Record glyph emission savings and remaining layout costs

docs/HANDOFF.md

@@ -423,7 +423,7 @@ The original proposal was implemented by Claude as `5fcace1` on
 `wip/region-and-placement` in `/home/bob/repos/iris-pr18`. It kept the fraction
 reference stable but failed five layout cases and three draw-count cases.
 The first correction is `efb416b`, exact recomposition is `2ed5503`, and the
-current performance/alignment continuation is **`7601aa2`**, pushed to origin's
+current performance continuation is **`a7307d9`**, pushed to origin's
 `wip/region-and-placement`, in the isolated checkout
 `/home/bob/repos/iris-layout-experiment`; **it is not on PR #18**. The original
 checkout is unchanged. Keep the experiment's build directory separate: Cargo
@@ -507,7 +507,7 @@ center. Do not describe those two centers as the same thing or compensate with
 an arbitrary offset. The parked padding API change is still separate.
 
 Ordinary verification passes: formatting, workspace clippy with
-`layout-diagnostics`, workspace tests (100 suite tests and 21 core tests), and
+`layout-diagnostics`, workspace tests (101 suite tests and 21 core tests), and
 the fast generated cases. The corrected `tabs` resize from 1920x1200 to
 900x1200 matches a cold image with zero differing pixels. The release oracle
 passes 2000 seeds at depth 4 and 1000 at depth 6; the shrinker passes 400 at
@@ -557,6 +557,51 @@ with each range less than 1.3% of its median. Additional depth-6 fixtures,
 7.0% for size and 8.4% for resize; seed 13 improves resize by 9.2%, while its
 size samples overlap and establish no cycle improvement.
 
+**Glyph emission now composes its origin once per text draw.** Glyph offsets
+and bitmap sizes are pixel lengths, so composition adds them unchanged. The
+fresh draw resolves the shared frame/extent origin once, then adds each glyph's
+offset and size. Retained local geometry is unchanged; a regression compares
+fresh drawing with retained recomposition for fractional origins, negative
+pixel offsets, changing frames and extents, and region nodes. This adds no
+retained state and changes no layout decisions.
+
+Against `7601aa2`, nine alternating pairs of direct, uninstrumented release
+executables under `perf stat -e cycles:u,instructions:u`, seed 1, depth 8,
+2000 frames per process, including setup:
+
+| phase | cycles before → after (billions) | instructions before → after (billions) |
+| --- | ---: | ---: |
+| size | 4.5164 → 4.0909 (-9.4%) | 12.1709 → 10.3993 (-14.6%) |
+| resize | 5.3025 → 4.8044 (-9.4%) | 14.2150 → 12.1999 (-14.2%) |
+| many | 5.2743 → 4.8071 (-8.9%) | 14.0989 → 12.0868 (-14.3%) |
+
+All nine samples per side were retained; before/after cycle ranges do not
+overlap. No scans or builds ran during measurement. All instrumented work
+counters match `7601aa2` on all six phases (50 frames, cold once), including
+the draw and primitive counts in the table above: this reduces the cost of
+emitting the repeated glyphs, not the repeated layout evaluations themselves.
+
+Additional depth-6 fixtures, nine alternating pairs and 4000 frames each:
+seed 3's size updates improve cycles by 5.0% and instructions by 8.0%; seed
+13's resizes improve cycles by 8.5% and instructions by 13.6%, both with
+nonoverlapping cycle ranges. Seed 3's resize and seed 13's size ranges overlap.
+On seed 1/depth 8, cold (one frame/process) and scroll (300,000 frames) also
+have overlapping cycle ranges. Repaint (1,000,000 frames) measures 1.1% more
+cycles, with nonoverlapping ranges and just 0.02% more instructions: this slice
+has a small measured repaint cost alongside the text-heavy gains. All nine
+samples per side were retained in these checks too.
+
+The verification runs above pass again at this head. All five reference
+renders (`view`, `minimal`,
+`random`, `tabs`, `text`) match `7601aa2` pixel for pixel at 1920x1200; `tabs`
+resized to 900x1200 matches a cold render there. The GPU probe reported Venus
+on the RX 7900 XT in this session.
+
+Two trials were removed: combining frame recomposition with extent movement
+saved under 0.8% of instructions and no cycles; special-casing equal endpoint
+fractions in `UiSpan::within` cost 11% cycles and 7% instructions despite still
+inlining. Keep that hot general composition small.
+
 **The remaining gap has not been established as an unavoidable correctness
 cost.** A temporary failure trace of one frame on seed 1, depth 8 counted
 39/108 failed reuse attempts in size, 65/151 in resize, and 45/109 in many