Repository navigation
fix(three): make material position stage and space explicit - #311
Draft
thejustinwalsh wants to merge 4 commits into
Draft
thejustinwalsh wants to merge 4 commits into
thejustinwalsh wants to merge 4 commits into
Conversation
Docs report 📚Every changed package is attested at this head. A reviewer verifies each claim after merge.
✅ attested · |
size-limit report 📦
|
This was referenced Oct 18, 2026
Package performance:
|
| statistic | baseline | candidate |
|---|---|---|
| Median workload p50 | 4.350ms | 3.850ms |
| Workloads ≤1 ms | 19 | 18 |
| Workloads ≤2 ms | 41 | 41 |
| Workloads ≤3 ms | 47 | 47 |
| Workloads >3 ms | 52 | 52 |
Largest candidate workloads (including neutral results):
| workload | baseline | candidate | Δ time |
|---|---|---|---|
| 1000 labels 100 edits interleaved edit-read publication | 318.51ms | 318.64ms | +0.130ms |
| 1000 labels 100 edits immediate measurements with publication | 240.97ms | 240.13ms | -0.840ms |
| mutate one trailing span across 1000 formatted labels | 111.82ms | 114.26ms | +2.440ms |
| 100 labels interleaved edit-read publication | 60.25ms | 60.85ms | +0.600ms |
| scattered character assignment in one large paragraph with 8192 paint spans | 45.60ms | 44.87ms | -0.730ms |
faster ◀ ┊ ▶ slower Δ p50 p baseline candidate
+ issue 216 reporter workaround ├─●─┤┊ -9.2% .028 ▁▇█▄▂▃▆▅▂▂ ▃█▃▁▁▃▅▂▁▁
+ write one paragraph and text mutation ├────●┤ ┊ -11.4% .002 ▂█▄▂▂▁▁▃▂▁ █▆▃▁▁▁▃▂▁▁
+ write 22k UTF16 text into owned array control ├──●──┤┊ -12.6% .005 ▄█▄▂▂▁▂▁▂▂ ██▂▁▁▂▁▁▁▁
cached measurement ├┼─●───┤ +7.5% .279 █▄▂▁▁▁▂▄▂▁ █▄▁▁▁▁▂▄▄▁
reflow after width change ├─●─┤ +7.3% .083 ▆█▅▃▅▄▅▃▁▁ ▂▆█▅▅▅▆▄▂▂
100 labels first same-length scene publication ├─●┤ +6.4% .161 ▃█▄▄▃▃▃▃▂▂ ▃█▅▄▅▃▃▂▂▁
normalize 1000 equivalent formatted flow updates ┊●┤ +4.6% <.001 ▂▇█▅▄▃▃▂▁▁ ▁▂██▅▄▄▂▃▁
type one character into a long paragraph and publish a frame ├●┤ +4.3% .065 ▄█▆▃▂▃▄▂▂▁ ▁█▇▅▂▂▄▂▂▂
bind and normalize 1000 framework…ped formatted flow updates ├┼●┤ +3.8% .279 ▂▅▆█▄▅▅▄▃▂ ▁▁▆██▆▅▅▄▂
100 labels last same-length engine publication ├┼●─┤ +3.6% .328 ▃█▂▃▃▃▂▁▂▁ ▃█▃▅▃▂▃▂▂▂
publish after text change ├┼●┤ +3.6% .195 ▅█▄▃▃▃▂▃▂▁ ▂█▆▃▄▃▃▂▂▁
measurement then publication after width change ├●┤ +3.1% .083 ▁▁▁▇█▇▆▅▂▁ ▁▁▁▇█▇▇▃▅▂
100 labels first length-changing engine publication ├●┤ +3.0% .161 ▃█▄▃▃▃▃▂▁▁ ▆█▆▄▄▃▄▁▂▁
copy per-glyph metrics from 100 unchanged labels ├┼●┤ +2.7% .798 ▃█▅▂▂▂▁▃▁▁ ▃█▄▃▂▁▁▂▂▂
edit and independently copy one label among 1000 labels ├●┤ +2.6% .195 ▁▄▆█▆▆▄▂▂▁ ▁▄▃▆▆█▆▂▂▁
normalize 1000 equivalent retained label updates ├──┼●┤ +2.3% .959 ▂█▄▄▆▄▃▂▂▂ ▁█▃▂▅▃▄▂▁▁
mutate one trailing span across 1000 formatted labels ├● +2.2% .038 ▁▄█▅▇▆▃▂▁▁ ▁▄█▆▇▄▅▄▂▂
measure after text change ├●┤ +2.2% .161 █▅▇▅▂▁▃▂▂▁ ▄█▄▆▃▂▃▂▂▁
100 labels first color-only engine publication ├┼●┤ +1.9% .574 ▂█▃▁▂▃▂▃▂▁ ▅█▅▁▃▅▂▃▃▂
100 labels 100 edits immediate me…ements without publication ├● +1.8% .130 ▂█▄▃▂▃▂▁▂▁ ▂█▅▃▂▂▂▂▂▁
relayout after font-size change ├●┤ +1.6% .442 █▄▄▅▂▂▁▁▂▁ ▂█▃▄▃▂▁▁▁▁
10 labels last color-only engine publication ●─┤ +1.5% .195 ▄██▄▂▃▃▅▆▂ ▂█▇▇▂▂▅▆▆▂
rewrite 64 labels in one group ├─●─┤ +1.4% .798 ▄▆▄▅█▃▄▂▂▁ ▆█▅▅█▅▅▂▂▂
edit and independently copy one label among 1 labels ●┤ +1.3% .195 ▃▇█▆▆▃▂▂▃▃ ▂▇█▇▆▄▃▂▄▃
10 labels last length-changing scene publication ├●┤ +1.2% .798 ▁▅█▆▂▄▂▃▃▁ ▂▇█▇▂▅▅▄▃▂
