cmd/compile: make []byte("...") more efficient

Do []byte(string) conversions more efficiently when the string
is a constant. Instead of calling stringtobyteslice, allocate
just the space we need and encode the initialization directly.

[]byte("foo") rewrites to the following pseudocode:

var s [3]byte // on heap or stack, depending on whether b escapes
s = *(*[3]byte)(&"foo"[0]) // initialize s from the string
b = s[:]

which generates this assembly:

	0x001d 00029 (tmp1.go:9)	LEAQ	type.[3]uint8(SB), AX
	0x0024 00036 (tmp1.go:9)	MOVQ	AX, (SP)
	0x0028 00040 (tmp1.go:9)	CALL	runtime.newobject(SB)
	0x002d 00045 (tmp1.go:9)	MOVQ	8(SP), AX
	0x0032 00050 (tmp1.go:9)	MOVBLZX	go.string."foo"+2(SB), CX
	0x0039 00057 (tmp1.go:9)	MOVWLZX	go.string."foo"(SB), DX
	0x0040 00064 (tmp1.go:9)	MOVW	DX, (AX)
	0x0043 00067 (tmp1.go:9)	MOVB	CL, 2(AX)
// Then the slice is b = {AX, 3, 3}

The generated code is still not optimal, as it still does load/store
from read-only memory instead of constant stores.  Next CL...

Update #26498
Fixes #10170

Change-Id: I4b990b19f9a308f60c8f4f148934acffefe0a5bd
Reviewed-on: https://go-review.googlesource.com/c/140698
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

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

@@ -798,9 +798,8 @@ func (e *EscState) esc(n *Node, parent *Node) {
 	// gathered here.
 	if n.Esc != EscHeap && n.Type != nil &&
 		(n.Type.Width > maxStackVarSize ||
-			(n.Op == ONEW || n.Op == OPTRLIT) && n.Type.Elem().Width >= 1<<16 ||
+			(n.Op == ONEW || n.Op == OPTRLIT) && n.Type.Elem().Width >= maxImplicitStackVarSize ||
 			n.Op == OMAKESLICE && !isSmallMakeSlice(n)) {
-
 		// isSmallMakeSlice returns false for non-constant len/cap.
 		// If that's the case, print a more accurate escape reason.
 		var msgVerb, escapeMsg string

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

@@ -13,8 +13,18 @@ import (
 )
 
 const (
-	BADWIDTH        = types.BADWIDTH
+	BADWIDTH = types.BADWIDTH
+
+	// maximum size variable which we will allocate on the stack.
+	// This limit is for explicit variable declarations like "var x T" or "x := ...".
 	maxStackVarSize = 10 * 1024 * 1024
+
+	// maximum size of implicit variables that we will allocate on the stack.
+	//   p := new(T)          allocating T on the stack
+	//   p := &T{}            allocating T on the stack
+	//   s := make([]T, n)    allocating [n]T on the stack
+	//   s := []byte("...")   allocating [n]byte on the stack
+	maxImplicitStackVarSize = 64 * 1024
 )
 
 // isRuntimePkg reports whether p is package runtime.

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

@@ -367,7 +367,7 @@ func isSmallMakeSlice(n *Node) bool {
 	}
 	t := n.Type
 
-	return smallintconst(l) && smallintconst(r) && (t.Elem().Width == 0 || r.Int64() < (1<<16)/t.Elem().Width)
+	return smallintconst(l) && smallintconst(r) && (t.Elem().Width == 0 || r.Int64() < maxImplicitStackVarSize/t.Elem().Width)
 }
 
 // walk the whole tree of the body of an
@@ -1204,7 +1204,7 @@ opswitch:
 
 	case ONEW:
 		if n.Esc == EscNone {
-			if n.Type.Elem().Width >= 1<<16 {
+			if n.Type.Elem().Width >= maxImplicitStackVarSize {
 				Fatalf("large ONEW with EscNone: %v", n)
 			}
 			r := temp(n.Type.Elem())
@@ -1593,8 +1593,36 @@ opswitch:
 
 		n = mkcall("slicerunetostring", n.Type, init, a, n.Left)
 
