Runtime-generated FSR/AFMF presents may suppress simulation intervals and report 0 Simulation FPS
I maintainer di solito rispondono entro 2 giorni
Nessuno ha ancora preso questa issue.
Valutazione
- Difficoltà
- 4/5
- Tempo stimato
- 3-5 giorni
- Idoneità per principianti
- 45/100
- Tipo di issue
- Bug
- Chiarezza
- Specificata chiaramente
- Stato di attività
- Attiva
- Stack tecnologico
- cpp
- Ambito
- performance
Direzione di ricerca
Start with ComputeMsBetweenSimulationStarts() and the UnifiedSwapChain state updates for lastSimStartTime; read PcLatencyFrameGenerationTests::GeneratedPresentDoesNotTakeApplicationPclData for the existing PCL behavior. Add an end-to-end test for generated/application presents sharing a timestamp, including repeated pairs and the listed regression cases. Done means marked application frames produce intervals while generated rows and missing-marker frames do not, with unit coverage and a live or PresentBench2 ETL confirmation.
Scritto dal modello di indicizzazione a partire dal testo della issue.
Descrizione
Summary
Runtime frame generation using separate generated presents, such as FSR-FG or AFMF, may prevent PresentMon from producing msBetweenSimStarts samples when PCL simulation markers are enabled.
This is currently a source-supported hypothesis, not a confirmed live-hardware reproduction.
Suspected symptoms
Affected workloads may show:
MsBetweenSimulationStartasNA.- Missing
PM_METRIC_BETWEEN_SIMULATION_STARTframe values. - Simulation FPS or Simulation Frame Time reporting 0 despite active rendering.
- No valid simulation samples across repeated generated/application frame pairs.
This likely affects runtime-generated FSR/AFMF presents using PCL markers. It is not expected to affect ordinary presentation, application-provider markers, XeSS-FG's currently tested ordering, or generated displays attached to the application's own present.
Root-cause hypothesis
With FSR-style runtime frame generation:
- The generated present completes before its application present.
- Both presents receive the application's PCL simulation-start timestamp.
- The generated row emits no simulation interval because it is not an application frame.
- Despite that, it advances the swap chain's global
lastSimStartTime. - The application row then carries the same timestamp.
ComputeMsBetweenSimulationStarts()requirescurrentSimStartTime > lastSimStartTime, so the application row also emits no interval.
Repeating this sequence may leave the capture with no valid simulation interval samples.
Existing test PcLatencyFrameGenerationTests::GeneratedPresentDoesNotTakeApplicationPclData confirms the prerequisite that generated and application presents can receive the same PCL timestamp.
When this behavior arose
The underlying simulation interval logic has existed since at least commit ee39f134 on January 5, 2026.
The suspected duplicate-timestamp interaction took its current form in commit 081c989b on November 18, 2026, merged through #725. That change correctly allowed a generated present to read PCL data without claiming it, so the following application present could receive the same data.
Before #725, the generated present could consume the timing entry and leave the application present without PCL data. After #725, both presents can carry the same timestamp, exposing the lastSimStartTime interaction described above.
The existing CSV and frame-query simulation interval surfaces predate #741. That PR only makes the behavior more visible by deriving Simulation FPS and Simulation Frame Time from the same field.
Validation plan
End-to-end unit test
Construct a generated AMD_FSR_FG present followed by an application present sharing one PCL simulation timestamp, and feed them through UnifiedSwapChain in their expected completion/display order.
Verify that:
- The generated row has no simulation interval.
- The application row should produce an interval relative to the previous application marker.
- Repeated generated/application pairs continue producing application intervals.
- Missing-marker application frames, dropped frames, XeSS-style ordering, and source transitions do not regress.
This test should exercise actual state updates rather than assigning lastSimStartTime manually.
PresentBench2 experiment
Extend PresentBench2 with optional real FSR-FG support:
- Add a DX12
--use-fsr-fgmode and capability reporting. - Supply FSR-FG depth, motion-vector, and timing inputs.
