cmd/compile: allow non-pointer writes in the middle of a write barrier

This lets us combine more write barriers, getting rid of some of the
test+branch and gcWriteBarrier* calls.
With the new write barriers, it's easy to add a few non-pointer writes
to the set of values written.

We allow up to 2 non-pointer writes between pointer writes. This is enough
for, for example, adjacent slice fields.

Fixes #62126

Change-Id: I872d0fa9cc4eb855e270ffc0223b39fde1723c4b
Reviewed-on: https://go-review.googlesource.com/c/go/+/521498
TryBot-Result: Gopher Robot <[email protected]>
Run-TryBot: Keith Randall <[email protected]>
Reviewed-by: Cherry Mui <[email protected]>
Reviewed-by: Keith Randall <[email protected]>

Author: randall77

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

@@ -250,6 +250,7 @@ func writebarrier(f *Func) {
 		// to a new block.
 		var last *Value
 		var start, end int
+		var nonPtrStores int
 		values := b.Values
 	FindSeq:
 		for i := len(values) - 1; i >= 0; i-- {
@@ -261,8 +262,17 @@ func writebarrier(f *Func) {
 					last = w
 					end = i + 1
 				}
+				nonPtrStores = 0
 			case OpVarDef, OpVarLive:
 				continue
+			case OpStore:
+				if last == nil {
+					continue
+				}
+				nonPtrStores++
+				if nonPtrStores > 2 {
+					break FindSeq
+				}
 			default:
 				if last == nil {
 					continue
@@ -484,6 +494,10 @@ func writebarrier(f *Func) {
 				mem.Aux = w.Aux
 			case OpVarDef, OpVarLive:
 				mem = bEnd.NewValue1A(pos, w.Op, types.TypeMem, w.Aux, mem)
+			case OpStore:
+				ptr := w.Args[0]
+				val := w.Args[1]
+				mem = bEnd.NewValue3A(pos, OpStore, types.TypeMem, w.Aux, ptr, val, mem)
 			}
 		}
 

test/codegen/writebarrier.go

@@ -0,0 +1,55 @@
+// asmcheck
+
+// Copyright 2023 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 codegen
+
+func combine2string(p *[2]string, a, b string) {
+	// amd64:`.*runtime[.]gcWriteBarrier4\(SB\)`
+	// arm64:`.*runtime[.]gcWriteBarrier4\(SB\)`
+	p[0] = a
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[1] = b
+}
+
+func combine4string(p *[4]string, a, b, c, d string) {
+	// amd64:`.*runtime[.]gcWriteBarrier8\(SB\)`
+	// arm64:`.*runtime[.]gcWriteBarrier8\(SB\)`
+	p[0] = a
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[1] = b
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[2] = c
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[3] = d
+}
+
+func combine2slice(p *[2][]byte, a, b []byte) {
+	// amd64:`.*runtime[.]gcWriteBarrier4\(SB\)`
+	// arm64:`.*runtime[.]gcWriteBarrier4\(SB\)`
+	p[0] = a
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[1] = b
+}
+
+func combine4slice(p *[4][]byte, a, b, c, d []byte) {
+	// amd64:`.*runtime[.]gcWriteBarrier8\(SB\)`
+	// arm64:`.*runtime[.]gcWriteBarrier8\(SB\)`
+	p[0] = a
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[1] = b
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[2] = c
+	// amd64:-`.*runtime[.]gcWriteBarrier`
+	// arm64:-`.*runtime[.]gcWriteBarrier`
+	p[3] = d
+}