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

ARM64 supports load/store instructions with a memory operand that the address is calculated by base register + index register. In this CL, 1. Some rules are added to the compile's ARM64 backend to emit such efficient instructions. 2. A wrong rule of load combination is fixed. The go1 benchmark does show improvement. name old time/op new time/op delta BinaryTree17-4 44.5s ± 2% 44.1s ± 1% -0.81% (p=0.000 n=28+29) Fannkuch11-4 32.7s ± 3% 30.5s ± 0% -6.79% (p=0.000 n=30+26) FmtFprintfEmpty-4 499ns ± 0% 506ns ± 5% +1.39% (p=0.003 n=25+30) FmtFprintfString-4 1.07µs ± 0% 1.04µs ± 4% -3.17% (p=0.000 n=23+30) FmtFprintfInt-4 1.15µs ± 4% 1.13µs ± 0% -1.55% (p=0.000 n=30+23) FmtFprintfIntInt-4 1.77µs ± 4% 1.74µs ± 0% -1.71% (p=0.000 n=30+24) FmtFprintfPrefixedInt-4 2.37µs ± 5% 2.12µs ± 0% -10.56% (p=0.000 n=30+23) FmtFprintfFloat-4 3.03µs ± 1% 3.03µs ± 4% -0.13% (p=0.003 n=25+30) FmtManyArgs-4 7.38µs ± 1% 7.43µs ± 4% +0.59% (p=0.003 n=25+30) GobDecode-4 101ms ± 6% 95ms ± 5% -5.55% (p=0.000 n=30+30) GobEncode-4 78.0ms ± 4% 78.8ms ± 6% +1.05% (p=0.000 n=30+30) Gzip-4 4.25s ± 0% 4.27s ± 4% +0.45% (p=0.003 n=24+30) Gunzip-4 428ms ± 1% 420ms ± 0% -1.88% (p=0.000 n=23+23) HTTPClientServer-4 549µs ± 1% 541µs ± 1% -1.56% (p=0.000 n=29+29) JSONEncode-4 194ms ± 0% 188ms ± 4% ~ (p=0.417 n=23+30) JSONDecode-4 890ms ± 5% 831ms ± 0% -6.55% (p=0.000 n=30+23) Mandelbrot200-4 47.3ms ± 2% 46.5ms ± 0% ~ (p=0.980 n=30+26) GoParse-4 43.1ms ± 6% 43.8ms ± 6% +1.65% (p=0.000 n=30+30) RegexpMatchEasy0_32-4 1.06µs ± 0% 1.07µs ± 3% ~ (p=0.092 n=23+30) RegexpMatchEasy0_1K-4 5.53µs ± 0% 5.51µs ± 0% -0.24% (p=0.000 n=25+25) RegexpMatchEasy1_32-4 1.02µs ± 3% 1.01µs ± 0% -1.27% (p=0.000 n=30+24) RegexpMatchEasy1_1K-4 7.26µs ± 0% 7.33µs ± 0% +0.95% (p=0.000 n=23+26) RegexpMatchMedium_32-4 1.84µs ± 7% 1.79µs ± 1% ~ (p=0.333 