- Emit
AMD_FSR_FGframe-type events for generated presents. - Run with
--emit-pcl-markers. - Capture ETL and inspect frame type, PCL frame ID, simulation timestamps,
MsBetweenSimulationStart, Simulation FPS, and Simulation Frame Time.
Compare frame generation disabled, FSR-FG enabled, and XeSS-FG enabled.
FSR-FG is cross-vendor, so AMD hardware is not inherently required. Use hardware explicitly supported by the selected FSR SDK version.
Fix proposals
Option A: Update simulation history only for application work
Prevent generated-only rows from advancing lastSimStartTime. Application frames without markers should continue leaving the previous valid marker intact.
This is the smallest fix, but all other users of lastSimStartTime must be audited.
Option B: Track application simulation history separately
Add a dedicated lastAppSimStartTime, update it only from application frames with valid markers, and use it for msBetweenSimStarts.
This avoids changing other consumers of lastSimStartTime and is likely safer, but requires coordinated V1, V2, bootstrap, dropped-frame, and reset handling.
Acceptance criteria
- Generated rows remain missing for application simulation interval metrics.
- Each marked application frame after initialization produces a valid interval.
- No CPU fallback is introduced.
- Missing markers do not erase prior valid application marker history.
- Existing PC latency, animation, XeSS-FG, CSV, and frame-query behavior does not regress.
- Unit coverage and a live or PresentBench2 ETL confirm the result.
- Lingua principale
- C++
- Stelle
- 2.6k
- Fork
- 201
- Merge medio
- 2g 21h
- PR unite (30g)
- 16
Preparare l'ambiente
Questo progetto non fornisce container di sviluppo, Dockerfile né guida per i contributori, quindi l'ambiente è a tuo carico: parti dal suo README e consulta la nostra guida al primo contributo per i passaggi generali.
Come iniziare
- Leggi tutta la issue e poi la guida ai contributi del progetto.
- Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
- Fai un fork del repository e lavora su un branch.
- Apri una pull request che faccia riferimento al numero della issue.
Altre issue di GameTechDev/PresentMon
-
Difficoltà 2/5 1-3 ore Idoneità per principianti 75/100
GameTechDev/PresentMon#745 ·
I maintainer di solito rispondono entro 2 giorni
-
Difficoltà 1/5 Meno di un'ora Idoneità per principianti 80/100
GameTechDev/PresentMon#684 · 1 commento ·
I maintainer di solito rispondono entro 2 giorni
-
Difficoltà 4/5 3-5 giorni Idoneità per principianti 45/100
GameTechDev/PresentMon#743 ·
I maintainer di solito rispondono entro 2 giorni
-
Reduce nonpaged memory reserved by the PresentMon ETW session at startupForse già presa @markgalvan-intel l’ha presa 3 giorni fa. ApertaPresentData
GameTechDev/PresentMon#740 · 1 assegnatario ·
I maintainer di solito rispondono entro 2 giorni
-
CLI usability issuesForse già presa @planetchili l’ha presa 10 giorni fa. Aperta
GameTechDev/PresentMon#734 · 1 assegnatario ·
I maintainer di solito rispondono entro 2 giorni
Tutte le issue di GameTechDev/PresentMon
Issue simili
-
Difficoltà 1/5 Meno di un'ora Idoneità per principianti 91/100
I maintainer di solito rispondono entro 1 giorno
-
Difficoltà 2/5 1-3 ore Idoneità per principianti 64/100
I maintainer di solito rispondono entro 1 giorno
-
Difficoltà 2/5 1-3 ore Idoneità per principianti 69/100
maplibre/maplibre-native#5072 ·
I maintainer di solito rispondono entro 1 giorno
-
bug grammar sev-5
Difficoltà 2/5 1-3 ore Idoneità per principianti 78/100
-
[UBSan] spawn_simulator passes null child_options.args to uv_spawn (execvp nonnull violation)Aperta
Difficoltà 2/5 1-3 ore Idoneità per principianti 79/100
tenstorrent/tt-umd#4017 ·
I maintainer di solito rispondono entro 3 giorni