cmd/compile: optimize ARM64 with shifted register indexed load/store

ARM64 supports efficient instructions which combine shift, addition, load/store together. Such as "MOVD (R0)(R1<<3), R2" and "MOVWU R6, (R4)(R1<<2)". This CL optimizes the compiler to emit such efficient instuctions. And below is some test data. 1. binary size before/after binary size change pkg/linux_arm64 +80.1KB pkg/tool/linux_arm64 +121.9KB go -4.3KB gofmt -64KB 2. go1 benchmark There is big improvement for the test case Fannkuch11, and slight improvement for sme others, excluding noise. name old time/op new time/op delta BinaryTree17-4 43.9s ± 2% 44.0s ± 2% ~ (p=0.820 n=30+30) Fannkuch11-4 30.6s ± 2% 24.5s ± 3% -19.93% (p=0.000 n=25+30) FmtFprintfEmpty-4 500ns ± 0% 499ns ± 0% -0.11% (p=0.000 n=23+25) FmtFprintfString-4 1.03µs ± 0% 1.04µs ± 3% ~ (p=0.065 n=29+30) FmtFprintfInt-4 1.15µs ± 3% 1.15µs ± 4% -0.56% (p=0.000 n=30+30) FmtFprintfIntInt-4 1.80µs ± 5% 1.82µs ± 0% ~ (p=0.094 n=30+24) FmtFprintfPrefixedInt-4 2.17µs ± 5% 2.20µs ± 0% ~ (p=0.100 n=30+23) FmtFprintfFloat-4 3.08µs ± 3% 3.09µs ± 4% ~ (p=0.123 n=30+30) FmtManyArgs-4 7.41µs ± 4% 7.17µs ± 1% -3.26% (p=0.000 n=30+23) GobDecode-4 93.7ms ± 0% 94.7ms ± 4% ~ (p=0.685 n=24+30) GobEncode-4 78.7ms ± 7% 77.1ms ± 0% ~ (p=0.729 n=30+23) Gzip-4 4.01s ± 0% 3.97s ± 5% -1.11% (p=0.037 n=24+30) Gunzip-4 389ms ± 4% 384ms ± 0% ~ (p=0.155 n=30+23) HTTPClientServer-4 536µs ± 1% 537µs ± 1% ~ (p=0.236 n=30+30) JSONEncode-4 179ms ± 1% 182ms ± 6% ~ (p=0.763 n=24+30) JSONDecode-4 843ms ± 0% 839ms ± 6% -0.42% (p=0.003 n=25+30) Mandelbrot200-4 46.5ms ± 0% 46.5ms ± 0% +0.02% (p=0.000 n=26+26) GoParse-4 44.3ms ± 6% 43.3ms ± 0% ~ (p=0.067 n=30+27) RegexpMatchEasy0_32-4 1.07µs ± 7% 1.07µs ± 4% ~ (p=0.835 n=30+30) RegexpMatchEasy0_1K-4 5.51µs ± 0% 5.49µs ± 0% -0.35% (p=0.000 n=23+26) RegexpMatchEasy1_32-4 1.01µs ± 0% 1.02µs ± 4% +0.96% (p=0.014 n=24+30) RegexpMatchEasy1_1K-4 7.43µs ± 0% 7.18µs ± 0% -3.41% (p=0.000 n=23+24) RegexpMatchMedium_32-4 1.78µs ± 0% 1.81µs ± 4% +1.47% (p=0.012 n=23+30) RegexpMatchMedium_1K-4 547µs ± 1% 542µs ± 3% -0.90% (p=0.003 n=24+30) RegexpMatchHard_32-4 30.4µs ± 0% 29.7µs ± 0% -2.15% (p=0.000 n=19+23) RegexpMatchHard_1K-4 913µs ± 0% 915µs ± 6% +0.25% (p=0.012 n=24+30) Revcomp-4 6.32s ± 1% 6.42s ± 4% ~ (p=0.342 n=25+30) Template-4 868ms ± 6% 878ms ± 6% +1.15% (p=0.000 n=30+30) TimeParse-4 4.57µs ± 4% 4.59µs ± 3% +0.65% (p=0.010 n=29+30) TimeFormat-4 4.51µs ± 0% 4.50µs ± 0% -0.27% (p=0.000 n=27+24) [Geo mean] 695µs 689µs -0.92% name old speed new speed delta GobDecode-4 8.19MB/s ± 0% 8.12MB/s ± 4% ~ (p=0.680 n=24+30) GobEncode-4 9.76MB/s ± 7% 9.96MB/s ± 0% ~ (p=0.616 n=30+23) Gzip-4 4.84MB/s ± 0% 4.89MB/s ± 4% +1.16% (p=0.030 n=24+30) Gunzip-4 49.9MB/s ± 4% 50.6MB/s ± 0% ~ (p=0.162 n=30+23) JSONEncode-4 10.9MB/s ± 1% 10.7MB/s ± 6% ~ (p=0.575 n=24+30) JSONDecode-4 2.30MB/s ± 0% 2.32MB/s ± 5% +0.72% (p=0.003 n=22+30) GoParse-4 1.31MB/s ± 6% 1.34MB/s ± 0% +2.26% (p=0.002 n=30+27) RegexpMatchEasy0_32-4 30.0MB/s ± 6% 30.0MB/s ± 4% ~ (p=1.000 n=30+30) RegexpMatchEasy0_1K-4 186MB/s ± 0% 187MB/s ± 0% +0.35% (p=0.000 n=23+26) RegexpMatchEasy1_32-4 31.8MB/s ± 0% 31.5MB/s ± 4% -0.92% (p=0.012 n=25+30) RegexpMatchEasy1_1K-4 138MB/s ± 0% 143MB/s ± 0% +3.53% (p=0.000 n=23+24) RegexpMatchMedium_32-4 560kB/s ± 0% 553kB/s ± 4% -1.19% (p=0.005 n=23+30) RegexpMatchMedium_1K-4 1.87MB/s ± 0% 1.89MB/s ± 3% +1.04% (p=0.002 n=24+30) RegexpMatchHard_32-4 1.05MB/s ± 0% 1.08MB/s ± 0% +2.40% (p=0.000 n=19+23) RegexpMatchHard_1K-4 1.12MB/s ± 0% 1.12MB/s ± 5% +0.12% (p=0.006 n=25+30) Revcomp-4 40.2MB/s ± 1% 39.6MB/s ± 4% ~ (p=0.242 n=25+30) Template-4 2.24MB/s ± 6% 2.21MB/s ± 6% -1.15% (p=0.000 n=30+30) [Geo mean] 7.87MB/s 7.91MB/s +0.44% Change-Id: If374cb7abf83537aa0a176f73c0f736f7800db03 Reviewed-on: https://go-review.googlesource.com/108735Reviewed-by: Cherry Zhang <cherryyz@google.com> Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>

cmd/compile: optimize ARM64 with shifted register indexed load/store
ARM64 supports efficient instructions which combine shift, addition, load/store together. Such as "MOVD (R0)(R1<<3), R2" and "MOVWU R6, (R4)(R1<<2)". This CL optimizes the compiler to emit such efficient instuctions. And below is some test data. 1. binary size before/after binary size change pkg/linux_arm64 +80.1KB pkg/tool/linux_arm64 +121.9KB go -4.3KB gofmt -64KB 2. go1 benchmark There is big improvement for the test case Fannkuch11, and slight improvement for sme others, excluding noise. name old time/op new time/op delta BinaryTree17-4 43.9s ± 2% 44.0s ± 2% ~ (p=0.820 n=30+30) Fannkuch11-4 30.6s ± 2% 24.5s ± 3% -19.93% (p=0.000 n=25+30) FmtFprintfEmpty-4 500ns ± 0% 499ns ± 0% -0.11% (p=0.000 n=23+25) FmtFprintfString-4 1.03µs ± 0% 1.04µs ± 3% ~ (p=0.065 n=29+30) FmtFprintfInt-4 1.15µs ± 3% 1.15µs ± 4% -0.56% (p=0.000 n=30+30) FmtFprintfIntInt-4 1.80µs ± 5% 1.82µs ± 0% ~ (p=0.094 n=30+24) FmtFprintfPrefixedInt-4 2.17µs ± 5% 2.20µs ± 0% ~ (p=0.100 n=30+23) FmtFprintfFloat-4 3.08µs ± 3% 3.09µs ± 4% ~ (p=0.123 n=30+30) FmtManyArgs-4 7.41µs ± 4% 7.17µs ± 1% -3.26% (p=0.000 n=30+23) GobDecode-4 93.7ms ± 0% 94.7ms ± 4% ~ (p=0.685 n=24+30) GobEncode-4 78.7ms ± 7% 77.1ms ± 0% ~ (p=0.729 n=30+23) Gzip-4 4.01s ± 0% 3.97s ± 5% -1.11% (p=0.037 n=24+30) Gunzip-4 389ms ± 4% 384ms ± 0% ~ (p=0.155 n=30+23) HTTPClientServer-4 536µs ± 1% 537µs ± 1% ~ (p=0.236 n=30+30) JSONEncode-4 179ms ± 1% 182ms ± 6% ~ (p=0.763 n=24+30) JSONDecode-4 843ms ± 0% 839ms ± 6% -0.42% (p=0.003 n=25+30) Mandelbrot200-4 46.5ms ± 0% 46.5ms ± 0% +0.02% (p=0.000 n=26+26) GoParse-4 44.3ms ± 6% 43.3ms ± 0% ~ (p=0.067 n=30+27) RegexpMatchEasy0_32-4 1.07µs ± 7% 1.07µs ± 4% ~ (p=0.835 n=30+30) RegexpMatchEasy0_1K-4 5.51µs ± 0% 5.49µs ± 0% -0.35% (p=0.000 n=23+26) RegexpMatchEasy1_32-4 1.01µs ± 0% 1.02µs ± 4% +0.96% (p=0.014 n=24+30) RegexpMatchEasy1_1K-4 7.43µs ± 0% 7.18µs ± 0% -3.41% (p=0.000 n=23+24) RegexpMatchMedium_32-4 1.78µs ± 0% 1.81µs ± 4% +1.47% (p=0.012 n=23+30) RegexpMatchMedium_1K-4 547µs ± 1% 542µs ± 3% -0.90% (p=0.003 n=24+30) RegexpMatchHard_32-4 30.4µs ± 0% 29.7µs ± 0% -2.15% (p=0.000 n=19+23) RegexpMatchHard_1K-4 913µs ± 0% 915µs ± 6% +0.25% (p=0.012 n=24+30) Revcomp-4 6.32s ± 1% 6.42s ± 4% ~ (p=0.342 n=25+30) Template-4 868ms ± 6% 878ms ± 6% +1.15% (p=0.000 n=30+30) TimeParse-4 4.57µs ± 4% 4.59µs ± 3% +0.65% (p=0.010 n=29+30) TimeFormat-4 4.51µs ± 0% 4.50µs ± 0% -0.27% (p=0.000 n=27+24) [Geo mean] 695µs 689µs -0.92% name old speed new speed delta GobDecode-4 8.19MB/s ± 0% 8.12MB/s ± 4% ~ (p=0.680 n=24+30) GobEncode-4 9.76MB/s ± 7% 9.96MB/s ± 0% ~ (p=0.616 n=30+23) Gzip-4 4.84MB/s ± 0% 4.89MB/s ± 4% +1.16% (p=0.030 n=24+30) Gunzip-4 49.9MB/s ± 4% 50.6MB/s ± 0% ~ (p=0.162 n=30+23) JSONEncode-4 10.9MB/s ± 1% 10.7MB/s ± 6% ~ (p=0.575 n=24+30) JSONDecode-4 2.30MB/s ± 0% 2.32MB/s ± 5% +0.72% (p=0.003 n=22+30) GoParse-4 1.31MB/s ± 6% 1.34MB/s ± 0% +2.26% (p=0.002 n=30+27) RegexpMatchEasy0_32-4 30.0MB/s ± 6% 30.0MB/s ± 4% ~ (p=1.000 n=30+30) RegexpMatchEasy0_1K-4 186MB/s ± 0% 187MB/s ± 0% +0.35% (p=0.000 n=23+26) RegexpMatchEasy1_32-4 31.8MB/s ± 0% 31.5MB/s ± 4% -0.92% (p=0.012 n=25+30) RegexpMatchEasy1_1K-4 138MB/s ± 0% 143MB/s ± 0% +3.53% (p=0.000 n=23+24) RegexpMatchMedium_32-4 560kB/s ± 0% 553kB/s ± 4% -1.19% (p=0.005 n=23+30) RegexpMatchMedium_1K-4 1.87MB/s ± 0% 1.89MB/s ± 3% +1.04% (p=0.002 n=24+30) RegexpMatchHard_32-4 1.05MB/s ± 0% 1.08MB/s ± 0% +2.40% (p=0.000 n=19+23) RegexpMatchHard_1K-4 1.12MB/s ± 0% 1.12MB/s ± 5% +0.12% (p=0.006 n=25+30) Revcomp-4 40.2MB/s ± 1% 39.6MB/s ± 4% ~ (p=0.242 n=25+30) Template-4 2.24MB/s ± 6% 2.21MB/s ± 6% -1.15% (p=0.000 n=30+30) [Geo mean] 7.87MB/s 7.91MB/s +0.44% Change-Id: If374cb7abf83537aa0a176f73c0f736f7800db03 Reviewed-on: https://go-review.googlesource.com/108735Reviewed-by: Cherry Zhang <cherryyz@google.com> Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