n=30+23) RegexpMatchMedium_1K-4 553µs ± 0% 547µs ± 0% -1.14% (p=0.000 n=24+22) RegexpMatchHard_32-4 30.8µs ± 1% 30.3µs ± 0% -1.40% (p=0.000 n=24+24) RegexpMatchHard_1K-4 928µs ± 0% 929µs ± 5% +0.12% (p=0.013 n=23+30) Revcomp-4 8.13s ± 4% 6.32s ± 1% -22.23% (p=0.000 n=30+23) Template-4 899ms ± 6% 854ms ± 1% -5.01% (p=0.000 n=30+24) TimeParse-4 4.66µs ± 4% 4.59µs ± 1% -1.57% (p=0.000 n=30+23) TimeFormat-4 4.58µs ± 0% 4.61µs ± 0% +0.57% (p=0.000 n=26+24) [Geo mean] 717µs 698µs -2.55% name old speed new speed delta GobDecode-4 7.63MB/s ± 6% 8.08MB/s ± 5% +5.88% (p=0.000 n=30+30) GobEncode-4 9.85MB/s ± 4% 9.75MB/s ± 6% -1.04% (p=0.000 n=30+30) Gzip-4 4.56MB/s ± 0% 4.55MB/s ± 4% -0.36% (p=0.003 n=24+30) Gunzip-4 45.3MB/s ± 1% 46.2MB/s ± 0% +1.92% (p=0.000 n=23+23) JSONEncode-4 10.0MB/s ± 0% 10.4MB/s ± 4% ~ (p=0.403 n=23+30) JSONDecode-4 2.18MB/s ± 5% 2.33MB/s ± 0% +6.91% (p=0.000 n=30+23) GoParse-4 1.34MB/s ± 5% 1.32MB/s ± 5% -1.66% (p=0.000 n=30+30) RegexpMatchEasy0_32-4 30.2MB/s ± 0% 29.8MB/s ± 3% ~ (p=0.099 n=23+30) RegexpMatchEasy0_1K-4 185MB/s ± 0% 186MB/s ± 0% +0.24% (p=0.000 n=25+25) RegexpMatchEasy1_32-4 31.4MB/s ± 3% 31.8MB/s ± 0% +1.24% (p=0.000 n=30+24) RegexpMatchEasy1_1K-4 141MB/s ± 0% 140MB/s ± 0% -0.94% (p=0.000 n=23+26) RegexpMatchMedium_32-4 541kB/s ± 6% 560kB/s ± 0% +3.45% (p=0.000 n=30+23) RegexpMatchMedium_1K-4 1.85MB/s ± 0% 1.87MB/s ± 0% +1.08% (p=0.000 n=24+23) RegexpMatchHard_32-4 1.04MB/s ± 1% 1.06MB/s ± 1% +1.48% (p=0.000 n=24+24) RegexpMatchHard_1K-4 1.10MB/s ± 0% 1.10MB/s ± 5% +0.15% (p=0.004 n=23+30) Revcomp-4 31.3MB/s ± 4% 40.2MB/s ± 1% +28.52% (p=0.000 n=30+23) Template-4 2.16MB/s ± 6% 2.27MB/s ± 1% +5.18% (p=0.000 n=30+24) [Geo mean] 7.57MB/s 7.79MB/s +2.98% fixes #24907 Change-Id: I94afd0e3f53d62a1cf5e452f3dd6daf61be21785 Reviewed-on: https://go-review.googlesource.com/107376 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>

