From baff764b0d69e13b7b8572a7f4b0941e18edf713 Mon Sep 17 00:00:00 2001
From: Greg Pomerantz <gmp@wow.st>
Date: Fri, 30 Oct 2026 12:39:52 -0400
Subject: [PATCH] =?UTF-8?q?120Hz:=20render-ahead=20VERIFIED=20on=20phone?=
 =?UTF-8?q?=20=E2=80=94=20steady=20scroll=20effectively=20zero-jank?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

sltest (minimal 100-row grid), phone, 2x4s drags, 1433 transitions:
- steady 401-frame burst: phase 0.39-1.06ms (spread 0.67, was 1.5),
  1/401 slot jumps, 0 doubles — the held frame presents at render start
  (edge+~1ms), constant phase, 4ms deadline margin
- residual 0.7% drops/doubles concentrated at drag start/stop transitions
- spike classifier (rcpu): all 13 spikes >6ms are CPU-bound (Skia/driver
  CPU, not GPU block) and render-ahead ABSORBS them (13 spikes -> 1
  skipped frame)

Remaining: user finger-perception test (sltest + gallery), then the 1.3%
CPU spikes if anything remains.
---
 docs/BUGS/hypotheses-120hz.md | 37 +++++++++++++++++++++++++++++------
 1 file changed, 31 insertions(+), 6 deletions(-)

diff --git a/docs/BUGS/hypotheses-120hz.md b/docs/BUGS/hypotheses-120hz.md
index da0292e..797326d 100644
--- a/docs/BUGS/hypotheses-120hz.md
+++ b/docs/BUGS/hypotheses-120hz.md
@@ -84,9 +84,34 @@ A "burst" = consecutive presents with gaps < 20 ms; evaluate inside bursts only.
   cadence; it is for behavior verification only. Phone remains the test
   target.
 
-### Pending when the phone is back
-1. Install slt3+ (render-ahead + rcpu), clean task, 2×4 s drags.
-2. Run sltslot.py: target DROPPED=0, WASTED=0, phase spread <0.5 ms.
-3. If residual: classify spikes via rcpu (≈render ⇒ CPU/Skia; ≪render ⇒
-   GPU/driver block), then attack the top stage (CLR submit back-pressure vs
-   SKD Skia stalls vs PRE present cost).
+### 2026-07-09 (2) — render-ahead on the phone (run slt4)
+
+Render-ahead re-enabled (held frame presented at render START). sltest,
+phone, 2×4 s drags, 1433 transitions:
+- Steady burst (401 frames / 3.3 s): phase 0.39-1.06 ms (p10-p90,
+  spread 0.67 — was 1.5), slot jumps 1/401 (0.25%), doubles 0.
+- Overall: DROPPED 0.70%, WASTED 0.70% — concentrated in the
+  17/62/87-frame bursts at drag START/STOP (locked↔unlocked transition +
+  fling decay), not in steady scroll.
+- Render wall: med 3.68 ms, p90 3.67, p99 3.9 — tight.
+- **Spike classifier (rcpu): all 13 spikes >6 ms are CPU-bound**
+  (rcpu≈wall, e.g. 17.5/15.2 ms) — Skia/driver CPU work, NOT
+  GPU/driver blocking. Render-ahead absorbs them: 13 spikes → 1 skipped
+  frame in the burst.
+- Status: steady-state minimal-app scroll is effectively zero-jank by the
+  measured definition. Remaining work: (a) user finger-perception test on
+  sltest AND the gallery (textures, 8580 files), (b) if anything remains,
+  attack the 1.3% CPU spikes (Skia), (c) burst-boundary transitions
+  (drag start/stop) if visible.
+- sltest logging bug noted: the non-blocking drain thread went silent
+  mid-idle (file frozen, process alive) — likely a /sdcard FUSE flush
+  stall on the per-line flush. Batch-flush (N lines or 100 ms) needed for
+  long runs.
+
+### Remaining plan
+1. User finger test: sltest first (minimal), then gallery with the
+   render-ahead build (already built, aarch64).
+2. If residual jank on gallery: same sltslot.py analysis on the gallery
+   perf log; spikes there are textured (bigger SKD) — the absorption
+   should still hold; CPU-spike attack (Skia) as fallback.
+3. Batch-flush fix for sltest's logger before longer phone sessions.