cmd/compile: add MOVBEWstore support for GOAMD64>=3

This CL add MOVBE support for 16-bit version, but MOVBEWload is excluded because it does not satisfy zero extented. For golang#51724 Change-Id: I3fadf20bcbb9b423f6355e6a1e340107e8e621ac Reviewed-on: https://go-review.googlesource.com/c/go/+/396617 Reviewed-by: Keith Randall <[email protected]> Run-TryBot: Keith Randall <[email protected]> TryBot-Result: Gopher Robot <[email protected]> Trust: Emmanuel Odeke <[email protected]>
robertodauria · Apr 3, 2022 · ba6df85 · ba6df85
1 parent 3451844
commit ba6df85
Show file tree

Hide file tree

Showing 6 changed files with 82 additions and 8 deletions.
diff --git a/src/cmd/compile/internal/amd64/ssa.go b/src/cmd/compile/internal/amd64/ssa.go
@@ -778,7 +778,7 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
 	case ssa.OpAMD64MOVQstore, ssa.OpAMD64MOVSSstore, ssa.OpAMD64MOVSDstore, ssa.OpAMD64MOVLstore, ssa.OpAMD64MOVWstore, ssa.OpAMD64MOVBstore, ssa.OpAMD64MOVOstore,
 		ssa.OpAMD64ADDQmodify, ssa.OpAMD64SUBQmodify, ssa.OpAMD64ANDQmodify, ssa.OpAMD64ORQmodify, ssa.OpAMD64XORQmodify,
 		ssa.OpAMD64ADDLmodify, ssa.OpAMD64SUBLmodify, ssa.OpAMD64ANDLmodify, ssa.OpAMD64ORLmodify, ssa.OpAMD64XORLmodify,
-		ssa.OpAMD64MOVBEQstore, ssa.OpAMD64MOVBELstore:
+		ssa.OpAMD64MOVBEQstore, ssa.OpAMD64MOVBELstore, ssa.OpAMD64MOVBEWstore:
 		p := s.Prog(v.Op.Asm())
 		p.From.Type = obj.TYPE_REG
 		p.From.Reg = v.Args[1].Reg()

diff --git a/src/cmd/compile/internal/ssa/gen/AMD64.rules b/src/cmd/compile/internal/ssa/gen/AMD64.rules
@@ -2225,6 +2225,8 @@
 (BSWAP(Q|L) x:(MOV(Q|L)load [i] {s} p mem))    && x.Uses == 1 && buildcfg.GOAMD64 >= 3 => (MOVBE(Q|L)load [i] {s} p mem)
 (BSWAP(Q|L) (MOVBE(Q|L)load [i] {s} p m))    => (MOV(Q|L)load [i] {s} p m)
 (MOVBE(Q|L)store [i] {s} p (BSWAP(Q|L) x) m) => (MOV(Q|L)store [i] {s} p x m)
+(MOVWstore [i] {s} p x:(ROLWconst [8] w) mem)   && x.Uses == 1 && buildcfg.GOAMD64 >= 3 => (MOVBEWstore [i] {s} p w mem)
+(MOVBEWstore [i] {s} p x:(ROLWconst [8] w) mem) && x.Uses == 1 => (MOVWstore [i] {s} p w mem)
 
 (ORQ                   x0:(MOVBELload [i0] {s} p mem)
     sh:(SHLQconst [32] x1:(MOVBELload [i1] {s} p mem)))

diff --git a/src/cmd/compile/internal/ssa/gen/AMD64Ops.go b/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
@@ -924,6 +924,8 @@ func init() {
 		{name: "TZCNTL", argLength: 1, reg: gp11, asm: "TZCNTL", clobberFlags: true},
 