cmd/compile: optimize ARM64 with register indexed load/store
ARM64 supports load/store instructions with a memory operand that the address is calculated by base register + index register. In this CL, 1. Some rules are added to the compile's ARM64 backend to emit such efficient instructions. 2. A wrong rule of load combination is fixed. The go1 benchmark does show improvement. name old time/op new time/op delta BinaryTree17-4 44.5s ± 2% 44.1s ± 1% -0.81% (p=0.000 n=28+29) Fannkuch11-4 32.7s ± 3% 30.5s ± 0% -6.79% (p=0.000 n=30+26) FmtFprintfEmpty-4 499ns ± 0% 506ns ± 5% +1.39% (p=0.003 n=25+30) FmtFprintfString-4 1.07µs ± 0% 1.04µs ± 4% -3.17% (p=0.000 n=23+30) FmtFprintfInt-4 1.15µs ± 4% 1.13µs ± 0% -1.55% (p=0.000 n=30+23) FmtFprintfIntInt-4 1.77µs ± 4% 1.74µs ± 0% -1.71% (p=0.000 n=30+24) FmtFprintfPrefixedInt-4 2.37µs ± 5% 2.12µs ± 0% -10.56% (p=0.000 n=30+23) FmtFprintfFloat-4 3.03µs ± 1% 3.03µs ± 4% -0.13% (p=0.003 n=25+30) FmtManyArgs-4 7.38µs ± 1% 7.43µs ± 4% +0.59% (p=0.003 n=25+30) GobDecode-4 101ms ± 6% 95ms ± 5% -5.55% (p=0.000 n=30+30) GobEncode-4 78.0ms ± 4% 78.8ms ± 6% +1.05% (p=0.000 n=30+30) Gzip-4 4.25s ± 0% 4.27s ± 4% +0.45% (p=0.003 n=24+30) Gunzip-4 428ms ± 1% 420ms ± 0% -1.88% (p=0.000 n=23+23) HTTPClientServer-4 549µs ± 1% 541µs ± 1% -1.56% (p=0.000 n=29+29) JSONEncode-4 194ms ± 0% 188ms ± 4% ~ (p=0.417 n=23+30) JSONDecode-4 890ms ± 5% 831ms ± 0% -6.55% (p=0.000 n=30+23) Mandelbrot200-4 47.3ms ± 2% 46.5ms ± 0% ~ (p=0.980 n=30+26) GoParse-4 43.1ms ± 6% 43.8ms ± 6% +1.65% (p=0.000 n=30+30) RegexpMatchEasy0_32-4 1.06µs ± 0% 1.07µs ± 3% ~ (p=0.092 n=23+30) RegexpMatchEasy0_1K-4 5.53µs ± 0% 5.51µs ± 0% -0.24% (p=0.000 n=25+25) RegexpMatchEasy1_32-4 1.02µs ± 3% 1.01µs ± 0% -1.27% (p=0.000 n=30+24) RegexpMatchEasy1_1K-4 7.26µs ± 0% 7.33µs ± 0% +0.95% (p=0.000 n=23+26) RegexpMatchMedium_32-4 1.84µs ± 7% 1.79µs ± 1% ~ (p=0.333 n=30+23) RegexpMatchMedium_1K-4 553µs ± 0% 547µs ± 0% -1.14% (p=0.000 n=24+22) RegexpMatchHard_32-4 30.8µs ± 1% 30.3µs ± 0% -1.40% (p=0.000 n=24+24) RegexpMatchHard_1K-4 928µs ± 0% 929µs ± 5% +0.12% (p=0.013 n=23+30) Revcomp-4 8.13s ± 4% 6.32s ± 1% -22.23% (p=0.000 n=30+23) Template-4 899ms ± 6% 854ms ± 1% -5.01% (p=0.000 n=30+24) TimeParse-4 4.66µs ± 4% 4.59µs ± 1% -1.57% (p=0.000 n=30+23) TimeFormat-4 4.58µs ± 0% 4.61µs ± 0% +0.57% (p=0.000 n=26+24) [Geo mean] 717µs 698µs -2.55% name old speed new speed delta GobDecode-4 7.63MB/s ± 6% 8.08MB/s ± 5% +5.88% (p=0.000 n=30+30) GobEncode-4 9.85MB/s ± 4% 9.75MB/s ± 6% -1.04% (p=0.000 n=30+30) Gzip-4 4.56MB/s ± 0% 4.55MB/s ± 4% -0.36% (p=0.003 n=24+30) Gunzip-4 45.3MB/s ± 1% 46.2MB/s ± 0% +1.92% (p=0.000 n=23+23) JSONEncode-4 10.0MB/s ± 0% 10.4MB/s ± 4% ~ (p=0.403 n=23+30) JSONDecode-4 2.18MB/s ± 5% 2.33MB/s ± 0% +6.91% (p=0.000 n=30+23) GoParse-4 1.34MB/s ± 5% 1.32MB/s ± 5% -1.66% (p=0.000 n=30+30) RegexpMatchEasy0_32-4 30.2MB/s ± 0% 29.8MB/s ± 3% ~ (p=0.099 n=23+30) RegexpMatchEasy0_1K-4 185MB/s ± 0% 186MB/s ± 0% +0.24% (p=0.000 n=25+25) RegexpMatchEasy1_32-4 31.4MB/s ± 3% 31.8MB/s ± 0% +1.24% (p=0.000 n=30+24) RegexpMatchEasy1_1K-4 141MB/s ± 0% 140MB/s ± 0% -0.94% (p=0.000 n=23+26) RegexpMatchMedium_32-4 541kB/s ± 6% 560kB/s ± 0% +3.45% (p=0.000 n=30+23) RegexpMatchMedium_1K-4 1.85MB/s ± 0% 1.87MB/s ± 0% +1.08% (p=0.000 n=24+23) RegexpMatchHard_32-4 1.04MB/s ± 1% 1.06MB/s ± 1% +1.48% (p=0.000 n=24+24) RegexpMatchHard_1K-4 1.10MB/s ± 0% 1.10MB/s ± 5% +0.15% (p=0.004 n=23+30) Revcomp-4 31.3MB/s ± 4% 40.2MB/s ± 1% +28.52% (p=0.000 n=30+23) Template-4 2.16MB/s ± 6% 2.27MB/s ± 1% +5.18% (p=0.000 n=30+24) [Geo mean] 7.57MB/s 7.79MB/s +2.98% fixes #24907 Change-Id: I94afd0e3f53d62a1cf5e452f3dd6daf61be21785 Reviewed-on: https://go-review.googlesource.com/107376 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
