysicalMemoryPool::Allocµs per callPhysicalMemoryPool::Allocms per framesceKernelAllocateDirectMemorycalls per framesceKernelAllocateDirectMemoryµs per callsceKernelAllocateDirectMemoryms per frameCounters come from a local probe build that is not part of this PR (
review-probe.patchin the evidence archive): the same counters and timers were inserted into main2185dbf92eand into main with all eleven series PRs merged (#7464 in its rebased form), anchored on function definitions present in both trees. Runs were in the mirrored order main, series with #7464 reduced to its wake only, series, series, reduced series, main, in the scene above withAPS5_PROFILE_DRAW=1; values are medians of two runs over the same 8-second window as the trace, starting 20 s after the first flip, per flip. These runs carry profiling overhead and are not used for FPS. Each optimised function below is changed by this PR only; counters show where the work goes, not an isolated FPS gain.Other homebrew checks
UI Lab: the 21 default concepts were rendered and visually reviewed in a navigation cycle for this topic. Artwork was present; this covers default navigation, not all themes or every widget interaction.
Known baseline rendering issue: vkQuake still has intermittent flickering and an intermittently black floor on main and the tested performance branches. These changes do not fix it. Yamagi Quake II gameplay stops at unsupported last-provoking-vertex flat interpolation, and UZDoom at unsupported clip/cull-distance handling on the baseline; neither supplies a valid gameplay FPS comparison.
Dependencies, series and overlap
#6044 relocates the test files. Its MemoryPoolTest change is a relocation, not this allocator algorithm; whichever lands later must retain the model test at the current path.
The project overlap scan compared the local topics against 785 open PRs and checked merges into main
e5c2b3c325e485ba7fd417c47c46f48aa2c89115. This topic merges cleanly in that snapshot. Some other PRs already conflict with main, so the tool cannot compare all their shared-file hunks; the full report is retained with the common series evidence package. No exact duplicate was identified in the inspected implementations. This is not a certification that every conflicting foreign branch is composition-tested.Conventions and whitespace checks passed for the tested topic range. The benchmark base remains pinned for comparability; this is not a fresh-current-main build/test claim. Measurements and focused tests refer to the pinned source revisions above; no full-suite CI run is claimed.
Re-checked on 2026-10-21: this head still merges cleanly into main
2185dbf92e. #6658, which the earlier combined runs carried, merged into main on 2026-10-20.Combined series result
Re-measured on 2026-10-21 on current main
2185dbf92e(Windows 11 Pro build 26300, Ryzen 7 5800H, RTX 3070 Laptop GPU, NVIDIA 617.42, WinLibs GCC 15.2.0 r7, Release; SuperTuxKart (PPSA69421) tutorial prompt over the main menu, 1920x1080 internal rendering in a maximized 2560x1369 window, warm shader caches, audio muted, no input;APS5_TRACE_SECONDS=20:8withAPS5_TRACE_EVENTS=1048576, console output not redirected, no profiling switches). Runs in the order A-B-C-D-D-C-B-A, two per configuration; FPS is presents over the 8-s window.2185dbf92e#7464 and #7465 make the largest step together: the title's per-frame host imports disappear (see their PRs). With #7464 reduced to its CPU-label wake only, the whole series measured 10.07 / 9.01 FPS against main 9.24 / 9.30 in a separate block of the same session. Combined numbers are not the gain of any single PR. The combined builds are local merges of the published (#7465-#7470, #7472, #7474) and rebased (#7464, #7471, #7473) heads into main
2185dbf92e; the commits in their summary lines are those merges, not pushed. Summary lines, traces and digests: series-evidence.zip.Checklist
6447ed1ed7fd; merges cleanly into current main2185dbf92e(git merge-tree, 2026-10-21); related overlaps are listed aboveNotImplemented_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)The current head includes the review correction and validation recorded below.
Series index
These are the ten performance topics from the same profiling effort, followed by a separate UI Lab cache-preservation fix. Membership in the series does not imply a dependency.
Reproduction evidence
Series traces, summary lines, module hashes, focused-test logs and UI contact sheets. Game measurements use the pinned base and source revisions described above; assets contain no game binaries.
Update of 2026-10-21: 7469-evidence.zip (traces, summary lines,
trace_digest.pyandperformance_compare.pyoutput, loaded-module hashes), counters where.json and the probe review-probe.patch with probe_window.py.Review correction (2026-10-18)
Thanks for testing the allocation semantics and measuring the page-granular case. Fixed in
809c0009e3cf: after eight allocation-node probes the search falls back to the previous reverse-page search from the current candidate. On this machine the 16,384 one-page prefix goes from 92.16 us in the old PR to 7.50 us after the correction (7.43 us in the pinned main); the one-4-GiB-block case remains 0.085 us. These are metadata-only medians over five 2,000-cycle trials, not FPS. All 17 MemoryPoolTest cases pass on Windows and native Linux, including the 32,768-operation independent first-fit model and a new 16,384-block bounds/alignment case.The three corrections (#7467, #7469 and #7470) were also checked together against the previous series in a same-scene SuperTuxKart A-B-B-A repeat: 32.680 -> 32.722 FPS (1920x1080 internal rendering, maximized 2560x1369 client, warm caches, trace seconds 20-28, RTX 3070 Laptop/NVIDIA 617.42). Focus and loaded-module hashes were verified. This combined check does not assign a separate FPS gain to this PR. #7473 is being validated separately.