aaf73c6d · Ben Shi · Cherry Zhang · ceda47d0 · aaf73c6d · aaf73c6d
Commit aaf73c6d authored Apr 22, 2018 by Ben Shi Committed by Cherry Zhang Apr 27, 2018
6 changed files
--- a/src/cmd/compile/internal/arm64/ssa.go
+++ b/src/cmd/compile/internal/arm64/ssa.go
@@ -92,6 +92,23 @@ func genshift(s *gc.SSAGenState, as obj.As, r0, r1, r int16, typ int64, n int64)
 	return p
 }

+// generate the memory operand for the indexed load/store instructions
+func genIndexedOperand(v *ssa.Value) obj.Addr {
+	// Reg: base register, Index: (shifted) index register
+	mop := obj.Addr{Type: obj.TYPE_MEM, Reg: v.Args[0].Reg()}
+	switch v.Op {
+	case ssa.OpARM64MOVDloadidx8, ssa.OpARM64MOVDstoreidx8, ssa.OpARM64MOVDstorezeroidx8:
+		mop.Index = arm64.REG_LSL | 3<<5 | v.Args[1].Reg()&31
+	case ssa.OpARM64MOVWloadidx4, ssa.OpARM64MOVWUloadidx4, ssa.OpARM64MOVWstoreidx4, ssa.OpARM64MOVWstorezeroidx4:
+		mop.Index = arm64.REG_LSL | 2<<5 | v.Args[1].Reg()&31
+	case ssa.OpARM64MOVHloadidx2, ssa.OpARM64MOVHUloadidx2, ssa.OpARM64MOVHstoreidx2, ssa.OpARM64MOVHstorezeroidx2:
+		mop.Index = arm64.REG_LSL | 1<<5 | v.Args[1].Reg()&31
+	default: // not shifted
+		mop.Index = v.Args[1].Reg()
+	}
+	return mop
+}
+
 func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 	switch v.Op {
 	case ssa.OpCopy, ssa.OpARM64MOVDreg:
@@ -351,12 +368,14 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		ssa.OpARM64MOVHUloadidx,
 		ssa.OpARM64MOVWloadidx,
 		ssa.OpARM64MOVWUloadidx,
-		ssa.OpARM64MOVDloadidx:
+		ssa.OpARM64MOVDloadidx,
+		ssa.OpARM64MOVHloadidx2,
+		ssa.OpARM64MOVHUloadidx2,
+		ssa.OpARM64MOVWloadidx4,
+		ssa.OpARM64MOVWUloadidx4,
+		ssa.OpARM64MOVDloadidx8:
 		p := s.Prog(v.Op.Asm())
-		p.From.Type = obj.TYPE_MEM
-		p.From.Name = obj.NAME_NONE
-		p.From.Reg = v.Args[0].Reg()
-		p.From.Index = v.Args[1].Reg()
+		p.From = genIndexedOperand(v)
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = v.Reg()
 	case ssa.OpARM64LDAR,
@@ -384,14 +403,14 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 	case ssa.OpARM64MOVBstoreidx,
 		ssa.OpARM64MOVHstoreidx,
 		ssa.OpARM64MOVWstoreidx,
-		ssa.OpARM64MOVDstoreidx:
+		ssa.OpARM64MOVDstoreidx,