34f5f8a5 · Ben Shi · Cherry Zhang · d5a52e70 · 34f5f8a5 · 34f5f8a5
Commit 34f5f8a5 authored Apr 16, 2018 by Ben Shi Committed by Cherry Zhang Apr 19, 2018
6 changed files
--- a/src/cmd/compile/internal/arm64/ssa.go
+++ b/src/cmd/compile/internal/arm64/ssa.go
@@ -341,6 +341,20 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		gc.AddAux(&p.From, v)
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = v.Reg()
+	case ssa.OpARM64MOVBloadidx,
+		ssa.OpARM64MOVBUloadidx,
+		ssa.OpARM64MOVHloadidx,
+		ssa.OpARM64MOVHUloadidx,
+		ssa.OpARM64MOVWloadidx,
+		ssa.OpARM64MOVWUloadidx,
+		ssa.OpARM64MOVDloadidx:
+		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.To.Type = obj.TYPE_REG
+		p.To.Reg = v.Reg()
 	case ssa.OpARM64LDAR,
 		ssa.OpARM64LDARW:
 		p := s.Prog(v.Op.Asm())
@@ -363,6 +377,17 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		p.To.Type = obj.TYPE_MEM
 		p.To.Reg = v.Args[0].Reg()
 		gc.AddAux(&p.To, v)
+	case ssa.OpARM64MOVBstoreidx,
+		ssa.OpARM64MOVHstoreidx,
+		ssa.OpARM64MOVWstoreidx,
+		ssa.OpARM64MOVDstoreidx:
+		p := s.Prog(v.Op.Asm())
+		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
@@ -381,6 +406,17 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		p.To.Type = obj.TYPE_MEM
 		p.To.Reg = v.Args[0].Reg()
 		gc.AddAux(&p.To, v)
+	case ssa.OpARM64MOVBstorezeroidx,
+		ssa.OpARM64MOVHstorezeroidx,
+		ssa.OpARM64MOVWstorezeroidx,
+		ssa.OpARM64MOVDstorezeroidx:
+		p := s.Prog(v.Op.Asm())
+		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
@@ -144,6 +144,7 @@ func init() {
 		gp2flags1 = regInfo{inputs: []regMask{gp, gp}, outputs: []regMask{gp}}
 		gp22      = regInfo{inputs: []regMask{gpg, gpg}, outputs: []regMask{gp, gp}}
 		gpload    = regInfo{inputs: []regMask{gpspsbg}, outputs: []regMask{gp}}
+		gp2load   = regInfo{inputs: []regMask{gpspsbg, gpg}, outputs: []regMask{gp}}
 		gpstore   = regInfo{inputs: []regMask{gpspsbg, gpg}}
 		gpstore0  = regInfo{inputs: []regMask{gpspsbg}}
 		gpstore2  = regInfo{inputs: []regMask{gpspsbg, gpg, gpg}}
@@ -318,6 +319,15 @@ func init() {
 		{name: "FMOVSload", argLength: 2, reg: fpload, aux: "SymOff", asm: "FMOVS", typ: "Float32", faultOnNilArg0: true, symEffect: "Read"}, // load from arg0 + auxInt + aux.  arg1=mem.
 		{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: "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.
 		{name: "MOVWstore", argLength: 3, reg: gpstore, aux: "SymOff", asm: "MOVW", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"},   // store 4 bytes of arg1 to arg0 + auxInt + aux.  arg2=mem.
@@ -326,12 +336,24 @@ func init() {
 		{name: "FMOVSstore", argLength: 3, reg: fpstore, aux: "SymOff", asm: "FMOVS", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // store 4 bytes of arg1 to arg0 + auxInt + aux.  arg2=mem.
 		{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: "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.
 		{name: "MOVWstorezero", argLength: 2, reg: gpstore0, aux: "SymOff", asm: "MOVW", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // store 4 bytes of zero to arg0 + auxInt + aux.  arg1=mem.
 		{name: "MOVDstorezero", argLength: 2, reg: gpstore0, aux: "SymOff", asm: "MOVD", typ: "Mem", faultOnNilArg0: true, symEffect: "Write"}, // store 8 bytes of zero to arg0 + auxInt + aux.  arg1=mem.
 		{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: "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
@@ -1152,6 +1152,13 @@ const (
