cmd/compile/internal/ssa: combine 2 byte loads + shifts into word load + rolw 8 on AMD64

... and same for stores. This does for binary.BigEndian.Uint16() what
was already done for Uint32 and Uint64 with BSWAP in 10f75748 (CL 32222).

Here is how generated code changes e.g. for the following function
(omitting saying the same prologue/epilogue):

	func get16(b [2]byte) uint16 {
		return binary.BigEndian.Uint16(b[:])
	}

"".get16 t=1 size=21 args=0x10 locals=0x0

	// before
        0x0000 00000 (x.go:15)  MOVBLZX "".b+9(FP), AX
        0x0005 00005 (x.go:15)  MOVBLZX "".b+8(FP), CX
        0x000a 00010 (x.go:15)  SHLL    $8, CX
        0x000d 00013 (x.go:15)  ORL     CX, AX

	// after
	0x0000 00000 (x.go:15)	MOVWLZX	"".b+8(FP), AX
	0x0005 00005 (x.go:15)	ROLW	$8, AX

encoding/binary is speedup overall a bit:

name                    old time/op    new time/op    delta
ReadSlice1000Int32s-4     4.83µs ± 0%    4.83µs ± 0%     ~     (p=0.206 n=4+5)
ReadStruct-4              1.29µs ± 2%    1.28µs ± 1%   -1.27%  (p=0.032 n=4+5)
ReadInts-4                 384ns ± 1%     385ns ± 1%     ~     (p=0.968 n=4+5)
WriteInts-4                534ns ± 3%     526ns ± 0%   -1.54%  (p=0.048 n=4+5)
WriteSlice1000Int32s-4    5.02µs ± 0%    5.11µs ± 3%     ~     (p=0.175 n=4+5)
PutUint16-4               0.59ns ± 0%    0.49ns ± 2%  -16.95%  (p=0.016 n=4+5)
PutUint32-4               0.52ns ± 0%    0.52ns ± 0%     ~     (all equal)
PutUint64-4               0.53ns ± 0%    0.53ns ± 0%     ~     (all equal)
PutUvarint32-4            19.9ns ± 0%    19.9ns ± 1%     ~     (p=0.556 n=4+5)
PutUvarint64-4            54.5ns ± 1%    54.2ns ± 0%     ~     (p=0.333 n=4+5)

name                    old speed      new speed      delta
ReadSlice1000Int32s-4    829MB/s ± 0%   828MB/s ± 0%     ~     (p=0.190 n=4+5)
ReadStruct-4            58.0MB/s ± 2%  58.7MB/s ± 1%   +1.30%  (p=0.032 n=4+5)
ReadInts-4              78.0MB/s ± 1%  77.8MB/s ± 1%     ~     (p=0.968 n=4+5)
WriteInts-4             56.1MB/s ± 3%  57.0MB/s ± 0%     ~     (p=0.063 n=4+5)
WriteSlice1000Int32s-4   797MB/s ± 0%   783MB/s ± 3%     ~     (p=0.190 n=4+5)
PutUint16-4             3.37GB/s ± 0%  4.07GB/s ± 2%  +20.83%  (p=0.016 n=4+5)
PutUint32-4             7.73GB/s ± 0%  7.72GB/s ± 0%     ~     (p=0.556 n=4+5)
PutUint64-4             15.1GB/s ± 0%  15.1GB/s ± 0%     ~     (p=0.905 n=4+5)
PutUvarint32-4           201MB/s ± 0%   201MB/s ± 0%     ~     (p=0.905 n=4+5)
PutUvarint64-4           147MB/s ± 1%   147MB/s ± 0%     ~     (p=0.286 n=4+5)

( "a bit" only because most of the time is spent in reflection-like things
  there, not actual bytes decoding. Even for direct PutUint16 benchmark the
  looping adds overhead and lowers visible benefit. For code-generated encoders /
  decoders actual effect is more than 20% )

Adding Uint32 and Uint64 raw benchmarks too for completeness.

