mirror of
https://github.com/ruby/ruby.git
synced 2026-08-05 23:40:42 +08:00
35b7652058
When every element is a String sharing one ASCII-compatible fast-path encoding (UTF-8/US-ASCII/ASCII-8BIT) and the separator is compatible, build the result with a single memcpy pass instead of appending element by element through rb_str_buf_append. Other cases fall back to the existing path.
45 lines
2.2 KiB
YAML
45 lines
2.2 KiB
YAML
prelude: |
|
|
# --- must NOT take any fast path: prove added precompute-loop checks don't regress ---
|
|
# all multibyte UTF-8 (VALID coderange). 7-bit gate bails at elem 0; same-encoding gate TAKES it.
|
|
mb_utf8 = Array.new(1000) { "café" }
|
|
# ASCII except a trailing multibyte elem: WORST CASE for 7-bit gate (scans all, then falls back).
|
|
tail_mb = Array.new(1000) { "a" * 8 }; tail_mb[-1] = "café"
|
|
# non-String elements -> ary_join_1 (to_s). Gate bails at type check before coderange.
|
|
nonstr_int = Array.new(1000) { |i| i }
|
|
# ASCII strings except a trailing Integer: precompute scans strings, then mixed fallback.
|
|
mixed_tail = (Array.new(999) { "a" * 8 } << 42)
|
|
# UTF-16LE: not ASCII-compatible. Both gates bail immediately.
|
|
utf16 = Array.new(1000) { "ab".encode("UTF-16LE") }
|
|
# nested arrays -> recursive fallback.
|
|
nested = Array.new(500) { [1, 2] }
|
|
|
|
# 7-bit ASCII elements but a multibyte separator (3-byte em dash). Both gates bail at sep check.
|
|
ascii_elems = Array.new(1000) { "a" * 8 }
|
|
sep_mb = "—"
|
|
|
|
# --- fast-path ELIGIBLE variants: confirm benefit generalizes beyond single chars / find crossover ---
|
|
# frozen ASCII elements (read-only path).
|
|
frozen_elems = Array.new(1000) { ("a" * 8).freeze }
|
|
# large elements: where memcpy should dominate and the win decays toward 1x.
|
|
big_e256 = Array.new(1000) { "a" * 256 }
|
|
big_e1k = Array.new(1000) { "a" * 1024 }
|
|
big_e4k = Array.new(1000) { "a" * 4096 }
|
|
|
|
benchmark:
|
|
# regression (fallback / gate worst-case) — all use a 1-byte ASCII separator unless noted
|
|
reg_multibyte_utf8: mb_utf8.join(" ")
|
|
reg_tail_multibyte: tail_mb.join(" ")
|
|
reg_nonstring_int: nonstr_int.join(" ")
|
|
reg_mixed_tail_int: mixed_tail.join(" ")
|
|
reg_utf16: utf16.join(" ".encode("UTF-16LE"))
|
|
reg_nested: nested.join(" ")
|
|
reg_multibyte_sep: ascii_elems.join(sep_mb)
|
|
|
|
# fast-path eligible variants
|
|
fp_frozen: frozen_elems.join(" ")
|
|
fp_nosep: ascii_elems.join
|
|
fp_multichar_sep: ascii_elems.join(", ")
|
|
fp_big_e256: big_e256.join(" ")
|
|
fp_big_e1k: big_e1k.join(" ")
|
|
fp_big_e4k: big_e4k.join(" ")
|