borrow per-glyph metrics from 100 promoted labels ●─┤ +1.1% .328 ▁▁▄█▄▂▁▁▁▁ ▁▁▄█▄▂▁▁▁▁
publication after width change ├●┤ +1.1% .878 ▁▅█▅▆▅▃▃▁▂ ▁▆█▇▆▆▃▃▁▁
10 labels first same-length engine publication ├●┤ +1.0% .442 ▄█▇▆▆▇▄▃▂▂ ▁▅█▃▅▄▄▂▂▁
100 labels interleaved edit-read publication ● +1.0% .505 ▂▅█▅▃▂▂▂▂▁ ▁▇▇█▅▂▃▃▁▁
100 labels last length-changing engine publication ●─┤ +0.9% .328 ▃█▄▄▃▃▃▃▂▂ ▇█▅▆▃▅▄▃▂▁
1000 labels no-op scene traversal ├●┤ +0.9% .442 ▃█▇▅▂▂▂▂▂▁ ▁█▇▄▂▂▂▂▂▁
100 labels batched edit-read publication ├●┤ +0.9% .798 ▂██▄▄▄▄▂▂▁ ▂█▆▅▂▃▄▂▂▁
mount and publish a 22k-glyph paragraph ● +0.7% .959 ▃█▃▅▂▂▁▁▁▁ ▃█▆▄▄▂▂▂▂▁
10 labels first color-only scene publication ├●┤ +0.6% .505 ▁█▆▅▆▄▇▅▃▂ ▂█▇▄▆▃▆▇▆▂
1000 labels last color-only scene publication ├●┤ +0.5% .505 ▄█▅▇▅▅▂▂▂▁ ▃▆▅█▅▄▂▁▁▁
type one character into a long Fredoka paragraph and measure ├●┤ +0.5% .959 ▆██▇▅▅▂▁▂▁ ▁█▆▆▅▅▂▁▁▁
1000 labels first length-changing engine publication ● +0.3% .721 ▁▄█▆▆▂▁▁▁▂ ▁▄█▇▃▂▂▁▁▁
10 labels no-op scene traversal ├●─┤ +0.2% .878 ▁▃▃▃█▅▆▅▃▂ ▃▃▃▆▇█▆▅▄▁
1000 labels last same-length engine publication ├● +0.1% .959 ▂▅▇█▇▅▄▃▃▂ ▂▇▆██▇▅▄▂▁
type one character into a long paragraph and measure ├●┤ +0.1% .721 ▅█▅▇▆▁▁▂▂▁ ▁█▆▅▄▂▁▂▂▂
1000 labels 100 edits interleaved edit-read publication ● 0.0% .959 ▃█▄▁▁▁▁▁▂▁ ▁█▇▁▁▁▁▁▁▁
100 labels no-op scene traversal ├●┤ -0.1% .798 ▂█▄▂▃▃▃▂▃▁ ▄█▅▂▄▃▃▃▂▁
1000 labels first color-only scene publication ●┤ -0.1% .645 ▄▇▆█▄▂▂▂▂▂ ▄█▄█▄▂▃▂▂▁
1000 labels first color-only engine publication ├●┤ -0.1% .959 ▂█▆▅▇▆▃▁▁▁ ▃█▄█▆▅▄▂▁▁
per-glyph inspection after width change ├─●┤ -0.1% .574 ▂▂▃▁▂▃▄█▇▁ ▁▂▃▁▂▃▃█▅▁
reorder then edit one of 1000 retained labels ├●┤ -0.1% .959 ▃█▇▆▄▃▂▃▁▂ ▄▇█▇▄▂▃▂▁▁
move 1000 retained TypeGPU labels ├─●─┤ -0.2% .721 ▁▇█▇▃▂▂▁▂▁ ▂▆█▅▃▂▁▂▁▁
10 labels first same-length scene publication ●┤ -0.3% .721 ▁▅█▅▃▂▄▃▃▂ ▂██▆▃▄▄▅▃▂
1000 labels 100 edits immediate m…surements with publication ├● -0.3% .442 ▃▅▃▁██▇▅▂▁ ▃▇▂▁█▆▃▂▁▂
mount a label and publish its first edit ├─●─┤ -0.3% .798 ▂██▄▇▆▅▅▃▂ ▃▇█▄▄▆▅▄▄▂
mount 100 labels and borrow glyphs from each once ├● -0.3% .442 ▂▇█▄▂▃▃▂▁▁ ▂█▇▄▂▃▂▂▂▂
rewrite 256 labels in one group ├●┤ -0.4% .721 ▃██▇▂▂▂▁▂▁ ▁▇█▆▂▂▂▁▁▁
100 labels 100 edits immediate measurements with publication ● -0.4% .721 ▂█▇▅▄▂▃▂▃▁ ▃█▇▄▃▃▂▄▁▁
rewrite 512 labels in one group ├●┤ -0.5% .959 ▂▇▆█▄▃▂▂▁▁ ▃▆█▆▄▃▂▁▂▁
edit and measure one of 100 retained labels ├●──┤ -0.7% .959 ▂▆▇█▆▃▅▃▂▂ ▃▆█▅▆▄▃▅▃▁
1000 labels 100 edits batched edit-read publication ├●┤ -0.7% .505 ▁▄█▅▅▃▃▃▂▁ ▃▆█▅▅▄▃▂▁▁
read every outline from an unchanged copy ├● -0.7% .442 █▃▂▂▁▁▁▁▁▁ █▂▂▁▁▁▁▁▁▁
1000 labels first length-changing scene publication ├●┤ -0.8% .721 ▁▅██▆▄▃▂▁▁ ▁▆██▄▅▂▁▂▁
1000 labels first same-length engine publication ├● -0.8% .050 ▁▅█▆▆▃▃▁▁▁ ▂▅█▇▃▂▁▁▁▁
100 labels last length-changing scene publication ├●──┤ -1.0% .878 ▄██▅▅▅▄▅▃▂ ▃▇█▅▃▅▅▅▂▁
1000 labels last same-length scene publication ├●┤ -1.0% .195 ▂▇█▄▃▂▁▁▁▁ ▁▅█▄▂▂▁▁▁▁
1000 labels 100 edits immediate m…ements without publication ├● -1.0% .161 ▃▇▄█▆▃▃▃▃▂ ▃█▇▆▆▂▂▃▃▂
reuse 1000 unchanged Vue property snapshots ├●┤ -1.1% .959 ▁█▂▁▁▂▂▁▂▁ ▃█▂▂▂▂▂▂▂▁
publication after font-size change ├● -1.1% .328 ▂▅█▇▄▆▅▄▂▂ ▂█▇▅▆▆▄▃▁▂
long unsafe-boundary paragraph text edit ├● -1.2% .645 ▃▆██▅▄▃▂▁▂ ▅▆██▅▃▃▂▂▂