NOTE I had to adjust load-combining rule for bswap case to match first 2 bytes
loads as result of "2-bytes load+shift" -> "loadw + rorw 8" rewrite. Reason is:
for loads+shift, even e.g. into uint16 var

	var b []byte
	var v uin16
	v = uint16(b[1]) | uint16(b[0])<<8

the compiler eventually generates L(ong) shift - SHLLconst [8], probably
because it is more straightforward / other reasons to work on the whole
register. This way 2 bytes rewriting rule is using SHLLconst (not SHLWconst) in
its pattern, and then it always gets matched first, even if 2-byte rule comes
syntactically after 4-byte rule in AMD64.rules because 4-bytes rule seemingly
needs more applyRewrite() cycles to trigger. If 2-bytes rule gets matched for
inner half of

	var b []byte
	var v uin32
	v = uint32(b[3]) | uint32(b[2])<<8 | uint32(b[1])<<16 | uint32(b[0])<<24

and we keep 4-byte load rule unchanged, the result will be MOVW + RORW $8 and
then series of byte loads and shifts - not one MOVL + BSWAPL.

There is no such problem for stores: there compiler, since it probably knows
store destination is 2 bytes wide, uses SHRWconst 8 (not SHRLconst 8) and thus
2-byte store rule is not a subset of rule for 4-byte stores.

Fixes #17151  (int16 was last missing piece there)

Change-Id: Idc03ba965bfce2b94fef456b02ff6742194748f6
Reviewed-on: https://go-review.googlesource.com/34636Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

src/cmd/compile/internal/gc/asm_test.go

@@ -221,6 +221,38 @@ func f(b []byte, i int, v uint32) {
 `,
 		[]string{"\tBSWAPL\t"},
 	},
+	{"amd64", "linux", `
+import "encoding/binary"
+func f(b []byte) uint16 {
+	return binary.BigEndian.Uint16(b)
+}
+`,
+		[]string{"\tROLW\t\\$8,"},
+	},
+	{"amd64", "linux", `
+import "encoding/binary"
+func f(b []byte, i int) uint16 {
+	return binary.BigEndian.Uint16(b[i:])
+}
+`,
+		[]string{"\tROLW\t\\$8,"},
+	},
+	{"amd64", "linux", `
+import "encoding/binary"
+func f(b []byte, v uint16) {
+	binary.BigEndian.PutUint16(b, v)
+}
+`,
+		[]string{"\tROLW\t\\$8,"},
+	},
+	{"amd64", "linux", `
+import "encoding/binary"
+func f(b []byte, i int, v uint16) {
+	binary.BigEndian.PutUint16(b[i:], v)
+}
+`,
+		[]string{"\tROLW\t\\$8,"},
+	},
 	{"386", "linux", `
 import "encoding/binary"
 func f(b []byte) uint32 {

src/cmd/compile/internal/ssa/gen/AMD64.rules

@@ -1558,58 +1558,72 @@
   && clobber(o5)
   -> @mergePoint(b,x0,x1,x2,x3,x4,x5,x6,x7) (MOVQloadidx1 <v.Type> [i] {s} p idx mem)
 
-// Combine byte loads + shifts into larger (unaligned) loads + bswap
-(ORL o1:(ORL o0:(ORL
+// Combine 2 byte loads + shifts into (unaligned) word load + rolw 8
+(ORL
                        x0:(MOVBload [i] {s} p mem)
     s0:(SHLLconst [8]  x1:(MOVBload [i-1] {s} p mem)))
-    s1:(SHLLconst [16] x2:(MOVBload [i-2] {s} p mem)))
-    s2:(SHLLconst [24] x3:(MOVBload [i-3] {s} p mem)))
   && x0.Uses == 1
   && x1.Uses == 1
+  && s0.Uses == 1
+  && mergePoint(b,x0,x1) != nil
+  && clobber(x0)
+  && clobber(x1)
+  && clobber(s0)
+  -> @mergePoint(b,x0,x1) (ROLWconst <v.Type> [8] (MOVWload [i-1] {s} p mem))
+
+(ORL
+                       x0:(MOVBloadidx1 [i] {s} p idx mem)
+    s0:(SHLLconst [8]  x1:(MOVBloadidx1 [i-1] {s} p idx mem)))
+  && x0.Uses == 1
+  && x1.Uses == 1
+  && s0.Uses == 1
+  && mergePoint(b,x0,x1) != nil
+  && clobber(x0)
+  && clobber(x1)
+  && clobber(s0)
+  -> @mergePoint(b,x0,x1) (ROLWconst <v.Type> [8] (MOVWloadidx1 <v.Type> [i-1] {s} p idx mem))
+
+// Combine byte loads + shifts into larger (unaligned) loads + bswap
+// (for L version first 2 bytes loads are matched as result of above 2-bytes load+shift rewrite)
+(ORL o1:(ORL o0:(ROLWconst [8] x01:(MOVWload [i1] {s} p mem))
+    s1:(SHLLconst [16] x2:(MOVBload [i1-1] {s} p mem)))
+    s2:(SHLLconst [24] x3:(MOVBload [i1-2] {s} p mem)))
+  && x01.Uses == 1
   && x2.Uses == 1
   && x3.Uses == 1
-  && s0.Uses == 1
   && s1.Uses == 1
   && s2.Uses == 1
   && o0.Uses == 1
   && o1.Uses == 1
-  && mergePoint(b,x0,x1,x2,x3) != nil
-  && clobber(x0)
-  && clobber(x1)
+  && mergePoint(b,x01,x2,x3) != nil
+  && clobber(x01)
   && clobber(x2)
   && clobber(x3)
-  && clobber(s0)
   && clobber(s1)
   && clobber(s2)
   && clobber(o0)
   && clobber(o1)
-  -> @mergePoint(b,x0,x1,x2,x3) (BSWAPL <v.Type> (MOVLload [i-3] {s} p mem))
+  -> @mergePoint(b,x01,x2,x3) (BSWAPL <v.Type> (MOVLload [i1-2] {s} p mem))
 