+		ssa.OpARM64MOVHstoreidx2,
+		ssa.OpARM64MOVWstoreidx4,
+		ssa.OpARM64MOVDstoreidx8:
 		p := s.Prog(v.Op.Asm())
+		p.To = genIndexedOperand(v)
 		p.From.Type = obj.TYPE_REG
 		p.From.Reg = v.Args[2].Reg()
-		p.To.Type = obj.TYPE_MEM
-		p.To.Name = obj.NAME_NONE
-		p.To.Reg = v.Args[0].Reg()
-		p.To.Index = v.Args[1].Reg()
 	case ssa.OpARM64STP:
 		p := s.Prog(v.Op.Asm())
 		p.From.Type = obj.TYPE_REGREG
@@ -413,14 +432,14 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 	case ssa.OpARM64MOVBstorezeroidx,
 		ssa.OpARM64MOVHstorezeroidx,
 		ssa.OpARM64MOVWstorezeroidx,
-		ssa.OpARM64MOVDstorezeroidx:
+		ssa.OpARM64MOVDstorezeroidx,
+		ssa.OpARM64MOVHstorezeroidx2,
+		ssa.OpARM64MOVWstorezeroidx4,
+		ssa.OpARM64MOVDstorezeroidx8:
 		p := s.Prog(v.Op.Asm())
+		p.To = genIndexedOperand(v)
 		p.From.Type = obj.TYPE_REG
 		p.From.Reg = arm64.REGZERO
-		p.To.Type = obj.TYPE_MEM
-		p.To.Name = obj.NAME_NONE
-		p.To.Reg = v.Args[0].Reg()
-		p.To.Index = v.Args[1].Reg()
 	case ssa.OpARM64MOVQstorezero:
 		p := s.Prog(v.Op.Asm())
 		p.From.Type = obj.TYPE_REGREG

--- a/src/cmd/compile/internal/ssa/gen/ARM64.rules
+++ b/src/cmd/compile/internal/ssa/gen/ARM64.rules
--- a/src/cmd/compile/internal/ssa/gen/ARM64Ops.go
+++ b/src/cmd/compile/internal/ssa/gen/ARM64Ops.go
@@ -324,13 +324,20 @@ func init() {
 		{name: "FMOVDload", argLength: 2, reg: fpload, aux: "SymOff", asm: "FMOVD", typ: "Float64", faultOnNilArg0: true, symEffect: "Read"}, // load from arg0 + auxInt + aux.  arg1=mem.

 		// register indexed load
-		{name: "MOVDloadidx", argLength: 3, reg: gp2load, asm: "MOVD", faultOnNilArg0: true},   // load 64-bit dword from arg0 + arg1, arg2 = mem.
-		{name: "MOVWloadidx", argLength: 3, reg: gp2load, asm: "MOVW", faultOnNilArg0: true},   // load 32-bit word from arg0 + arg1, sign-extended to 64-bit, arg2=mem.
-		{name: "MOVWUloadidx", argLength: 3, reg: gp2load, asm: "MOVWU", faultOnNilArg0: true}, // load 32-bit word from arg0 + arg1, zero-extended to 64-bit, arg2=mem.
-		{name: "MOVHloadidx", argLength: 3, reg: gp2load, asm: "MOVH", faultOnNilArg0: true},   // load 16-bit word from arg0 + arg1, sign-extended to 64-bit, arg2=mem.
-		{name: "MOVHUloadidx", argLength: 3, reg: gp2load, asm: "MOVHU", faultOnNilArg0: true}, // load 16-bit word from arg0 + arg1, zero-extended to 64-bit, arg2=mem.
-		{name: "MOVBloadidx", argLength: 3, reg: gp2load, asm: "MOVB", faultOnNilArg0: true},   // load 8-bit word from arg0 + arg1, sign-extended to 64-bit, arg2=mem.