1000 labels first same-length scene publication ├●┤ -1.3% .279 ▁▅█▇▂▂▁▁▁▁ ▂██▇▄▃▂▁▁▂
publish 1024 retained Text instances ├● -1.3% .234 ▂▃▄█▇▅▆▅▃▂ ▁▆▆▇█▄▄▂▃▁
broad character assignment in one…raph with 8192 paint spans ├● -1.4% .038 ▁▂█▇▅▃▂▂▂▁ ▂▅█▆▃▂▂▁▂▁
edit and sparsely borrow one of 1000 retained labels ├●┤ -1.4% .645 ▂▇█▅▃▅▄▅▄▂ ▁█▇▃▃▃▅▃▃▂
10 labels last color-only scene publication ├─●┤ -1.5% .382 ▁▆█▇▇▅▇█▇▃ ▃▇▆▅▆▄▄█▅▃
100 labels last same-length scene publication ├─●┤ -1.5% .645 ▄█▄▄▄▄▃▃▃▁ ▃█▅▄▄▄▃▃▂▁
rewrite 128 labels in one group ├●─┤ -1.6% .959 ▃█▆▅▅▃▄▃▁▁ ▄█▆▄▅▃▄▃▁▁
scattered character assignment in…raph with 8192 paint spans ├● -1.6% .105 ▂██▅▂▂▂▁▁▁ ▂█▃▃▂▂▂▁▁▁
1000 labels last length-changing engine publication ●┤ -1.7% .021 ▁▄██▄▅▃▁▁▁ ▁▄█▄▃▂▁▁▁▁
100 labels first length-changing scene publication ├●┼┤ -1.9% .234 ▂▇█▅▃▄▅▃▂▂ ▄█▆▃▃▅▃▃▁▁
100 labels last color-only scene publication ├─●┼┤ -2.0% .382 ▅█▅▃▃▃▂▄▄▂ ▂█▃▃▂▂▃▃▃▁
10 labels first length-changing scene publication ●┤ -2.1% .130 ▂▇█▆▃▅▅▃▄▂ ▂█▅▃▃▅▃▄▂▁
10 labels first length-changing engine publication ●┤ -2.1% .234 ▂█▇▆▇▅▅▃▄▂ ▃█▅▄▃▅▄▃▁▂
copy per-glyph metrics after text change ●┤ -2.1% .105 ▁▄▄█▅▄▃▃▂▂ ▁▇▄█▅▃▂▃▂▁
copy per-glyph metrics with outlines after text change ●┤ -2.1% .234 ▁▁▁▁▄█▄▃▄▂ ▁▁▁▂▅█▅▄▂▂
100 labels first color-only scene publication ├●┼┤ -2.2% .328 ▅█▅▃▂▄▃▂▄▁ ▂█▄▂▁▃▂▂▃▁
100 labels first same-length engine publication ├●┼┤ -2.4% .442 ▃█▂▃▂▃▂▂▂▁ ▃█▁▂▃▃▂▂▂▁
measurement after width change ├●┼┤ -2.4% .505 ▃█▄▄▂▁▃▂▁▁ ▄█▆▃▃▂▃▁▁▁
edit and measure one of 1000 retained labels ├─●┤ -2.6% .105 ▃█▅▃▃▅▃▃▃▁ ▃█▃▂▃▂▃▂▂▂
10 labels last same-length engine publication ●┤ -2.7% .105 ▁██▆█▆▄▂▃▂ ▂▆██▅▄▂▄▃▁
1000 labels last color-only engine publication ●┤ -2.7% .234 ▂▆▆██▅▃▂▂▂ ▃█▅▆█▄▂▁▁▁
cached per-glyph inspection ├●┼─┤ -2.9% .798 ▂▃▆█▇▂▄▄▂▂ ▁▂▇█▃▂▃▃▃▁
long unsafe-boundary paragraph reflow ├●┤ -2.9% .083 ▂▆▅▄▇█▆▄▄▂ ▆▆▅▄█▇▅▄▂▁
100 labels last color-only engine publication ├●┼┤ -3.2% .382 ▅█▄▂▂▄▃▂▄▂ ▃█▃▁▂▃▃▃▃▁
long reporter workaround paragraph reflow ├●┤ -3.2% .195 ▃█▄▃▆▆▅▃▃▁ █▇▃▇▇▅▅▄▂▁
10 labels last length-changing engine publication ●┤ -3.4% .005 ▂▅▇█▅▄▄▄▂▂ ▂██▆▄▃▂▃▂▁
inspection then publication after width change ├●┼─┤ -3.4% .442 ▃█▄▁▅▆▄▅▄▂ ▃█▄▁▅▅▄▅▃▁
common-safe word-wrap reflow ├●┤ -3.4% .028 ▁██▂▂▂▄█▃▂ ▃█▃▂▁▃▅▄▁▁
reorder 1000 unchanged retained labels ├─●┤ -4.7% .003 ▂█▆▇▆▅▂▁▁▁ ▃▇█▅▂▃▂▁▁▁
issue 216 unsafe-boundary word-wrap reflow ├●─┤ -7.0% .065 ▁█▇▄▂▂▇▅▃▁ ▃█▂▃▁▄▅▃▁▁
measure 1000 unchanged retained labels ├─●─┼┤ -7.4% .234 ▁▇▆▅█▇▅▅▃▂ ▁█▇▆█▇▂▁▁▁
write 1000 paragraph-order mutations ├──●─┼──────────────▶ -7.5% .279 ▆▄█▂▂▁▂▁▁▁ ▅█▇▄▂▁▁▁▂▁
write 1000 paragraph and text mutations ├───●──┤ -9.1% .065 ▃█▇▇▆▅▂▂▂▁ ▃█▆▄▂▁▁▁▁▁
borrow glyphs from 1000 steadily promoted retained labels ├──────●──────┼───┤ -23.4% .382 ▅▆▅▃▁▁▃▆█▇ ▄▆▆▃▁▁▂██▁
└──────────────┴──────────────┘
-50% 0 +50%| status | bench | baseline | candidate | Δ time | Δ p50 | p | 95% CI |
|---|---|---|---|---|---|---|---|
| 🟢 faster | issue 216 reporter workaround | 868.71µs | 788.53µs | -0.080ms | -9.2% | .028 | -15.6..-2.9% |
| 🟢 faster | write one paragraph and text mutation | 53.39µs | 47.31µs | -0.006ms | -11.4% | .002 | -28.0..-6.0% |
| 🟢 faster | write 22k UTF16 text into owned array control | 187.37µs | 163.78µs | -0.024ms | -12.6% | .005 | -21.7..-4.2% |
96 neutral, 9 skipped