-(ORL o1:(ORL o0:(ORL
-                       x0:(MOVBloadidx1 [i] {s} p idx mem)
-    s0:(SHLLconst [8]  x1:(MOVBloadidx1 [i-1] {s} p idx mem)))
-    s1:(SHLLconst [16] x2:(MOVBloadidx1 [i-2] {s} p idx mem)))
-    s2:(SHLLconst [24] x3:(MOVBloadidx1 [i-3] {s} p idx mem)))
-  && x0.Uses == 1
-  && x1.Uses == 1
+(ORL o1:(ORL o0:(ROLWconst [8] x01:(MOVWloadidx1 [i1] {s} p idx mem))
+    s1:(SHLLconst [16] x2:(MOVBloadidx1 [i1-1] {s} p idx mem)))
+    s2:(SHLLconst [24] x3:(MOVBloadidx1 [i1-2] {s} p idx mem)))
+  && x01.Uses == 1
   && x2.Uses == 1
   && x3.Uses == 1
-  && s0.Uses == 1
   && s1.Uses == 1
   && s2.Uses == 1
   && o0.Uses == 1
   && o1.Uses == 1
-  && mergePoint(b,x0,x1,x2,x3) != nil
-  && clobber(x0)
-  && clobber(x1)
+  && mergePoint(b,x01,x2,x3) != nil
+  && clobber(x01)
   && clobber(x2)
   && clobber(x3)
-  && clobber(s0)
   && clobber(s1)
   && clobber(s2)
   && clobber(o0)
   && clobber(o1)
-  -> @mergePoint(b,x0,x1,x2,x3) (BSWAPL <v.Type> (MOVLloadidx1 <v.Type> [i-3] {s} p idx mem))
+  -> @mergePoint(b,x01,x2,x3) (BSWAPL <v.Type> (MOVLloadidx1 <v.Type> [i1-2] {s} p idx mem))
 
 (ORQ o5:(ORQ o4:(ORQ o3:(ORQ o2:(ORQ o1:(ORQ o0:(ORQ
                        x0:(MOVBload [i] {s} p mem)
@@ -1719,6 +1733,19 @@
   && clobber(o5)
   -> @mergePoint(b,x0,x1,x2,x3,x4,x5,x6,x7) (BSWAPQ <v.Type> (MOVQloadidx1 <v.Type> [i-7] {s} p idx mem))
 
