cmd/compile: don't short-circuit copies whose source is volatile

Current optimization: When we copy a->b and then b->c, we might as well
copy a->c instead of b->c (then b might be dead and go away).

*Except* if a is a volatile location (might be clobbered by a call).
In that case, we really do want to copy a immediately, because there
might be a call before we can do the a->c copy.

User calls can't happen in between, because the rule matches up the
memory states. But calls inserted for memory barriers, particularly
runtime.typedmemmove, can.

(I guess we could introduce a register-calling-convention version
of runtime.typedmemmove, but that seems a bigger change than this one.)

Fixes #43570

Change-Id: Ifa518bb1a6f3a8dd46c352d4fd54ea9713b3eb1a
Reviewed-on: https://go-review.googlesource.com/c/go/+/282492
Trust: Keith Randall <khr@golang.org>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

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

@@ -2512,7 +2512,7 @@
 (Move {t1} [s] dst tmp1 midmem:(Move {t2} [s] tmp2 src _))
 	&& t1.Compare(t2) == types.CMPeq
 	&& isSamePtr(tmp1, tmp2)
-	&& isStackPtr(src)
+	&& isStackPtr(src) && !isVolatile(src)
 	&& disjoint(src, s, tmp2, s)
 	&& (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
 	=> (Move {t1} [s] dst src midmem)
@@ -2521,7 +2521,7 @@
 (Move {t1} [s] dst tmp1 midmem:(VarDef (Move {t2} [s] tmp2 src _)))
 	&& t1.Compare(t2) == types.CMPeq
 	&& isSamePtr(tmp1, tmp2)
-	&& isStackPtr(src)
+	&& isStackPtr(src) && !isVolatile(src)
 	&& disjoint(src, s, tmp2, s)
 	&& (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
 	=> (Move {t1} [s] dst src midmem)

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

@@ -13637,7 +13637,7 @@ func rewriteValuegeneric_OpMove(v *Value) bool {
 		return true
 	}
 	// match: (Move {t1} [s] dst tmp1 midmem:(Move {t2} [s] tmp2 src _))
-	// cond: t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
+	// cond: t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && !isVolatile(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
 	// result: (Move {t1} [s] dst src midmem)
 	for {
 		s := auxIntToInt64(v.AuxInt)
@@ -13651,7 +13651,7 @@ func rewriteValuegeneric_OpMove(v *Value) bool {
 		t2 := auxToType(midmem.Aux)
 		src := midmem.Args[1]
 		tmp2 := midmem.Args[0]
-		if !(t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))) {
+		if !(t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && !isVolatile(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))) {
 			break
 		}
 		v.reset(OpMove)
@@ -13661,7 +13661,7 @@ func rewriteValuegeneric_OpMove(v *Value) bool {
 		return true
 	}
 	// match: (Move {t1} [s] dst tmp1 midmem:(VarDef (Move {t2} [s] tmp2 src _)))
-	// cond: t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
+	// cond: t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && !isVolatile(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))
 	// result: (Move {t1} [s] dst src midmem)
 	for {
 		s := auxIntToInt64(v.AuxInt)
@@ -13679,7 +13679,7 @@ func rewriteValuegeneric_OpMove(v *Value) bool {
 		t2 := auxToType(midmem_0.Aux)
 		src := midmem_0.Args[1]
 		tmp2 := midmem_0.Args[0]
-		if !(t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))) {
+		if !(t1.Compare(t2) == types.CMPeq && isSamePtr(tmp1, tmp2) && isStackPtr(src) && !isVolatile(src) && disjoint(src, s, tmp2, s) && (disjoint(src, s, dst, s) || isInlinableMemmove(dst, src, s, config))) {
 			break
 		}
 		v.reset(OpMove)

test/fixedbugs/issue43570.go

@@ -0,0 +1,40 @@
+// run
+
+// Copyright 2021 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package main
+
+import "fmt"
+
+type T [8]*int
+
+//go:noinline
+func f(x int) T {
+	return T{}
+}
+
+//go:noinline
+func g(x int, t T) {
+	if t != (T{}) {
+		panic(fmt.Sprintf("bad: %v", t))
+	}
+}
+
+func main() {
+	const N = 10000
+	var q T
+	func() {
+		for i := 0; i < N; i++ {
+			q = f(0)
+			g(0, q)
+			sink = make([]byte, 1024)
+		}
+	}()
+	// Note that the closure is a trick to get the write to q to be a
+	// write to a pointer that is known to be non-nil and requires
+	// a write barrier.
+}
+
+var sink []byte