 	OpARM64MOVDload
 	OpARM64FMOVSload
 	OpARM64FMOVDload
+	OpARM64MOVDloadidx
+	OpARM64MOVWloadidx
+	OpARM64MOVWUloadidx
+	OpARM64MOVHloadidx
+	OpARM64MOVHUloadidx
+	OpARM64MOVBloadidx
+	OpARM64MOVBUloadidx
 	OpARM64MOVBstore
 	OpARM64MOVHstore
 	OpARM64MOVWstore
@@ -1159,11 +1166,19 @@ const (
 	OpARM64STP
 	OpARM64FMOVSstore
 	OpARM64FMOVDstore
+	OpARM64MOVBstoreidx
+	OpARM64MOVHstoreidx
+	OpARM64MOVWstoreidx
+	OpARM64MOVDstoreidx
 	OpARM64MOVBstorezero
 	OpARM64MOVHstorezero
 	OpARM64MOVWstorezero
 	OpARM64MOVDstorezero
 	OpARM64MOVQstorezero
+	OpARM64MOVBstorezeroidx
+	OpARM64MOVHstorezeroidx
+	OpARM64MOVWstorezeroidx
+	OpARM64MOVDstorezeroidx
 	OpARM64FMOVDgpfp
 	OpARM64FMOVDfpgp
 	OpARM64MOVBreg
@@ -15070,6 +15085,111 @@ var opcodeTable = [...]opInfo{
 			},
 		},
 	},
+	{
+		name:           "MOVDloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
+	{
+		name:           "MOVWloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVW,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
+	{
+		name:           "MOVWUloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVWU,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
+	{
+		name:           "MOVHloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVH,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
+	{
+		name:           "MOVHUloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVHU,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
+	{
+		name:           "MOVBloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVB,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
+	{
+		name:           "MOVBUloadidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVBU,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
+		},
+	},
 	{
 		name:           "MOVBstore",
 		auxType:        auxSymOff,
@@ -15169,6 +15289,58 @@ var opcodeTable = [...]opInfo{
 			},
 		},
 	},
+	{
+		name:           "MOVBstoreidx",
+		argLen:         4,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVB,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{2, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
+	{
+		name:           "MOVHstoreidx",
+		argLen:         4,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVH,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{2, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
+	{
+		name:           "MOVWstoreidx",
+		argLen:         4,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVW,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{2, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
+	{
+		name:           "MOVDstoreidx",
+		argLen:         4,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{2, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
 	{
 		name:           "MOVBstorezero",
 		auxType:        auxSymOff,
@@ -15234,6 +15406,54 @@ var opcodeTable = [...]opInfo{
 			},
 		},
 	},
+	{
+		name:           "MOVBstorezeroidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVB,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
+	{
+		name:           "MOVHstorezeroidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVH,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
+	{
+		name:           "MOVWstorezeroidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVW,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
+	{
+		name:           "MOVDstorezeroidx",
+		argLen:         3,
+		faultOnNilArg0: true,
+		asm:            arm64.AMOVD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
+				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
+			},
+		},
+	},
 	{
 		name:   "FMOVDgpfp",
 		argLen: 1,

