perf: cache grapheme widths in Window.print word wrap

[dyxushuai]

Problem

Window.print(.word) computed gwidth twice per word: once for the word width check and again per-grapheme while writing cells. This duplicates grapheme width work and adds overhead in word-wrapped text.

Fix

Cache grapheme slices + widths while computing the word width, then reuse that cache when writing cells. If the fixed buffer fills up, reuse the cached prefix and fall back to the original per-grapheme path for the remainder.

Note: Caching uses a fixed 4KB stack buffer to avoid heap allocation for the per-word cache
(ArrayListUnmanaged append).

Bench (local, zig build bench, iterations=200, 80x24)

Baseline = print + extra per-word gwidth pass to mirror the old double-work
Cached = current print implementation

Case	Baseline ns/frame	Cached ns/frame	Improvement	Speedup
Small	81,281	67,043	-17.5%	1.21x
Medium	318,911	264,232	-17.1%	1.21x
Large	632,170	526,237	-16.8%	1.20x
Overflow	4,600,554	3,894,809	-15.3%	1.18x

Improvement: Small -17.5% (1.21x); Medium -17.1% (1.21x); Large -16.8% (1.20x); Overflow -15.3% (1.18x).

Tests

• zig build test
• zig build bench

[Copilot]

Pull request overview

This PR optimizes word-wrapping performance in Window.print by caching grapheme slice positions and widths during the initial width calculation pass, eliminating redundant gwidth() calls when rendering. The optimization uses a fixed 4KB stack buffer to avoid heap allocations and falls back to the original per-grapheme iteration if the buffer overflows.

Key Changes:

• Introduced a caching mechanism that stores grapheme boundaries and widths in a fixed buffer during width calculation
• Refactored the word-rendering path into two branches: a cached path that reuses computed widths, and a fallback path for cache overflow scenarios
• Added comprehensive benchmarks demonstrating ~18-22% performance improvement across different text sizes

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/Window.zig	Implements grapheme width caching with WordPiece struct and FixedBufferAllocator, adds cached and fallback rendering paths
bench/bench.zig	Adds benchmark infrastructure and test cases (small/medium/large) to measure the caching optimization, includes helper iterators for baseline comparison

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.