+// Combine 2 byte stores + shift into rolw 8 + word store
+(MOVBstore [i] {s} p w
+  x0:(MOVBstore [i-1] {s} p (SHRWconst [8] w) mem))
+  && x0.Uses == 1
+  && clobber(x0)
+  -> (MOVWstore [i-1] {s} p (ROLWconst <w.Type> [8] w) mem)
+
+(MOVBstoreidx1 [i] {s} p idx w
+  x0:(MOVBstoreidx1 [i-1] {s} p idx (SHRWconst [8] w) mem))
+  && x0.Uses == 1
+  && clobber(x0)
+  -> (MOVWstoreidx1 [i-1] {s} p idx (ROLWconst <w.Type> [8] w) mem)
+
 // Combine stores + shifts into bswap and larger (unaligned) stores
 (MOVBstore [i] {s} p w
   x2:(MOVBstore [i-1] {s} p (SHRLconst [8] w)

src/cmd/compile/internal/ssa/rewriteAMD64.go

@@ -4148,6 +4148,52 @@ func rewriteValueAMD64_OpAMD64MOVBstore(v *Value, config *Config) bool {
 		v.AddArg(mem)
 		return true
 	}
+	// match: (MOVBstore [i] {s} p w   x0:(MOVBstore [i-1] {s} p (SHRWconst [8] w) mem))
+	// cond: x0.Uses == 1   && clobber(x0)
+	// result: (MOVWstore [i-1] {s} p (ROLWconst <w.Type> [8] w) mem)
+	for {
+		i := v.AuxInt
+		s := v.Aux
+		p := v.Args[0]
+		w := v.Args[1]
+		x0 := v.Args[2]
+		if x0.Op != OpAMD64MOVBstore {
+			break
+		}
+		if x0.AuxInt != i-1 {
+			break
+		}
+		if x0.Aux != s {
+			break
+		}
+		if p != x0.Args[0] {
+			break
+		}
+		x0_1 := x0.Args[1]
+		if x0_1.Op != OpAMD64SHRWconst {
+			break
+		}
+		if x0_1.AuxInt != 8 {
+			break
+		}
+		if w != x0_1.Args[0] {
+			break
+		}
+		mem := x0.Args[2]
+		if !(x0.Uses == 1 && clobber(x0)) {
+			break
+		}
+		v.reset(OpAMD64MOVWstore)
+		v.AuxInt = i - 1
+		v.Aux = s
+		v.AddArg(p)
+		v0 := b.NewValue0(v.Pos, OpAMD64ROLWconst, w.Type)
+		v0.AuxInt = 8
+		v0.AddArg(w)
+		v.AddArg(v0)
+		v.AddArg(mem)
+		return true
+	}
 	// match: (MOVBstore [i] {s} p w   x2:(MOVBstore [i-1] {s} p (SHRLconst [8] w)   x1:(MOVBstore [i-2] {s} p (SHRLconst [16] w)   x0:(MOVBstore [i-3] {s} p (SHRLconst [24] w) mem))))
 	// cond: x0.Uses == 1   && x1.Uses == 1   && x2.Uses == 1   && clobber(x0)   && clobber(x1)   && clobber(x2)
 	// result: (MOVLstore [i-3] {s} p (BSWAPL <w.Type> w) mem)
@@ -4875,6 +4921,57 @@ func rewriteValueAMD64_OpAMD64MOVBstoreidx1(v *Value, config *Config) bool {
 		v.AddArg(mem)
 		return true
 	}
+	// match: (MOVBstoreidx1 [i] {s} p idx w   x0:(MOVBstoreidx1 [i-1] {s} p idx (SHRWconst [8] w) mem))
+	// cond: x0.Uses == 1   && clobber(x0)
+	// result: (MOVWstoreidx1 [i-1] {s} p idx (ROLWconst <w.Type> [8] w) mem)
+	for {
+		i := v.AuxInt
+		s := v.Aux
+		p := v.Args[0]
+		idx := v.Args[1]
+		w := v.Args[2]
+		x0 := v.Args[3]
+		if x0.Op != OpAMD64MOVBstoreidx1 {
+			break
+		}
+		if x0.AuxInt != i-1 {
+			break
+		}
+		if x0.Aux != s {
+			break
+		}
+		if p != x0.Args[0] {
+			break
+		}
+		if idx != x0.Args[1] {
+			break
+		}
+		x0_2 := x0.Args[2]
+		if x0_2.Op != OpAMD64SHRWconst {
+			break
+		}
+		if x0_2.AuxInt != 8 {
