strings: use AVX2 for Index if available

IndexHard4-4      1.50ms ± 2%  0.71ms ± 0%  -52.36%  (p=0.000 n=20+19)

This also fixes a bug, that caused a string of length 16 to use
two 8-byte comparisons instead of one 16-byte. And adds a test for
cases when partial_match fails.

Change-Id: I1ee8fc4e068bb36c95c45de78f067c822c0d9df0
Reviewed-on: https://go-review.googlesource.com/22551
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

2 files changed, 26 insertions, 1 deletions

diff --git a/src/strings/strings_amd64.go b/src/strings/strings_amd64.go
index 434e2e9eb3..5e26ee2c97 100644
--- a/src/strings/strings_amd64.go
+++ b/src/strings/strings_amd64.go
@@ -9,7 +9,17 @@ package strings
 // indexShortStr returns the index of the first instance of c in s, or -1 if c is not present in s.
 // indexShortStr requires 2 <= len(c) <= shortStringLen
 func indexShortStr(s, c string) int // ../runtime/asm_$GOARCH.s
-const shortStringLen = 31
+func supportAVX2() bool             // ../runtime/asm_$GOARCH.s
+
+var shortStringLen int
+
+func init() {
+	if supportAVX2() {
+		shortStringLen = 63
+	} else {
+		shortStringLen = 31
+	}
+}
 
 // Index returns the index of the first instance of sep in s, or -1 if sep is not present in s.
 func Index(s, sep string) int {
diff --git a/src/strings/strings_test.go b/src/strings/strings_test.go
index 5fdf59c88d..cf7fde5bbd 100644
--- a/src/strings/strings_test.go
+++ b/src/strings/strings_test.go
@@ -86,32 +86,44 @@ var indexTests = []IndexTest{
 	{"32145678", "01234567", -1},
 	{"01234567", "01234567", 0},
 	{"x01234567", "01234567", 1},
+	{"x0123456x01234567", "01234567", 9},
 	{"xx01234567"[:9], "01234567", -1},
 	{"", "0123456789", -1},
 	{"3214567844", "0123456789", -1},
 	{"0123456789", "0123456789", 0},
 	{"x0123456789", "0123456789", 1},
+	{"x012345678x0123456789", "0123456789", 11},
 	{"xyz0123456789"[:12], "0123456789", -1},
 	{"x01234567x89", "0123456789", -1},
 	{"", "0123456789012345", -1},
 	{"3214567889012345", "0123456789012345", -1},
 	{"0123456789012345", "0123456789012345", 0},
 	{"x0123456789012345", "0123456789012345", 1},
+	{"x012345678901234x0123456789012345", "0123456789012345", 17},
 	{"", "01234567890123456789", -1},
 	{"32145678890123456789", "01234567890123456789", -1},
 	{"01234567890123456789", "01234567890123456789", 0},
 	{"x01234567890123456789", "01234567890123456789", 1},
+	{"x0123456789012345678x01234567890123456789", "01234567890123456789", 21},
 	{"xyz01234567890123456789"[:22], "01234567890123456789", -1},
 	{"", "0123456789012345678901234567890", -1},
 	{"321456788901234567890123456789012345678911", "0123456789012345678901234567890", -1},
 	{"0123456789012345678901234567890", "0123456789012345678901234567890", 0},
 	{"x0123456789012345678901234567890", "0123456789012345678901234567890", 1},
+	{"x012345678901234567890123456789x0123456789012345678901234567890", "0123456789012345678901234567890", 32},
 	{"xyz0123456789012345678901234567890"[:33], "0123456789012345678901234567890", -1},
 	{"", "01234567890123456789012345678901", -1},
 	{"32145678890123456789012345678901234567890211", "01234567890123456789012345678901", -1},
 	{"01234567890123456789012345678901", "01234567890123456789012345678901", 0},
 	{"x01234567890123456789012345678901", "01234567890123456789012345678901", 1},
+	{"x0123456789012345678901234567890x01234567890123456789012345678901", "01234567890123456789012345678901", 33},
 	{"xyz01234567890123456789012345678901"[:34], "01234567890123456789012345678901", -1},
+	{"xxxxxx012345678901234567890123456789012345678901234567890123456789012", "012345678901234567890123456789012345678901234567890123456789012", 6},
+	{"", "0123456789012345678901234567890123456789", -1},
+	{"xx012345678901234567890123456789012345678901234567890123456789012", "0123456789012345678901234567890123456789", 2},
+	{"xx012345678901234567890123456789012345678901234567890123456789012"[:41], "0123456789012345678901234567890123456789", -1},
+	{"xx012345678901234567890123456789012345678901234567890123456789012", "0123456789012345678901234567890123456xxx", -1},
+	{"xx0123456789012345678901234567890123456789012345678901234567890120123456789012345678901234567890123456xxx", "0123456789012345678901234567890123456xxx", 65},
 }
 
 var lastIndexTests = []IndexTest{
@@ -1315,6 +1327,9 @@ func benchmarkCountHard(b *testing.B, sep string) {
 func BenchmarkIndexHard1(b *testing.B) { benchmarkIndexHard(b, "<>") }
 func BenchmarkIndexHard2(b *testing.B) { benchmarkIndexHard(b, "</pre>") }
 func BenchmarkIndexHard3(b *testing.B) { benchmarkIndexHard(b, "<b>hello world</b>") }
+func BenchmarkIndexHard4(b *testing.B) {
+	benchmarkIndexHard(b, "<pre><b>hello</b><strong>world</strong></pre>")
+}
 
 func BenchmarkLastIndexHard1(b *testing.B) { benchmarkLastIndexHard(b, "<>") }
 func BenchmarkLastIndexHard2(b *testing.B) { benchmarkLastIndexHard(b, "</pre>") }