| status | bench | baseline | candidate | Δ time | Δ p50 | p | 95% CI |
|---|---|---|---|---|---|---|---|
| neutral | normalize 1000 equivalent retained label updates | 7.48ms | 7.65ms | +0.170ms | +2.3% | .959 | -8.8..+8.3% |
| neutral | mutate one trailing span across 1000 formatted labels | 111.82ms | 114.26ms | +2.440ms | +2.2% | .038 | +0.1..+3.1% |
| neutral | edit and measure one of 100 retained labels | 539.37µs | 535.72µs | -0.004ms | -0.7% | .959 | -4.6..+10.1% |
| neutral | rewrite 64 labels in one group | 4.36ms | 4.42ms | +0.060ms | +1.4% | .798 | -5.5..+5.2% |
| neutral | rewrite 128 labels in one group | 6.53ms | 6.43ms | -0.100ms | -1.6% | .959 | -4.7..+5.3% |
| neutral | rewrite 256 labels in one group | 10.62ms | 10.58ms | -0.040ms | -0.4% | .721 | -1.7..+4.4% |
| neutral | rewrite 512 labels in one group | 19.60ms | 19.51ms | -0.090ms | -0.5% | .959 | -3.2..+3.0% |
| neutral | 10 labels no-op scene traversal | 170.92µs | 171.28µs | +0.000360ms | +0.2% | .878 | -4.2..+5.4% |
| neutral | 10 labels first same-length scene publication | 1.27ms | 1.27ms | 0.000ms | -0.3% | .721 | -1.4..+4.4% |
| neutral | 10 labels first length-changing scene publication | 1.73ms | 1.69ms | -0.040ms | -2.1% | .130 | -4.9..+0.5% |
| neutral | 10 labels first color-only scene publication | 1.21ms | 1.22ms | +0.010ms | +0.6% | .505 | -2.2..+4.8% |
| neutral | 10 labels first same-length engine publication | 1.24ms | 1.26ms | +0.020ms | +1.0% | .442 | -2.4..+4.7% |
| neutral | 10 labels first length-changing engine publication | 1.70ms | 1.66ms | -0.040ms | -2.1% | .234 | -4.7..+0.9% |
| neutral | 10 labels last length-changing scene publication | 1.65ms | 1.67ms | +0.020ms | +1.2% | .798 | -2.2..+3.2% |
| neutral | 10 labels last color-only scene publication | 1.20ms | 1.18ms | -0.020ms | -1.5% | .382 | -7.4..+2.0% |
| neutral | 10 labels last same-length engine publication | 1.27ms | 1.23ms | -0.040ms | -2.7% | .105 | -4.2..+0.2% |
| neutral | 10 labels last length-changing engine publication | 1.64ms | 1.59ms | -0.050ms | -3.4% | .005 | -5.0..-1.2% |
| neutral | 10 labels last color-only engine publication | 995.44µs | 1.01ms | +0.015ms | +1.5% | .195 | -1.3..+5.0% |
| neutral | 100 labels no-op scene traversal | 262.87µs | 262.51µs | -0.000360ms | -0.1% | .798 | -3.4..+4.7% |
| neutral | 100 labels first same-length scene publication | 1.74ms | 1.86ms | +0.120ms | +6.4% | .161 | -1.6..+10.3% |
| neutral | 100 labels first length-changing scene publication | 2.54ms | 2.49ms | -0.050ms | -1.9% | .234 | -6.0..+1.8% |
| neutral | 100 labels first color-only scene publication | 1.44ms | 1.41ms | -0.030ms | -2.2% | .328 | -5.7..+2.4% |
| neutral | 100 labels first same-length engine publication | 1.56ms | 1.52ms | -0.040ms | -2.4% | .442 | -6.5..+3.7% |
| neutral | 100 labels first length-changing engine publication | 2.29ms | 2.36ms | +0.070ms | +3.0% | .161 | -1.4..+7.0% |
| neutral | 100 labels first color-only engine publication | 1.25ms | 1.27ms | +0.020ms | +1.9% | .574 | -4.2..+5.8% |
| neutral | 100 labels last same-length scene publication | 1.80ms | 1.77ms | -0.030ms | -1.5% | .645 | -7.9..+4.3% |
| neutral | 100 labels last length-changing scene publication | 2.26ms | 2.24ms | -0.020ms | -1.0% | .878 | -2.8..+8.5% |
| neutral | 100 labels last color-only scene publication | 1.45ms | 1.42ms | -0.030ms | -2.0% | .382 | -11.0..+3.7% |