 		// CPUID feature: MOVBE
+		// MOVBEWload does not satisfy zero extended, so only use MOVBEWstore
+		{name: "MOVBEWstore", argLength: 3, reg: gpstore, asm: "MOVBEW", aux: "SymOff", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // swap and store 2 bytes in arg1 to arg0+auxint+aux. arg2=mem
 		{name: "MOVBELload", argLength: 2, reg: gpload, asm: "MOVBEL", aux: "SymOff", typ: "UInt32", faultOnNilArg0: true, symEffect: "Read"}, // load and swap 4 bytes from arg0+auxint+aux. arg1=mem.  Zero extend.
 		{name: "MOVBELstore", argLength: 3, reg: gpstore, asm: "MOVBEL", aux: "SymOff", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // swap and store 4 bytes in arg1 to arg0+auxint+aux. arg2=mem
 		{name: "MOVBEQload", argLength: 2, reg: gpload, asm: "MOVBEQ", aux: "SymOff", typ: "UInt64", faultOnNilArg0: true, symEffect: "Read"}, // load and swap 8 bytes from arg0+auxint+aux. arg1=mem

diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
diff --git a/test/codegen/memcombine.go b/test/codegen/memcombine.go
@@ -105,20 +105,20 @@ func load_be32_idx(b []byte, idx int) {
 	sink32 = binary.BigEndian.Uint32(b[idx:])
 }
 
-func load_be16(b []byte) {
+func load_be16(b []byte) uint16 {
 	// amd64:`ROLW\s\$8`,-`MOVB`,-`OR`
 	// arm64:`REV16W`,`MOVHU\s\(R[0-9]+\),`,-`MOVB`
 	// ppc64le:`MOVHBR`
 	// s390x:`MOVHZ\s\(.*\),`,-`OR`,-`ORW`,-`SLD`,-`SLW`
-	sink16 = binary.BigEndian.Uint16(b)
+	return binary.BigEndian.Uint16(b)
 }
 
-func load_be16_idx(b []byte, idx int) {
+func load_be16_idx(b []byte, idx int) uint16 {
 	// amd64:`ROLW\s\$8`,-`MOVB`,-`OR`
 	// arm64:`REV16W`,`MOVHU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOVB`
 	// ppc64le:`MOVHBR`
 	// s390x:`MOVHZ\s\(.*\)\(.*\*1\),`,-`OR`,-`ORW`,-`SLD`,-`SLW`
-	sink16 = binary.BigEndian.Uint16(b[idx:])
+	return binary.BigEndian.Uint16(b[idx:])
 }
 
 func load_le_byte2_uint16(s []byte) uint16 {
@@ -463,15 +463,17 @@ func store_be32_idx(b []byte, idx int) {
 }
 
 func store_be16(b []byte) {
-	// amd64:`ROLW\s\$8`,-`SHR.`
+	// amd64/v1,amd64/v2:`ROLW\s\$8`,-`SHR.`
+	// amd64/v3:`MOVBEW`,-`ROLW`
 	// arm64:`MOVH`,`REV16W`,-`MOVB`
 	// ppc64le:`MOVHBR`
 	// s390x:`MOVH\s.*\(.*\)$`,-`SRW\s`,-`SRD\s`
 	binary.BigEndian.PutUint16(b, sink16)
 }
 
 func store_be16_idx(b []byte, idx int) {
-	// amd64:`ROLW\s\$8`,-`SHR.`
+	// amd64/v1,amd64/v2:`ROLW\s\$8`,-`SHR.`
+	// amd64/v3: `MOVBEW`
 	// arm64:`MOVH\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,`REV16W`,-`MOVB`
 	// ppc64le:`MOVHBR`
 	// s390x:`MOVH\s.*\(.*\)\(.*\*1\)$`,-`SRW\s`,-`SRD\s`
@@ -511,7 +513,8 @@ func store_le_byte_8(b []byte, val uint64) {
 func store_be_byte_2(b []byte, val uint16) {
 	_ = b[2]
 	// arm64:`REV16W`,`MOVH\sR[0-9]+,\s1\(R[0-9]+\)`,-`MOVB`
-	// amd64:`MOVW\s[A-Z]+,\s1\([A-Z]+\)`,-`MOVB`
+	// amd64/v1,amd64/v2:`MOVW\s[A-Z]+,\s1\([A-Z]+\)`,-`MOVB`
+	// amd64/v3: `MOVBEW`
 	b[1], b[2] = byte(val>>8), byte(val)
 }