-		// stringtoslicebyte(*32[byte], string) []byte;
 	case OSTRARRAYBYTE:
+		s := n.Left
+		if Isconst(s, CTSTR) {
+			sc := s.Val().U.(string)
+
+			// Allocate a [n]byte of the right size.
+			t := types.NewArray(types.Types[TUINT8], int64(len(sc)))
+			var a *Node
+			if n.Esc == EscNone && len(sc) <= maxImplicitStackVarSize {
+				a = nod(OADDR, temp(t), nil)
+			} else {
+				a = callnew(t)
+			}
+			p := temp(t.PtrTo()) // *[n]byte
+			init.Append(typecheck(nod(OAS, p, a), Etop))
+
+			// Copy from the static string data to the [n]byte.
+			if len(sc) > 0 {
+				as := nod(OAS,
+					nod(OIND, p, nil),
+					nod(OIND, convnop(nod(OSPTR, s, nil), t.PtrTo()), nil))
+				init.Append(typecheck(as, Etop))
+			}
+
+			// Slice the [n]byte to a []byte.
+			n.Op = OSLICEARR
+			n.Left = p
+			n = walkexpr(n, init)
+			break
+		}
 		a := nodnil()
 
 		if n.Esc == EscNone {
@@ -1604,7 +1632,8 @@ opswitch:
 			a = nod(OADDR, temp(t), nil)
 		}
 
-		n = mkcall("stringtoslicebyte", n.Type, init, a, conv(n.Left, types.Types[TSTRING]))
+		// stringtoslicebyte(*32[byte], string) []byte;
+		n = mkcall("stringtoslicebyte", n.Type, init, a, conv(s, types.Types[TSTRING]))
 
 	case OSTRARRAYBYTETMP:
 		// []byte(string) conversion that creates a slice

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

@@ -816,7 +816,7 @@
 // Decomposing StringMake and lowering of StringPtr and StringLen
 // happens in a later pass, dec, so that these operations are available
 // to other passes for optimizations.
-(StringPtr (StringMake (Const64 <t> [c]) _)) -> (Const64 <t> [c])
+(StringPtr (StringMake (Addr <t> {s} base) _)) -> (Addr <t> {s} base)
 (StringLen (StringMake _ (Const64 <t> [c]))) -> (Const64 <t> [c])
 (ConstString {s}) && config.PtrSize == 4 && s.(string) == "" ->
   (StringMake (ConstNil) (Const32 <typ.Int> [0]))

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

@@ -28748,9 +28748,9 @@ func rewriteValuegeneric_OpStringLen_0(v *Value) bool {
 	return false
 }
 func rewriteValuegeneric_OpStringPtr_0(v *Value) bool {
-	// match: (StringPtr (StringMake (Const64 <t> [c]) _))
+	// match: (StringPtr (StringMake (Addr <t> {s} base) _))
 	// cond:
-	// result: (Const64 <t> [c])
+	// result: (Addr <t> {s} base)
 	for {
 		v_0 := v.Args[0]
 		if v_0.Op != OpStringMake {
@@ -28758,14 +28758,16 @@ func rewriteValuegeneric_OpStringPtr_0(v *Value) bool {
 		}
 		_ = v_0.Args[1]
 		v_0_0 := v_0.Args[0]
-		if v_0_0.Op != OpConst64 {
+		if v_0_0.Op != OpAddr {
 			break
 		}
 		t := v_0_0.Type
-		c := v_0_0.AuxInt
-		v.reset(OpConst64)
+		s := v_0_0.Aux
+		base := v_0_0.Args[0]
+		v.reset(OpAddr)
 		v.Type = t
-		v.AuxInt = c
+		v.Aux = s
+		v.AddArg(base)
 		return true
 	}
 	return false

test/codegen/strings.go

@@ -13,3 +13,10 @@ func CountRunes(s string) int { // Issue #24923
 	// amd64:`.*countrunes`
 	return len([]rune(s))
 }
+
+func ToByteSlice() []byte { // Issue #24698
+	// amd64:`LEAQ\ttype\.\[3\]uint8`
+	// amd64:`CALL\truntime\.newobject`
+	// amd64:-`.*runtime.stringtoslicebyte`
+	return []byte("foo")
+}