-		{name: "MOVBUloadidx", argLength: 3, reg: gp2load, asm: "MOVBU", faultOnNilArg0: true}, // load 8-bit word from arg0 + arg1, zero-extended to 64-bit, arg2=mem.
+		{name: "MOVDloadidx", argLength: 3, reg: gp2load, asm: "MOVD"},   // load 64-bit dword from arg0 + arg1, arg2 = mem.
+		{name: "MOVWloadidx", argLength: 3, reg: gp2load, asm: "MOVW"},   // load 32-bit word from arg0 + arg1, sign-extended to 64-bit, arg2=mem.
+		{name: "MOVWUloadidx", argLength: 3, reg: gp2load, asm: "MOVWU"}, // load 32-bit word from arg0 + arg1, zero-extended to 64-bit, arg2=mem.
+		{name: "MOVHloadidx", argLength: 3, reg: gp2load, asm: "MOVH"},   // load 16-bit word from arg0 + arg1, sign-extended to 64-bit, arg2=mem.
+		{name: "MOVHUloadidx", argLength: 3, reg: gp2load, asm: "MOVHU"}, // load 16-bit word from arg0 + arg1, zero-extended to 64-bit, arg2=mem.
+		{name: "MOVBloadidx", argLength: 3, reg: gp2load, asm: "MOVB"},   // load 8-bit word from arg0 + arg1, sign-extended to 64-bit, arg2=mem.
+		{name: "MOVBUloadidx", argLength: 3, reg: gp2load, asm: "MOVBU"}, // load 8-bit word from arg0 + arg1, zero-extended to 64-bit, arg2=mem.
+
+		// shifted register indexed load
+		{name: "MOVHloadidx2", argLength: 3, reg: gp2load, asm: "MOVH"},   // load 16-bit half-word from arg0 + arg1*2, sign-extended to 64-bit, arg2=mem.
+		{name: "MOVHUloadidx2", argLength: 3, reg: gp2load, asm: "MOVHU"}, // load 16-bit half-word from arg0 + arg1*2, zero-extended to 64-bit, arg2=mem.
+		{name: "MOVWloadidx4", argLength: 3, reg: gp2load, asm: "MOVW"},   // load 32-bit word from arg0 + arg1*4, sign-extended to 64-bit, arg2=mem.
+		{name: "MOVWUloadidx4", argLength: 3, reg: gp2load, asm: "MOVWU"}, // load 32-bit word from arg0 + arg1*4, zero-extended to 64-bit, arg2=mem.
+		{name: "MOVDloadidx8", argLength: 3, reg: gp2load, asm: "MOVD"},   // load 64-bit double-word from arg0 + arg1*8, arg2 = mem.

 		{name: "MOVBstore", argLength: 3, reg: gpstore, aux: "SymOff", asm: "MOVB", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"},   // store 1 byte of arg1 to arg0 + auxInt + aux.  arg2=mem.
 		{name: "MOVHstore", argLength: 3, reg: gpstore, aux: "SymOff", asm: "MOVH", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"},   // store 2 bytes of arg1 to arg0 + auxInt + aux.  arg2=mem.
@@ -341,10 +348,15 @@ func init() {
 		{name: "FMOVDstore", argLength: 3, reg: fpstore, aux: "SymOff", asm: "FMOVD", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // store 8 bytes of arg1 to arg0 + auxInt + aux.  arg2=mem.

 		// register indexed store
-		{name: "MOVBstoreidx", argLength: 4, reg: gpstore2, asm: "MOVB", typ: "Mem", faultOnNilArg0: true}, // store 1 byte of arg2 to arg0 + arg1, arg3 = mem.
-		{name: "MOVHstoreidx", argLength: 4, reg: gpstore2, asm: "MOVH", typ: "Mem", faultOnNilArg0: true}, // store 2 bytes of arg2 to arg0 + arg1, arg3 = mem.
-		{name: "MOVWstoreidx", argLength: 4, reg: gpstore2, asm: "MOVW", typ: "Mem", faultOnNilArg0: true}, // store 4 bytes of arg2 to arg0 + arg1, arg3 = mem.
