Performance

Performance

Reviewed: 2026-11-01 · Source: b7140848 · Availability: current main; stable/main boundaries are stated below.

Start with measured outcomes: the October comparison includes Windows PGR2 cadence gains, Deck PGR2 CPU/tail regressions and mixed Conker results. Its builds and conditions differ from November main. A faster component or a passing suite cannot establish a universal game speedup.

Need	Reference
Delivered mechanism, default and availability	Implemented · Current status
Exact tested outcomes and limitations	Measured results
Latest stable versus main	Release guide
Configure a named build	Controls
Evaluate cadence, tails, CPU and correctness	Methodology
Current PR/source inventory	Change register
Measured experiments not in main	Researched
Remaining proposal scope	Unproven
Mechanism classification	Optimization catalog · Hot-path shortcuts

An implementation can be merged with bounded evidence; a measured experiment can close unmerged; a removed feature can have a preserved branch. Keep source presence, release availability, default, validation and measured benefit separate. Operational improvements include restored progress, correct bytes, fewer crashes and better diagnosis without manufactured FPS percentages.