Performance Research T02 Exact Clock and Presentation Pacing

Exact QPC scaling and presentation-deadline control

T02 — Implementation and validation notes

Field	Value
Evidence state	Existing QPC work and separate presentation-only research; no new gain measured
Inspected source	322986f4c5502b3fa887a9605ac8cd88e9b2c579
First experiment	Exact integer scaling oracle plus fixed-work pacing controls
Scope	Addendum dated September 15, 2026; not a current-main benchmark or implementation status

Unproven performance research · Performance research — common test protocol · Performance research — reproducible checks

Evidence and implementation reasoning

Do not present existing work as a new discovery. The retained history already includes an exact Windows QPC scaling fast path and removal of unsafe host-load timer manipulation and persistent NVIDIA power-policy changes. A replacement host-load design is scoped to presentation deadlines. Research history.

Remaining research. Count clock calls and identify redundant sampling, then optimize only the arithmetic or scheduling mechanism whose exact result/contract is established. Keep virtual device clocks, guest execution, and host presentation pacing distinct.

Expected gain. Usually a host-overhead or jitter reduction. Faster guest clocks, different presentation caps, or more work per wall-clock second can be useful stress configurations, but they do not establish that the same emulated workload runs more efficiently. QEMU explicitly distinguishes instruction-count time from cycle-accurate modeling. TCG icount.

Acceptance. Exact scaling over boundary/overflow cases; pause/resume; long and short sessions; frame/audio pacing; fixed-work versus normally paced runs. Do not increase reported throughput by changing the workload's time semantics.

First patch boundary

Optimize redundant sampling or proven arithmetic only. Do not change guest clocks, title caps or driver power policy as an implicit optimization.

host counter -> exact scale -> host deadline; guest virtual clock and device semantics remain unchanged

Locate the inspected seam before editing; names or ownership may have changed since the pinned review:

git grep -n -E 'QueryPerformanceCounter|QueryPerformanceFrequency|QPC|deadline' -- util ui

Proposed focused tests

These are implementation gates to add or exercise, not claims that production tests already exist or passed. Retained standalone checks are on the checks page.

Test ID	Setup / deliberate control	Required oracle
T02.01	Zero/small/large counter values, around integer division and floating-point precision boundaries, using the supported frequency domain.	Candidate scaling matches a sufficiently wide exact-integer reference with defined overflow handling.
T02.02	Pause/resume, long/short sessions and audio/video pacing transitions.	No drift or altered guest device clock semantics.
T02.03	Normally paced runs and fixed-work runs with identical guest-time settings.	Host overhead is distinguishable from merely executing different amounts of emulated work.

Counters and rejection criteria

Collect: clock_calls, scale_cpu_ns, deadline_error_ns, host_wakes, guest_work, presented_frames, audio_underruns, frame_tail_ns. Keep exact counts separate from sampled durations; use an uninstrumented build for performance qualification.

Reject or hold: Do not claim speed by changing time semantics. Retain separate reports for host pacing, guest progress and actually presented frames.

Follow the common protocol for source identities, previous-main/fixed-baseline controls, raw results, confidence and retail/XISO qualification. No gain becomes proven from a model, cache hit, local operation count, or shifted wait label alone.