| neutral | 100 labels last same-length engine publication | 1.60ms | 1.66ms | +0.060ms | +3.6% | .328 | -3.2..+10.0% |
| neutral | 100 labels last length-changing engine publication | 2.02ms | 2.04ms | +0.020ms | +0.9% | .328 | -1.6..+6.2% |
| neutral | 100 labels last color-only engine publication | 1.27ms | 1.22ms | -0.050ms | -3.2% | .382 | -7.7..+4.9% |
| neutral | 1000 labels no-op scene traversal | 1.25ms | 1.26ms | +0.010ms | +0.9% | .442 | -1.9..+4.2% |
| neutral | 1000 labels first same-length scene publication | 6.88ms | 6.80ms | -0.080ms | -1.3% | .279 | -2.5..+4.5% |
| neutral | 1000 labels first length-changing scene publication | 10.76ms | 10.67ms | -0.090ms | -0.8% | .721 | -1.8..+1.8% |
| neutral | 1000 labels first color-only scene publication | 5.44ms | 5.44ms | 0.000ms | -0.1% | .645 | -1.6..+2.6% |
| neutral | 1000 labels first same-length engine publication | 6.30ms | 6.25ms | -0.050ms | -0.8% | .050 | -2.4..-0.1% |
| neutral | 1000 labels first length-changing engine publication | 10.11ms | 10.14ms | +0.030ms | +0.3% | .721 | -1.2..+0.7% |
| neutral | 1000 labels first color-only engine publication | 4.62ms | 4.62ms | 0.000ms | -0.1% | .959 | -2.3..+3.1% |
| neutral | 1000 labels last same-length scene publication | 6.85ms | 6.78ms | -0.070ms | -1.0% | .195 | -1.7..+2.5% |
| neutral | 1000 labels last color-only scene publication | 5.42ms | 5.45ms | +0.030ms | +0.5% | .505 | -2.9..+1.8% |
| neutral | 1000 labels last same-length engine publication | 6.16ms | 6.16ms | 0.000ms | +0.1% | .959 | -1.9..+1.4% |
| neutral | 1000 labels last length-changing engine publication | 7.31ms | 7.18ms | -0.130ms | -1.7% | .021 | -4.9..-0.5% |
| neutral | 1000 labels last color-only engine publication | 4.79ms | 4.66ms | -0.130ms | -2.7% | .234 | -4.8..+1.5% |
| neutral | 100 labels 100 edits immediate measurements without publication | 12.25ms | 12.47ms | +0.220ms | +1.8% | .130 | -0.6..+2.3% |
| neutral | 100 labels 100 edits immediate measurements with publication | 40.07ms | 39.89ms | -0.180ms | -0.4% | .721 | -1.4..+0.8% |
| neutral | 100 labels batched edit-read publication | 18.04ms | 18.21ms | +0.170ms | +0.9% | .798 | -2.8..+3.3% |
| neutral | 100 labels interleaved edit-read publication | 60.25ms | 60.85ms | +0.600ms | +1.0% | .505 | -0.8..+1.5% |
| neutral | 1000 labels 100 edits immediate measurements without publication | 14.73ms | 14.59ms | -0.140ms | -1.0% | .161 | -2.5..+0.2% |
| neutral | 1000 labels 100 edits immediate measurements with publication | 240.97ms | 240.13ms | -0.840ms | -0.3% | .442 | -4.2..+1.1% |
| neutral | 1000 labels 100 edits batched edit-read publication | 26.31ms | 26.11ms | -0.200ms | -0.7% | .505 | -2.9..+1.8% |
| neutral | 1000 labels 100 edits interleaved edit-read publication | 318.51ms | 318.64ms | +0.130ms | 0.0% | .959 | -1.0..+1.3% |
| neutral | mount and publish a 22k-glyph paragraph | 37.71ms | 37.99ms | +0.280ms | +0.7% | .959 | -1.3..+1.6% |
| neutral | mount a label and publish its first edit | 6.48ms | 6.46ms | -0.020ms | -0.3% | .798 | -6.2..+7.4% |
| neutral | mount 100 labels and borrow glyphs from each once | 31.85ms | 31.75ms | -0.100ms | -0.3% | .442 | -2.9..+1.1% |
| neutral | measure after text change | 6.21ms | 6.35ms | +0.140ms | +2.2% | .161 | -0.3..+5.5% |
