regexp: avoid allocation of input interface
Matters most for small inputs, because there is no real work
to amortize the allocation effort against.
benchmark old ns/op new ns/op delta
BenchmarkLiteral 613 473 -22.84%
BenchmarkNotLiteral 4981 4931 -1.00%
BenchmarkMatchClass 7289 7122 -2.29%
BenchmarkMatchClass_InRange 6618 6663 +0.68%
BenchmarkReplaceAll 7843 7233 -7.78%
BenchmarkAnchoredLiteralShortNonMatch 329 228 -30.70%
BenchmarkAnchoredLiteralLongNonMatch 322 228 -29.19%
BenchmarkAnchoredShortMatch 838 715 -14.68%
BenchmarkAnchoredLongMatch 824 715 -13.23%
benchmark old MB/s new MB/s speedup
BenchmarkMatchEasy0_32 119.73 196.61 1.64x
BenchmarkMatchEasy0_1K 540.58 538.33 1.00x
BenchmarkMatchEasy0_32K 732.57 714.00 0.97x
BenchmarkMatchEasy0_1M 726.44 708.36 0.98x
BenchmarkMatchEasy0_32M 707.77 691.45 0.98x
BenchmarkMatchEasy1_32 102.12 136.11 1.33x
BenchmarkMatchEasy1_1K 298.31 307.04 1.03x
BenchmarkMatchEasy1_32K 273.56 274.43 1.00x
BenchmarkMatchEasy1_1M 268.42 269.23 1.00x
BenchmarkMatchEasy1_32M 266.15 267.34 1.00x
BenchmarkMatchMedium_32 2.53 3.38 1.34x
BenchmarkMatchMedium_1K 9.37 9.57 1.02x
BenchmarkMatchMedium_32K 9.29 9.67 1.04x
BenchmarkMatchMedium_1M 9.42 9.66 1.03x
BenchmarkMatchMedium_32M 9.41 9.62 1.02x
BenchmarkMatchHard_32 6.66 6.75 1.01x
BenchmarkMatchHard_1K 6.81 6.85 1.01x
BenchmarkMatchHard_32K 6.79 6.85 1.01x
BenchmarkMatchHard_1M 6.82 6.83 1.00x
BenchmarkMatchHard_32M 6.80 6.80 1.00x
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/
5453076