Use an index in the loop over the subtags of a combined tag

[NullVoxPopuli-ai-agent]

The loop over the subtags of a combined tag uses an index in place of for...of. The change is three lines in validators.ts.

Alone, this change has no effect that the reactivity benchmark can separate from noise: the weighted mean is 1.0x of main. It is a part of #21650, and it merges alone.

case	main	this PR
kairo: mux	20.23 µs	19.72 µs (1.0x)
batch: 10 writes, 1 output	667 ns	664 ns (1.0x)
rows: 1000 rows, write 1	7.19 µs	7.10 µs (1.0x)
weighted geometric mean	1.0x	1.0x

What changes

• The loop reads subtag[i] with an index. A for...of loop over an array uses the iterator protocol.
• The loop keeps Math.max. A > comparison ignores the NaN revision of the volatile tag, and a cache that reads that tag then runs one time only.

All 20 cases

case	main	this PR
propagate: 1 chains x 1 deep	124 ns	121 ns (1.0x)
propagate: 10 chains x 10 deep	5.51 µs	5.47 µs (1.0x)
propagate: 100 chains x 100 deep	724.68 µs	708.19 µs (1.0x)
propagate: 1 chains x 1000 deep	57.18 µs	57.29 µs (1.0x)
propagate: 1000 chains x 1 deep	103.01 µs	101.40 µs (1.0x)
kairo: avoidable propagation	382 ns	361 ns (0.9x)
kairo: broad propagation	7.35 µs	7.10 µs (1.0x)
kairo: deep propagation	3.09 µs	3.08 µs (1.0x)
kairo: diamond	428 ns	424 ns (1.0x)
kairo: mux	20.23 µs	19.72 µs (1.0x)
kairo: repeated observers	365 ns	323 ns (0.9x)
kairo: triangle	755 ns	689 ns (0.9x)
kairo: unstable	574 ns	570 ns (1.0x)
rows: 1000 rows, write 1	7.19 µs	7.10 µs (1.0x)
rows: 1000 rows, write all	118.18 µs	118.70 µs (1.0x)
batch: 10 writes, 1 output	667 ns	664 ns (1.0x)
avoidable: write the same value	6 ns	6 ns (1.0x)
create: 1000 signals	17.84 µs	17.85 µs (1.0x)
create: 1000 computeds, read each	41.11 µs	37.95 µs (0.9x)
create: 1000 outputs	50.30 µs	49.03 µs (1.0x)
weighted geometric mean	1.0x	1.0x

The benchmark is https://github.com/NullVoxPopuli-ai-agent/ember-reactivity-bench. One measurement is the writes of one frame, then one flush that brings every output up to date. Each case of each build runs in its own process, pinned to one core. The numbers are the median of 6 mirrored rounds on Node 24.20.

The run has main, each of the five branches, and #21650 as columns: table and method.

Tests

• The full suite passes locally: 9495 pass, 18 skipped, 0 failed.
• type-check:internals, ESLint and Prettier pass.
• No new test. The tests of combined tags and of the volatile tag cover the loop.

The parts of #21650

#21650 has all five changes in one branch. It stays open as the reference for the combined numbers. The five branches merge with no conflict, and their merge is the same code as the head of #21650.

PR	change	reactivity benchmark, mean against main
#21663	Pool the trackers of tracking frames	0.6x
#21664	Reuse the combined tag of a tracking frame. Needs #21663	0.6x, with the pool
#21665 (this PR)	Index loop over the subtags of a combined tag	1.0x
#21666	Make the functions of a TrackedValue on their first use	0.9x, create is 0.5x
#21667	Store 0 first in the value field of a TrackedValue	1.0x, removes one deopt
#21650	all five	0.6x

Rendering was measured for the combined change only: pnpm bench shows 2.1% less script time, in this comment on #21650.

🤖 Generated with Claude Code