| neutral | publish after text change | 4.90ms | 5.08ms | +0.180ms | +3.6% | .195 | -2.8..+7.4% |
| neutral | reflow after width change | 1.62ms | 1.74ms | +0.120ms | +7.3% | .083 | -0.7..+13.1% |
| neutral | relayout after font-size change | 4.35ms | 4.42ms | +0.070ms | +1.6% | .442 | -4.7..+4.2% |
| neutral | type one character into a long paragraph and measure | 5.09ms | 5.10ms | +0.010ms | +0.1% | .721 | -2.3..+3.3% |
| neutral | type one character into a long Fredoka paragraph and measure | 4.35ms | 4.38ms | +0.030ms | +0.5% | .959 | -2.7..+4.0% |
| neutral | type one character into a long paragraph and publish a frame | 8.23ms | 8.59ms | +0.360ms | +4.3% | .065 | -0.1..+7.1% |
| neutral | reuse 1000 unchanged Vue property snapshots | 798.20µs | 789.10µs | -0.009ms | -1.1% | .959 | -3.5..+3.8% |
| neutral | normalize 1000 equivalent formatted flow updates | 15.48ms | 16.20ms | +0.720ms | +4.6% | <.001 | +2.5..+6.7% |
| neutral | bind and normalize 1000 framework-shaped formatted flow updates | 17.61ms | 18.27ms | +0.660ms | +3.8% | .279 | -2.5..+6.7% |
| neutral | copy per-glyph metrics after text change | 12.42ms | 12.16ms | -0.260ms | -2.1% | .105 | -4.8..+0.9% |
| neutral | copy per-glyph metrics from 100 unchanged labels | 499.42µs | 512.99µs | +0.014ms | +2.7% | .798 | -4.4..+6.0% |
| neutral | edit and independently copy one label among 1 labels | 3.31ms | 3.35ms | +0.040ms | +1.3% | .195 | -1.0..+3.6% |
| neutral | edit and independently copy one label among 1000 labels | 7.80ms | 8.01ms | +0.210ms | +2.6% | .195 | -1.1..+5.5% |
| neutral | borrow per-glyph metrics from 100 promoted labels | 1.03ms | 1.04ms | +0.010ms | +1.1% | .328 | -1.6..+6.1% |
| neutral | cached measurement | 1.25µs | 1.34µs | +0.000090ms | +7.5% | .279 | -4.2..+21.5% |
| neutral | measurement after width change | 2.84ms | 2.77ms | -0.070ms | -2.4% | .505 | -6.3..+2.8% |
| neutral | measurement then publication after width change | 5.46ms | 5.63ms | +0.170ms | +3.1% | .083 | -0.1..+6.6% |
| neutral | cached per-glyph inspection | 264.47µs | 256.90µs | -0.008ms | -2.9% | .798 | -7.1..+7.4% |
| neutral | per-glyph inspection after width change | 21.13ms | 21.10ms | -0.030ms | -0.1% | .574 | -5.6..+4.4% |
| neutral | inspection then publication after width change | 16.20ms | 15.65ms | -0.550ms | -3.4% | .442 | -6.9..+5.5% |
| neutral | publication after font-size change | 13.60ms | 13.45ms | -0.150ms | -1.1% | .328 | -2.3..+1.1% |
| neutral | publication after width change | 3.43ms | 3.47ms | +0.040ms | +1.1% | .878 | -3.2..+4.6% |
| neutral | copy per-glyph metrics with outlines after text change | 12.96ms | 12.70ms | -0.260ms | -2.1% | .234 | -3.8..+1.5% |
| neutral | read every outline from an unchanged copy | 2.93ms | 2.91ms | -0.020ms | -0.7% | .442 | -1.7..+0.9% |
| neutral | write 1000 paragraph-order mutations | 164.90µs | 152.53µs | -0.012ms | -7.5% | .279 | -15.3..+106.7% |
| neutral | write 1000 paragraph and text mutations | 409.01µs | 371.64µs | -0.037ms | -9.1% | .065 | -22.3..+0.1% |
| neutral | scattered character assignment in one large paragraph with 8192 paint spans | 45.60ms | 44.87ms | -0.730ms | -1.6% | .105 | -3.0..+0.3% |
| neutral | broad character assignment in one large paragraph with 8192 paint spans | 45.37ms | 44.73ms | -0.640ms | -1.4% | .038 | -3.4..-0.1% |