-		{name: "MOVDstoreidx", argLength: 4, reg: gpstore2, asm: "MOVD", typ: "Mem", faultOnNilArg0: true}, // store 8 bytes of arg2 to arg0 + arg1, arg3 = mem.
+		{name: "MOVBstoreidx", argLength: 4, reg: gpstore2, asm: "MOVB", typ: "Mem"}, // store 1 byte of arg2 to arg0 + arg1, arg3 = mem.
+		{name: "MOVHstoreidx", argLength: 4, reg: gpstore2, asm: "MOVH", typ: "Mem"}, // store 2 bytes of arg2 to arg0 + arg1, arg3 = mem.
+		{name: "MOVWstoreidx", argLength: 4, reg: gpstore2, asm: "MOVW", typ: "Mem"}, // store 4 bytes of arg2 to arg0 + arg1, arg3 = mem.
+		{name: "MOVDstoreidx", argLength: 4, reg: gpstore2, asm: "MOVD", typ: "Mem"}, // store 8 bytes of arg2 to arg0 + arg1, arg3 = mem.
+
+		// shifted register indexed store
+		{name: "MOVHstoreidx2", argLength: 4, reg: gpstore2, asm: "MOVH", typ: "Mem"}, // store 2 bytes of arg2 to arg0 + arg1*2, arg3 = mem.
+		{name: "MOVWstoreidx4", argLength: 4, reg: gpstore2, asm: "MOVW", typ: "Mem"}, // store 4 bytes of arg2 to arg0 + arg1*4, arg3 = mem.
+		{name: "MOVDstoreidx8", argLength: 4, reg: gpstore2, asm: "MOVD", typ: "Mem"}, // store 8 bytes of arg2 to arg0 + arg1*8, arg3 = mem.

 		{name: "MOVBstorezero", argLength: 2, reg: gpstore0, aux: "SymOff", asm: "MOVB", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // store 1 byte of zero to arg0 + auxInt + aux.  arg1=mem.
 		{name: "MOVHstorezero", argLength: 2, reg: gpstore0, aux: "SymOff", asm: "MOVH", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // store 2 bytes of zero to arg0 + auxInt + aux.  arg1=mem.
@@ -353,10 +365,15 @@ func init() {
 		{name: "MOVQstorezero", argLength: 2, reg: gpstore0, aux: "SymOff", asm: "STP", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"},  // store 16 bytes of zero to arg0 + auxInt + aux.  arg1=mem.

 		// register indexed store zero
-		{name: "MOVBstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVB", typ: "Mem", faultOnNilArg0: true}, // store 1 byte of zero to arg0 + arg1, arg2 = mem.
-		{name: "MOVHstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVH", typ: "Mem", faultOnNilArg0: true}, // store 2 bytes of zero to arg0 + arg1, arg2 = mem.
-		{name: "MOVWstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVW", typ: "Mem", faultOnNilArg0: true}, // store 4 bytes of zero to arg0 + arg1, arg2 = mem.
-		{name: "MOVDstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVD", typ: "Mem", faultOnNilArg0: true}, // store 8 bytes of zero to arg0 + arg1, arg2 = mem.
+		{name: "MOVBstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVB", typ: "Mem"}, // store 1 byte of zero to arg0 + arg1, arg2 = mem.
+		{name: "MOVHstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVH", typ: "Mem"}, // store 2 bytes of zero to arg0 + arg1, arg2 = mem.
+		{name: "MOVWstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVW", typ: "Mem"}, // store 4 bytes of zero to arg0 + arg1, arg2 = mem.
+		{name: "MOVDstorezeroidx", argLength: 3, reg: gpstore, asm: "MOVD", typ: "Mem"}, // store 8 bytes of zero to arg0 + arg1, arg2 = mem.
+
+		// shifted register indexed store zero
+		{name: "MOVHstorezeroidx2", argLength: 3, reg: gpstore, asm: "MOVH", typ: "Mem"}, // store 2 bytes of zero to arg0 + arg1*2, arg2 = mem.
+		{name: "MOVWstorezeroidx4", argLength: 3, reg: gpstore, asm: "MOVW", typ: "Mem"}, // store 4 bytes of zero to arg0 + arg1*4, arg2 = mem.