+			break
+		}
+		if w != x0_2.Args[0] {
+			break
+		}
+		mem := x0.Args[3]
+		if !(x0.Uses == 1 && clobber(x0)) {
+			break
+		}
+		v.reset(OpAMD64MOVWstoreidx1)
+		v.AuxInt = i - 1
+		v.Aux = s
+		v.AddArg(p)
+		v.AddArg(idx)
+		v0 := b.NewValue0(v.Pos, OpAMD64ROLWconst, w.Type)
+		v0.AuxInt = 8
+		v0.AddArg(w)
+		v.AddArg(v0)
+		v.AddArg(mem)
+		return true
+	}
 	// match: (MOVBstoreidx1 [i] {s} p idx w   x2:(MOVBstoreidx1 [i-1] {s} p idx (SHRLconst [8] w)   x1:(MOVBstoreidx1 [i-2] {s} p idx (SHRLconst [16] w)   x0:(MOVBstoreidx1 [i-3] {s} p idx (SHRLconst [24] w) mem))))
 	// cond: x0.Uses == 1   && x1.Uses == 1   && x2.Uses == 1   && clobber(x0)   && clobber(x1)   && clobber(x2)
 	// result: (MOVLstoreidx1 [i-3] {s} p idx (BSWAPL <w.Type> w) mem)
@@ -11866,19 +11963,11 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		v0.AddArg(mem)
 		return true
 	}
-	// match: (ORL o1:(ORL o0:(ORL                        x0:(MOVBload [i] {s} p mem)     s0:(SHLLconst [8]  x1:(MOVBload [i-1] {s} p mem)))     s1:(SHLLconst [16] x2:(MOVBload [i-2] {s} p mem)))     s2:(SHLLconst [24] x3:(MOVBload [i-3] {s} p mem)))
-	// cond: x0.Uses == 1   && x1.Uses == 1   && x2.Uses == 1   && x3.Uses == 1   && s0.Uses == 1   && s1.Uses == 1   && s2.Uses == 1   && o0.Uses == 1   && o1.Uses == 1   && mergePoint(b,x0,x1,x2,x3) != nil   && clobber(x0)   && clobber(x1)   && clobber(x2)   && clobber(x3)   && clobber(s0)   && clobber(s1)   && clobber(s2)   && clobber(o0)   && clobber(o1)
-	// result: @mergePoint(b,x0,x1,x2,x3) (BSWAPL <v.Type> (MOVLload [i-3] {s} p mem))
+	// match: (ORL                        x0:(MOVBload [i] {s} p mem)     s0:(SHLLconst [8]  x1:(MOVBload [i-1] {s} p mem)))
+	// cond: x0.Uses == 1   && x1.Uses == 1   && s0.Uses == 1   && mergePoint(b,x0,x1) != nil   && clobber(x0)   && clobber(x1)   && clobber(s0)
+	// result: @mergePoint(b,x0,x1) (ROLWconst <v.Type> [8] (MOVWload [i-1] {s} p mem))
 	for {
-		o1 := v.Args[0]
-		if o1.Op != OpAMD64ORL {
-			break
-		}
-		o0 := o1.Args[0]
-		if o0.Op != OpAMD64ORL {
-			break
-		}
-		x0 := o0.Args[0]
+		x0 := v.Args[0]
 		if x0.Op != OpAMD64MOVBload {
 			break
 		}
@@ -11886,7 +11975,7 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		s := x0.Aux
 		p := x0.Args[0]
 		mem := x0.Args[1]
-		s0 := o0.Args[1]
+		s0 := v.Args[1]
 		if s0.Op != OpAMD64SHLLconst {
 			break
 		}
@@ -11909,6 +11998,101 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if mem != x1.Args[1] {
 			break
 		}
+		if !(x0.Uses == 1 && x1.Uses == 1 && s0.Uses == 1 && mergePoint(b, x0, x1) != nil && clobber(x0) && clobber(x1) && clobber(s0)) {
+			break
+		}
+		b = mergePoint(b, x0, x1)
+		v0 := b.NewValue0(v.Pos, OpAMD64ROLWconst, v.Type)
+		v.reset(OpCopy)
+		v.AddArg(v0)
+		v0.AuxInt = 8