--- 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
@@ -22,7 +22,7 @@ var sink16 uint16
 func load_le64(b []byte) {
 	// amd64:`MOVQ\s\(.*\),`
 	// s390x:`MOVDBR\s\(.*\),`
-	// arm64:`MOVD\s\(R[0-9]+\),`
+	// arm64:`MOVD\s\(R[0-9]+\),`,-`MOV[BHW]`
 	// ppc64le:`MOVD\s`,-`MOV[BHW]Z`
 	sink64 = binary.LittleEndian.Uint64(b)
 }
@@ -30,7 +30,7 @@ func load_le64(b []byte) {
 func load_le64_idx(b []byte, idx int) {
 	// amd64:`MOVQ\s\(.*\)\(.*\*1\),`
 	// s390x:`MOVDBR\s\(.*\)\(.*\*1\),`
-	// arm64:`MOVD\s\(R[0-9]+\),`
+	// arm64:`MOVD\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[BHW]`
 	// ppc64le:`MOVD\s`,-`MOV[BHW]Z\s`
 	sink64 = binary.LittleEndian.Uint64(b[idx:])
 }
@@ -38,7 +38,7 @@ func load_le64_idx(b []byte, idx int) {
 func load_le32(b []byte) {
 	// amd64:`MOVL\s\(.*\),`           386:`MOVL\s\(.*\),`
 	// s390x:`MOVWBR\s\(.*\),`
-	// arm64:`MOVWU\s\(R[0-9]+\),`
+	// arm64:`MOVWU\s\(R[0-9]+\),`,-`MOV[BH]`
 	// ppc64le:`MOVWZ\s`
 	sink32 = binary.LittleEndian.Uint32(b)
 }
@@ -46,7 +46,7 @@ func load_le32(b []byte) {
 func load_le32_idx(b []byte, idx int) {
 	// amd64:`MOVL\s\(.*\)\(.*\*1\),`  386:`MOVL\s\(.*\)\(.*\*1\),`
 	// s390x:`MOVWBR\s\(.*\)\(.*\*1\),`
-	// arm64:`MOVWU\s\(R[0-9]+\),`
+	// arm64:`MOVWU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[BH]`
 	// ppc64le:`MOVWZ\s`
 	sink32 = binary.LittleEndian.Uint32(b[idx:])
 }
@@ -54,50 +54,54 @@ func load_le32_idx(b []byte, idx int) {
 func load_le16(b []byte) {
 	// amd64:`MOVWLZX\s\(.*\),`
 	// ppc64le:`MOVHZ\s`
+	// arm64:`MOVHU\s\(R[0-9]+\),`,-`MOVB`
 	sink16 = binary.LittleEndian.Uint16(b)
 }

