ripemd160: use bits.Rotate for rotates
Replace x << k | x >> (32 - k) with bits.RotateLeft32 and add
benchmark to check performance. This makes code cleaner and improves performance:
MillionA-6 34.8ms ± 1% 26.7ms ± 1% -23.41% (p=0.000 n=10+10)
Change-Id: I1b0c45f5ea12c7b53ab1ec5a26efc1903555fa66
Reviewed-on: https://go-review.googlesource.com/97915
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
diff --git a/ripemd160/ripemd160_test.go b/ripemd160/ripemd160_test.go
index 5df1b25..a1fbffd 100644
--- a/ripemd160/ripemd160_test.go
+++ b/ripemd160/ripemd160_test.go
@@ -50,15 +50,23 @@
}
}
-func TestMillionA(t *testing.T) {
+func millionA() string {
md := New()
for i := 0; i < 100000; i++ {
io.WriteString(md, "aaaaaaaaaa")
}
- out := "52783243c1697bdbe16d37f97f68f08325dc1528"
- s := fmt.Sprintf("%x", md.Sum(nil))
- if s != out {
+ return fmt.Sprintf("%x", md.Sum(nil))
+}
+
+func TestMillionA(t *testing.T) {
+ const out = "52783243c1697bdbe16d37f97f68f08325dc1528"
+ if s := millionA(); s != out {
t.Fatalf("RIPEMD-160 (1 million 'a') = %s, expected %s", s, out)
}
- md.Reset()
+}
+
+func BenchmarkMillionA(b *testing.B) {
+ for i := 0; i < b.N; i++ {
+ millionA()
+ }
}
diff --git a/ripemd160/ripemd160block.go b/ripemd160/ripemd160block.go
index 7bc8e6c..e0edc02 100644
--- a/ripemd160/ripemd160block.go
+++ b/ripemd160/ripemd160block.go
@@ -8,6 +8,10 @@
package ripemd160
+import (
+ "math/bits"
+)
+
// work buffer indices and roll amounts for one line
var _n = [80]uint{
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
@@ -59,16 +63,16 @@
i := 0
for i < 16 {
alpha = a + (b ^ c ^ d) + x[_n[i]]
- s := _r[i]
- alpha = (alpha<<s | alpha>>(32-s)) + e
- beta = c<<10 | c>>22
+ s := int(_r[i])
+ alpha = bits.RotateLeft32(alpha, s) + e
+ beta = bits.RotateLeft32(c, 10)
a, b, c, d, e = e, alpha, b, beta, d
// parallel line
alpha = aa + (bb ^ (cc | ^dd)) + x[n_[i]] + 0x50a28be6
- s = r_[i]
- alpha = (alpha<<s | alpha>>(32-s)) + ee
- beta = cc<<10 | cc>>22
+ s = int(r_[i])
+ alpha = bits.RotateLeft32(alpha, s) + ee
+ beta = bits.RotateLeft32(cc, 10)
aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
i++
@@ -77,16 +81,16 @@
// round 2
for i < 32 {
alpha = a + (b&c | ^b&d) + x[_n[i]] + 0x5a827999
- s := _r[i]
- alpha = (alpha<<s | alpha>>(32-s)) + e
- beta = c<<10 | c>>22
+ s := int(_r[i])
+ alpha = bits.RotateLeft32(alpha, s) + e
+ beta = bits.RotateLeft32(c, 10)
a, b, c, d, e = e, alpha, b, beta, d
// parallel line
alpha = aa + (bb&dd | cc&^dd) + x[n_[i]] + 0x5c4dd124
- s = r_[i]
- alpha = (alpha<<s | alpha>>(32-s)) + ee
- beta = cc<<10 | cc>>22
+ s = int(r_[i])
+ alpha = bits.RotateLeft32(alpha, s) + ee
+ beta = bits.RotateLeft32(cc, 10)
aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
i++
@@ -95,16 +99,16 @@
// round 3
for i < 48 {
alpha = a + (b | ^c ^ d) + x[_n[i]] + 0x6ed9eba1
- s := _r[i]
- alpha = (alpha<<s | alpha>>(32-s)) + e
- beta = c<<10 | c>>22
+ s := int(_r[i])
+ alpha = bits.RotateLeft32(alpha, s) + e
+ beta = bits.RotateLeft32(c, 10)
a, b, c, d, e = e, alpha, b, beta, d
// parallel line
alpha = aa + (bb | ^cc ^ dd) + x[n_[i]] + 0x6d703ef3
- s = r_[i]
- alpha = (alpha<<s | alpha>>(32-s)) + ee
- beta = cc<<10 | cc>>22
+ s = int(r_[i])
+ alpha = bits.RotateLeft32(alpha, s) + ee
+ beta = bits.RotateLeft32(cc, 10)
aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
i++
@@ -113,16 +117,16 @@
// round 4
for i < 64 {
alpha = a + (b&d | c&^d) + x[_n[i]] + 0x8f1bbcdc
- s := _r[i]
- alpha = (alpha<<s | alpha>>(32-s)) + e
- beta = c<<10 | c>>22
+ s := int(_r[i])
+ alpha = bits.RotateLeft32(alpha, s) + e
+ beta = bits.RotateLeft32(c, 10)
a, b, c, d, e = e, alpha, b, beta, d
// parallel line
alpha = aa + (bb&cc | ^bb&dd) + x[n_[i]] + 0x7a6d76e9
- s = r_[i]
- alpha = (alpha<<s | alpha>>(32-s)) + ee
- beta = cc<<10 | cc>>22
+ s = int(r_[i])
+ alpha = bits.RotateLeft32(alpha, s) + ee
+ beta = bits.RotateLeft32(cc, 10)
aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
i++
@@ -131,16 +135,16 @@
// round 5
for i < 80 {
alpha = a + (b ^ (c | ^d)) + x[_n[i]] + 0xa953fd4e
- s := _r[i]
- alpha = (alpha<<s | alpha>>(32-s)) + e
- beta = c<<10 | c>>22
+ s := int(_r[i])
+ alpha = bits.RotateLeft32(alpha, s) + e
+ beta = bits.RotateLeft32(c, 10)
a, b, c, d, e = e, alpha, b, beta, d
// parallel line
alpha = aa + (bb ^ cc ^ dd) + x[n_[i]]
- s = r_[i]
- alpha = (alpha<<s | alpha>>(32-s)) + ee
- beta = cc<<10 | cc>>22
+ s = int(r_[i])
+ alpha = bits.RotateLeft32(alpha, s) + ee
+ beta = bits.RotateLeft32(cc, 10)
aa, bb, cc, dd, ee = ee, alpha, bb, beta, dd
i++