+		v1 := b.NewValue0(v.Pos, OpAMD64MOVWload, config.fe.TypeUInt16())
+		v1.AuxInt = i - 1
+		v1.Aux = s
+		v1.AddArg(p)
+		v1.AddArg(mem)
+		v0.AddArg(v1)
+		return true
+	}
+	// match: (ORL                        x0:(MOVBloadidx1 [i] {s} p idx mem)     s0:(SHLLconst [8]  x1:(MOVBloadidx1 [i-1] {s} p idx mem)))
+	// cond: x0.Uses == 1   && x1.Uses == 1   && s0.Uses == 1   && mergePoint(b,x0,x1) != nil   && clobber(x0)   && clobber(x1)   && clobber(s0)
+	// result: @mergePoint(b,x0,x1) (ROLWconst <v.Type> [8] (MOVWloadidx1 <v.Type> [i-1] {s} p idx mem))
+	for {
+		x0 := v.Args[0]
+		if x0.Op != OpAMD64MOVBloadidx1 {
+			break
+		}
+		i := x0.AuxInt
+		s := x0.Aux
+		p := x0.Args[0]
+		idx := x0.Args[1]
+		mem := x0.Args[2]
+		s0 := v.Args[1]
+		if s0.Op != OpAMD64SHLLconst {
+			break
+		}
+		if s0.AuxInt != 8 {
+			break
+		}
+		x1 := s0.Args[0]
+		if x1.Op != OpAMD64MOVBloadidx1 {
+			break
+		}
+		if x1.AuxInt != i-1 {
+			break
+		}
+		if x1.Aux != s {
+			break
+		}
+		if p != x1.Args[0] {
+			break
+		}
+		if idx != x1.Args[1] {
+			break
+		}
+		if mem != x1.Args[2] {
+			break
+		}
+		if !(x0.Uses == 1 && x1.Uses == 1 && s0.Uses == 1 && mergePoint(b, x0, x1) != nil && clobber(x0) && clobber(x1) && clobber(s0)) {
+			break
+		}
+		b = mergePoint(b, x0, x1)
+		v0 := b.NewValue0(v.Pos, OpAMD64ROLWconst, v.Type)
+		v.reset(OpCopy)
+		v.AddArg(v0)
+		v0.AuxInt = 8
+		v1 := b.NewValue0(v.Pos, OpAMD64MOVWloadidx1, v.Type)
+		v1.AuxInt = i - 1
+		v1.Aux = s
+		v1.AddArg(p)
+		v1.AddArg(idx)
+		v1.AddArg(mem)
+		v0.AddArg(v1)
+		return true
+	}
+	// match: (ORL o1:(ORL o0:(ROLWconst [8] x01:(MOVWload [i1] {s} p mem))     s1:(SHLLconst [16] x2:(MOVBload [i1-1] {s} p mem)))     s2:(SHLLconst [24] x3:(MOVBload [i1-2] {s} p mem)))
+	// cond: x01.Uses == 1   && x2.Uses == 1   && x3.Uses == 1   && s1.Uses == 1   && s2.Uses == 1   && o0.Uses == 1   && o1.Uses == 1   && mergePoint(b,x01,x2,x3) != nil   && clobber(x01)   && clobber(x2)   && clobber(x3)   && clobber(s1)   && clobber(s2)   && clobber(o0)   && clobber(o1)
+	// result: @mergePoint(b,x01,x2,x3) (BSWAPL <v.Type> (MOVLload [i1-2] {s} p mem))
+	for {
+		o1 := v.Args[0]
+		if o1.Op != OpAMD64ORL {
+			break
+		}
+		o0 := o1.Args[0]
+		if o0.Op != OpAMD64ROLWconst {
+			break
+		}
+		if o0.AuxInt != 8 {
+			break
+		}
+		x01 := o0.Args[0]
+		if x01.Op != OpAMD64MOVWload {
+			break
+		}
+		i1 := x01.AuxInt
+		s := x01.Aux
+		p := x01.Args[0]
+		mem := x01.Args[1]
 		s1 := o1.Args[1]
 		if s1.Op != OpAMD64SHLLconst {
 			break
@@ -11920,7 +12104,7 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if x2.Op != OpAMD64MOVBload {
 			break
 		}
-		if x2.AuxInt != i-2 {
+		if x2.AuxInt != i1-1 {
 			break
 		}
 		if x2.Aux != s {
@@ -11943,7 +12127,7 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if x3.Op != OpAMD64MOVBload {