+		{name: "MOVDstorezeroidx8", argLength: 3, reg: gpstore, asm: "MOVD", typ: "Mem"}, // store 8 bytes of zero to arg0 + arg1*8, arg2 = mem.

 		{name: "FMOVDgpfp", argLength: 1, reg: gpfp, asm: "FMOVD"}, // move int64 to float64 (no conversion)
 		{name: "FMOVDfpgp", argLength: 1, reg: fpgp, asm: "FMOVD"}, // move float64 to int64 (no conversion)

--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
--- a/src/cmd/compile/internal/ssa/rewriteARM64.go
+++ b/src/cmd/compile/internal/ssa/rewriteARM64.go
--- a/test/codegen/memcombine.go
+++ b/test/codegen/memcombine.go
@@ -110,6 +110,21 @@ func load_byte2_uint16(s []byte) uint16 {
 	return uint16(s[0]) | uint16(s[1])<<8
 }

+func load_byte2_uint16_idx(s []byte, idx int) uint16 {
+	// arm64:`MOVHU\s\(R[0-9]+\)\(R[0-9]+<<1\)`,-`ORR`,-`MOVB`
+	return uint16(s[idx<<1]) | uint16(s[(idx<<1)+1])<<8
+}
+
+func load_byte4_uint32_idx(s []byte, idx int) uint32 {
+	// arm64:`MOVWU\s\(R[0-9]+\)\(R[0-9]+<<2\)`,-`ORR`,-`MOV[BH]`
+	return uint32(s[idx<<2]) | uint32(s[(idx<<2)+1])<<8 | uint32(s[(idx<<2)+2])<<16 | uint32(s[(idx<<2)+3])<<24
+}
+
+func load_byte8_uint64_idx(s []byte, idx int) uint64 {
+	// arm64:`MOVD\s\(R[0-9]+\)\(R[0-9]+<<3\)`,-`ORR`,-`MOV[BHW]`
+	return uint64(s[idx<<3]) | uint64(s[(idx<<3)+1])<<8 | uint64(s[(idx<<3)+2])<<16 | uint64(s[(idx<<3)+3])<<24 | uint64(s[(idx<<3)+4])<<32 | uint64(s[(idx<<3)+5])<<40 | uint64(s[(idx<<3)+6])<<48 | uint64(s[(idx<<3)+7])<<56
+}
+
 // Check load combining across function calls.

 func fcall_byte(a, b byte) (byte, byte) {
@@ -268,6 +283,32 @@ func zero_byte_16(b []byte) {
 	b[12], b[13], b[14], b[15] = 0, 0, 0, 0 // arm64:"STP",-"MOVB",-"MOVH",-"MOVW"
 }

+func zero_byte_2_idx(b []byte, idx int) {
+	// arm64: `MOVH\sZR,\s\(R[0-9]+\)\(R[0-9]+<<1\)`,-`MOVB`
+	b[(idx<<1)+0] = 0
+	b[(idx<<1)+1] = 0
+}
+
+func zero_byte_4_idx(b []byte, idx int) {
+	// arm64: `MOVW\sZR,\s\(R[0-9]+\)\(R[0-9]+<<2\)`,-`MOV[BH]`
+	b[(idx<<2)+0] = 0
+	b[(idx<<2)+1] = 0
+	b[(idx<<2)+2] = 0
+	b[(idx<<2)+3] = 0
+}
+
+func zero_byte_8_idx(b []byte, idx int) {
+	// arm64: `MOVD\sZR,\s\(R[0-9]+\)\(R[0-9]+<<3\)`,-`MOV[BHW]`
+	b[(idx<<3)+0] = 0
+	b[(idx<<3)+1] = 0
+	b[(idx<<3)+2] = 0
+	b[(idx<<3)+3] = 0
+	b[(idx<<3)+4] = 0
+	b[(idx<<3)+5] = 0
+	b[(idx<<3)+6] = 0
+	b[(idx<<3)+7] = 0
+}
+
 func zero_byte_30(a *[30]byte) {
 	*a = [30]byte{} // arm64:"STP",-"MOVB",-"MOVH",-"MOVW"
 }