| neutral | reorder 1000 unchanged retained labels | 6.40ms | 6.09ms | -0.310ms | -4.7% | .003 | -9.4..-1.6% |
| neutral | reorder then edit one of 1000 retained labels | 13.96ms | 13.95ms | -0.010ms | -0.1% | .959 | -1.8..+2.2% |
| neutral | measure 1000 unchanged retained labels | 449.77µs | 416.35µs | -0.033ms | -7.4% | .234 | -12.9..+4.3% |
| neutral | borrow glyphs from 1000 steadily promoted retained labels | 5.02ms | 3.85ms | -1.170ms | -23.4% | .382 | -45.5..+14.4% |
| neutral | edit and measure one of 1000 retained labels | 633.73µs | 617.12µs | -0.017ms | -2.6% | .105 | -9.0..+1.0% |
| neutral | edit and sparsely borrow one of 1000 retained labels | 687.44µs | 677.61µs | -0.010ms | -1.4% | .645 | -4.6..+2.2% |
| neutral | publish 1024 retained Text instances | 38.87ms | 38.35ms | -0.520ms | -1.3% | .234 | -4.1..+0.6% |
| neutral | move 1000 retained TypeGPU labels | 515.50µs | 514.72µs | -0.000780ms | -0.2% | .721 | -5.4..+5.0% |
| neutral | common-safe word-wrap reflow | 804.21µs | 776.66µs | -0.028ms | -3.4% | .028 | -7.2..-0.9% |
| neutral | issue 216 unsafe-boundary word-wrap reflow | 822.86µs | 765.40µs | -0.057ms | -7.0% | .065 | -11.3..+0.3% |
| neutral | long reporter workaround paragraph reflow | 1.84ms | 1.78ms | -0.060ms | -3.2% | .195 | -5.8..+1.4% |
| neutral | long unsafe-boundary paragraph reflow | 1.55ms | 1.51ms | -0.040ms | -2.9% | .083 | -7.9..+0.7% |
| neutral | long unsafe-boundary paragraph text edit | 4.46ms | 4.41ms | -0.050ms | -1.2% | .645 | -2.2..+1.3% |
| skipped: clock-confounded: neutral→slower · 2.45→2.59 | 10 labels first color-only engine publication | ||||||
| skipped: clock-confounded: neutral→slower · 2.54→2.61 | 10 labels last same-length scene publication | ||||||
| skipped: clock-confounded: neutral→slower · 2.44→2.61 | 1000 labels last length-changing scene publication | ||||||
| skipped: clock-confounded: slower→neutral · 2.42→2.54 | publish after color change | ||||||
| skipped: clock-confounded: slower→neutral · 2.60→2.55 | measure 100 unchanged labels | ||||||
| skipped: clock-confounded: neutral→faster · 2.61→2.46 | publication after text change | ||||||
| skipped: clock-confounded: faster→neutral · 2.83→2.96 | write 22k UTF16 text into Wasm request arena | ||||||
| skipped: clock-confounded: neutral→faster · 2.57→2.41 | unchanged character assignment in one large paragraph with 8192 paint spans | ||||||
| skipped: clock-confounded: neutral→faster · 2.53→2.39 | toggle one group containing 1000 retained labels |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Addresses #181. Custom Three material factories receive stage-safe
positionandparagraphPositionnodes.positionis draw-local;paragraphPositionuses paragraph-local, +Y-up coordinates. The default material retains its raw vertex graph so Bitmap and decoration shaders preserve vertex builtins through TypeGPU-to-TSL on WebGL2. Documentation covers direct/indexed transforms, displacement, paragraph texture mapping and inspection bounds.Built-in default cache misses now skip two unused
VaryingNodeconstructions. Custom factories retain both stage-safe nodes,undefinedstill falls back to the raw default, and the selected material is owned once. Cache hits are unchanged. A custom cache miss now constructs one additional context object; no end-to-end latency gain is claimed.Validation
Current head
