unicode/utf8: make acceptRanges bigger

This avoids bounds checks in the calling code.
The nominal increased size of the array in the binary
is compensated for by the decreased size of the functions that call it.

The benchmark changes are a bit scattered, but overall positive.

name                                 old time/op  new time/op  delta
RuneCountTenASCIIChars-8             8.86ns ± 4%  7.93ns ± 5%  -10.45%  (p=0.000 n=45+49)
RuneCountTenJapaneseChars-8          38.2ns ± 2%  37.2ns ± 1%   -2.63%  (p=0.000 n=44+41)
RuneCountInStringTenASCIIChars-8     7.82ns ± 2%  8.70ns ± 2%  +11.19%  (p=0.000 n=43+43)
RuneCountInStringTenJapaneseChars-8  39.3ns ± 9%  40.0ns ± 5%   +1.59%  (p=0.043 n=50+50)
ValidTenASCIIChars-8                 8.68ns ± 5%  8.74ns ± 5%     ~     (p=0.070 n=50+48)
ValidTenJapaneseChars-8              34.1ns ± 5%  36.8ns ± 4%   +8.09%  (p=0.000 n=45+50)
ValidStringTenASCIIChars-8           9.76ns ± 7%  8.33ns ± 3%  -14.59%  (p=0.000 n=48+47)
ValidStringTenJapaneseChars-8        37.7ns ± 8%  36.5ns ± 5%   -3.12%  (p=0.011 n=50+47)
EncodeASCIIRune-8                    2.60ns ± 1%  2.59ns ± 2%   -0.24%  (p=0.018 n=43+36)
EncodeJapaneseRune-8                 3.75ns ± 2%  4.56ns ± 6%  +21.71%  (p=0.000 n=41+50)
DecodeASCIIRune-8                    2.59ns ± 2%  2.59ns ± 2%     ~     (p=0.350 n=44+41)
DecodeJapaneseRune-8                 4.29ns ± 2%  4.31ns ± 2%   +0.61%  (p=0.001 n=48+39)
FullASCIIRune-8                      0.87ns ± 6%  0.29ns ± 5%  -67.31%  (p=0.000 n=49+43)
FullJapaneseRune-8                   0.65ns ± 6%  0.65ns ± 4%     ~     (p=0.375 n=50+49)
[Geo mean]                           7.02ns       6.51ns        -7.19%

Change-Id: I8d5d69c8d33ce2bff94785fba39a2203f9315cb0
Reviewed-on: https://go-review.googlesource.com/c/go/+/173537
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

diff --git a/src/unicode/utf8/utf8.go b/src/unicode/utf8/utf8.go
index db845ab2f2d321822d029ba2495f1b727232d544..b351cc351f914a0f78da13824ba8a7e289486f85 100644 (file)
--- a/src/unicode/utf8/utf8.go
+++ b/src/unicode/utf8/utf8.go
@@ -89,7 +89,8 @@ type acceptRange struct {
        hi uint8 // highest value for second byte.
 }
 
-var acceptRanges = [...]acceptRange{
+// acceptRanges has size 16 to avoid bounds checks in the code that uses it.
+var acceptRanges = [16]acceptRange{
        0: {locb, hicb},
        1: {0xA0, hicb},
        2: {locb, 0x9F},