 			break
 		}
-		if x3.AuxInt != i-3 {
+		if x3.AuxInt != i1-2 {
 			break
 		}
 		if x3.Aux != s {
@@ -11955,68 +12139,45 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if mem != x3.Args[1] {
 			break
 		}
-		if !(x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1 && x3.Uses == 1 && s0.Uses == 1 && s1.Uses == 1 && s2.Uses == 1 && o0.Uses == 1 && o1.Uses == 1 && mergePoint(b, x0, x1, x2, x3) != nil && clobber(x0) && clobber(x1) && clobber(x2) && clobber(x3) && clobber(s0) && clobber(s1) && clobber(s2) && clobber(o0) && clobber(o1)) {
+		if !(x01.Uses == 1 && x2.Uses == 1 && x3.Uses == 1 && s1.Uses == 1 && s2.Uses == 1 && o0.Uses == 1 && o1.Uses == 1 && mergePoint(b, x01, x2, x3) != nil && clobber(x01) && clobber(x2) && clobber(x3) && clobber(s1) && clobber(s2) && clobber(o0) && clobber(o1)) {
 			break
 		}
-		b = mergePoint(b, x0, x1, x2, x3)
+		b = mergePoint(b, x01, x2, x3)
 		v0 := b.NewValue0(v.Pos, OpAMD64BSWAPL, v.Type)
 		v.reset(OpCopy)
 		v.AddArg(v0)
 		v1 := b.NewValue0(v.Pos, OpAMD64MOVLload, config.fe.TypeUInt32())
-		v1.AuxInt = i - 3
+		v1.AuxInt = i1 - 2
 		v1.Aux = s
 		v1.AddArg(p)
 		v1.AddArg(mem)
 		v0.AddArg(v1)
 		return true
 	}
-	// match: (ORL o1:(ORL o0:(ORL                        x0:(MOVBloadidx1 [i] {s} p idx mem)     s0:(SHLLconst [8]  x1:(MOVBloadidx1 [i-1] {s} p idx mem)))     s1:(SHLLconst [16] x2:(MOVBloadidx1 [i-2] {s} p idx mem)))     s2:(SHLLconst [24] x3:(MOVBloadidx1 [i-3] {s} p idx mem)))
-	// cond: x0.Uses == 1   && x1.Uses == 1   && x2.Uses == 1   && x3.Uses == 1   && s0.Uses == 1   && s1.Uses == 1   && s2.Uses == 1   && o0.Uses == 1   && o1.Uses == 1   && mergePoint(b,x0,x1,x2,x3) != nil   && clobber(x0)   && clobber(x1)   && clobber(x2)   && clobber(x3)   && clobber(s0)   && clobber(s1)   && clobber(s2)   && clobber(o0)   && clobber(o1)
-	// result: @mergePoint(b,x0,x1,x2,x3) (BSWAPL <v.Type> (MOVLloadidx1 <v.Type> [i-3] {s} p idx mem))
+	// match: (ORL o1:(ORL o0:(ROLWconst [8] x01:(MOVWloadidx1 [i1] {s} p idx mem))     s1:(SHLLconst [16] x2:(MOVBloadidx1 [i1-1] {s} p idx mem)))     s2:(SHLLconst [24] x3:(MOVBloadidx1 [i1-2] {s} p idx mem)))
+	// cond: x01.Uses == 1   && x2.Uses == 1   && x3.Uses == 1   && s1.Uses == 1   && s2.Uses == 1   && o0.Uses == 1   && o1.Uses == 1   && mergePoint(b,x01,x2,x3) != nil   && clobber(x01)   && clobber(x2)   && clobber(x3)   && clobber(s1)   && clobber(s2)   && clobber(o0)   && clobber(o1)
+	// result: @mergePoint(b,x01,x2,x3) (BSWAPL <v.Type> (MOVLloadidx1 <v.Type> [i1-2] {s} p idx mem))
 	for {
 		o1 := v.Args[0]
 		if o1.Op != OpAMD64ORL {
 			break
 		}
 		o0 := o1.Args[0]
-		if o0.Op != OpAMD64ORL {
+		if o0.Op != OpAMD64ROLWconst {
 			break
 		}
-		x0 := o0.Args[0]
-		if x0.Op != OpAMD64MOVBloadidx1 {
+		if o0.AuxInt != 8 {