4bc98799passed the complete local Glyph check (tests, TypeScript/declaration consumers, lint, formatting and Rust fmt/Clippy), eight focused shader tests, both actual downstream example-package checks (13 raster tests and five renderer tests plus typechecks), and the 16-case material-position pixel proof through an isolated Portless route. The rendered proof covers stable TSL/TypeGPU, WebGPU/WebGL2, direct/indexed transforms and displacement. Owned listeners were closed. Fresh scoped Sol High review and integrating review found no actionable defect. The first refinement headf985944bfailed CI because the private helper mapped over the open material-context map; a downstream raster augmentation lacks its built-in-only field. The helper now names only its four built-in callers. All 238 JS, five Wasm and 227 declaration bytes match that prior candidate, so its rendered proof is reused. The failed CI receipt is retained. Exact-head4bc98799passes all five jobs in run 39293379442: Static checks, Docs report, Package size report, Check and Package performance report. The eight-block Ubuntu full-suite report classifies 3 faster / 0 slower / 96 neutral, with nine skipped workloads, against main canary0.1.0-canary-4e25a96a-20261015. These classifications retain the workflow’s materiality/statistical rules and exclusions; neutral is not proof of zero effect. This new result is separate from the historical measurements below; no isolated material-realization latency gain is claimed.Against a separate exact
e4498972build, onlythree/internal/material-realizer.jschanges among 238 JavaScript members: 10,862→10,725 raw bytes and 3,993→3,959 Node 24 gzip9 bytes. All 227 declaration members match byte-for-byte. Tracked Rust, generated ABI/Unicode and build-policy/dependency inputs match. Text-shaper and Slug Wasm match; three rebuilt baker Wasms differ for an unestablished reason. A subsequent authenticated binary-section inspection resolves all imported/exported function parameter/result vectors and finds them identical across all five modules. This is not code-semantic, memory-layout or lifetime equivalence proof. No baseline binaries were substituted. Verification is summarized in the committed change log; raw builds, signature receipts and logs remain on this machine.Performance admission remains open
Previous-head CI had 100 comparable workloads: 17 faster, 15 slower, 68 neutral, with eight exclusions (seven clock-confounded and one mode mismatch). Slower rows included unchanged-label inspection and width reflow. That comparison used the published main canary; its authenticated packages differed only in the material-realizer executable member. Cached inspection/copy calls do not reach the changed cold path, but indirect effects were not ruled out.
The previous coordinated local eight-block wave classified all six smoke cases neutral. Its glyphs comparison had two faster, one slower, eight neutral and one clock exclusion; the slower edit/copy row was +15.6% with slower candidate clocks and more GC. Width inspection was neutral (+2.7%, p=.234); reversed cold cases were three neutral/one exclusion. These historical observations do not establish a material-path slowdown or clear every flag, and are not reassigned to the current refinement. CPU timings do not establish GPU frame performance. Keep this draft unmerged.