 			break
 		}
-		i := x0.AuxInt
-		s := x0.Aux
-		p := x0.Args[0]
-		idx := x0.Args[1]
-		mem := x0.Args[2]
-		s0 := o0.Args[1]
-		if s0.Op != OpAMD64SHLLconst {
-			break
-		}
-		if s0.AuxInt != 8 {
-			break
-		}
-		x1 := s0.Args[0]
-		if x1.Op != OpAMD64MOVBloadidx1 {
-			break
-		}
-		if x1.AuxInt != i-1 {
-			break
-		}
-		if x1.Aux != s {
-			break
-		}
-		if p != x1.Args[0] {
-			break
-		}
-		if idx != x1.Args[1] {
-			break
-		}
-		if mem != x1.Args[2] {
+		x01 := o0.Args[0]
+		if x01.Op != OpAMD64MOVWloadidx1 {
 			break
 		}
+		i1 := x01.AuxInt
+		s := x01.Aux
+		p := x01.Args[0]
+		idx := x01.Args[1]
+		mem := x01.Args[2]
 		s1 := o1.Args[1]
 		if s1.Op != OpAMD64SHLLconst {
 			break
@@ -12028,7 +12189,7 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if x2.Op != OpAMD64MOVBloadidx1 {
 			break
 		}
-		if x2.AuxInt != i-2 {
+		if x2.AuxInt != i1-1 {
 			break
 		}
 		if x2.Aux != s {
@@ -12054,7 +12215,7 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if x3.Op != OpAMD64MOVBloadidx1 {
 			break
 		}
-		if x3.AuxInt != i-3 {
+		if x3.AuxInt != i1-2 {
 			break
 		}
 		if x3.Aux != s {
@@ -12069,15 +12230,15 @@ func rewriteValueAMD64_OpAMD64ORL(v *Value, config *Config) bool {
 		if mem != x3.Args[2] {
 			break
 		}
-		if !(x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1 && x3.Uses == 1 && s0.Uses == 1 && s1.Uses == 1 && s2.Uses == 1 && o0.Uses == 1 && o1.Uses == 1 && mergePoint(b, x0, x1, x2, x3) != nil && clobber(x0) && clobber(x1) && clobber(x2) && clobber(x3) && clobber(s0) && clobber(s1) && clobber(s2) && clobber(o0) && clobber(o1)) {
+		if !(x01.Uses == 1 && x2.Uses == 1 && x3.Uses == 1 && s1.Uses == 1 && s2.Uses == 1 && o0.Uses == 1 && o1.Uses == 1 && mergePoint(b, x01, x2, x3) != nil && clobber(x01) && clobber(x2) && clobber(x3) && clobber(s1) && clobber(s2) && clobber(o0) && clobber(o1)) {
 			break
 		}
-		b = mergePoint(b, x0, x1, x2, x3)
+		b = mergePoint(b, x01, x2, x3)
 		v0 := b.NewValue0(v.Pos, OpAMD64BSWAPL, v.Type)
 		v.reset(OpCopy)
 		v.AddArg(v0)
 		v1 := b.NewValue0(v.Pos, OpAMD64MOVLloadidx1, v.Type)
-		v1.AuxInt = i - 3
+		v1.AuxInt = i1 - 2
 		v1.Aux = s
 		v1.AddArg(p)
 		v1.AddArg(idx)

src/encoding/binary/binary_test.go

@@ -500,3 +500,27 @@ func BenchmarkWriteSlice1000Int32s(b *testing.B) {
 	}
 	b.StopTimer()
 }
+
+func BenchmarkPutUint16(b *testing.B) {
+	buf := [2]byte{}
+	b.SetBytes(2)
+	for i := 0; i < b.N; i++ {
+		BigEndian.PutUint16(buf[:], uint16(i))
+	}
+}
+
+func BenchmarkPutUint32(b *testing.B) {
+	buf := [4]byte{}
+	b.SetBytes(4)
+	for i := 0; i < b.N; i++ {
+		BigEndian.PutUint32(buf[:], uint32(i))
+	}
+}
+
+func BenchmarkPutUint64(b *testing.B) {
+	buf := [8]byte{}
+	b.SetBytes(8)
+	for i := 0; i < b.N; i++ {
+		BigEndian.PutUint64(buf[:], uint64(i))
+	}
+}