 func load_le16_idx(b []byte, idx int) {
 	// amd64:`MOVWLZX\s\(.*\),`
 	// ppc64le:`MOVHZ\s`
+	// arm64:`MOVHU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOVB`
 	sink16 = binary.LittleEndian.Uint16(b[idx:])
 }

 func load_be64(b []byte) {
 	// amd64:`BSWAPQ`
 	// s390x:`MOVD\s\(.*\),`
-	// arm64:`REV`
+	// arm64:`REV`,`MOVD\s\(R[0-9]+\),`,-`MOV[BHW]`,-`REVW`,-`REV16W`
 	sink64 = binary.BigEndian.Uint64(b)
 }

 func load_be64_idx(b []byte, idx int) {
 	// amd64:`BSWAPQ`
 	// s390x:`MOVD\s\(.*\)\(.*\*1\),`
-	// arm64:`REV`
+	// arm64:`REV`,`MOVD\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[WHB]`,-`REVW`,-`REV16W`
 	sink64 = binary.BigEndian.Uint64(b[idx:])
 }

 func load_be32(b []byte) {
 	// amd64:`BSWAPL`
 	// s390x:`MOVWZ\s\(.*\),`
-	// arm64:`REVW`
+	// arm64:`REVW`,`MOVWU\s\(R[0-9]+\),`,-`MOV[BH]`,-`REV16W`
 	sink32 = binary.BigEndian.Uint32(b)
 }

 func load_be32_idx(b []byte, idx int) {
 	// amd64:`BSWAPL`
 	// s390x:`MOVWZ\s\(.*\)\(.*\*1\),`
-	// arm64:`REVW`
+	// arm64:`REVW`,`MOVWU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[HB]`,-`REV16W`
 	sink32 = binary.BigEndian.Uint32(b[idx:])
 }

 func load_be16(b []byte) {
 	// amd64:`ROLW\s\$8`
+	// arm64: `REV16W`,`MOVHU\s\(R[0-9]+\),`,-`MOVB`
 	sink16 = binary.BigEndian.Uint16(b)
 }

 func load_be16_idx(b []byte, idx int) {
 	// amd64:`ROLW\s\$8`
+	// arm64: `REV16W`,`MOVHU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOVB`
 	sink16 = binary.BigEndian.Uint16(b[idx:])
 }

@@ -162,7 +166,7 @@ func store_le64(b []byte) {

 func store_le64_idx(b []byte, idx int) {
 	// amd64:`MOVQ\s.*\(.*\)\(.*\*1\)$`,-`SHR.`
-	// arm64:`MOVD`,-`MOV[WBH]`
+	// arm64:`MOVD\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,-`MOV[BHW]`
 	// ppc64le:`MOVD\s`,-`MOV[BHW]\s`
 	binary.LittleEndian.PutUint64(b[idx:], sink64)
 }
@@ -176,7 +180,7 @@ func store_le32(b []byte) {

 func store_le32_idx(b []byte, idx int) {
 	// amd64:`MOVL\s`
-	// arm64:`MOVW`,-`MOV[BH]`
+	// arm64:`MOVW\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,-`MOV[BH]`
 	// ppc64le:`MOVW\s`
 	binary.LittleEndian.PutUint32(b[idx:], sink32)
 }
@@ -190,32 +194,32 @@ func store_le16(b []byte) {

 func store_le16_idx(b []byte, idx int) {
 	// amd64:`MOVW\s`
-	// arm64:`MOVH`,-`MOVB`
+	// arm64:`MOVH\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,-`MOVB`
 	// ppc64le(DISABLED):`MOVH\s`
 	binary.LittleEndian.PutUint16(b[idx:], sink16)
 }

 func store_be64(b []byte) {
 	// amd64:`BSWAPQ`,-`SHR.`
-	// arm64:`MOVD`,`REV`,-`MOV[WBH]`
+	// arm64:`MOVD`,`REV`,-`MOV[WBH]`,-`REVW`,-`REV16W`
 	binary.BigEndian.PutUint64(b, sink64)
 }

 func store_be64_idx(b []byte, idx int) {
 	// amd64:`BSWAPQ`,-`SHR.`
-	// arm64:`MOVD`,`REV`,-`MOV[WBH]`
+	// arm64:`REV`,`MOVD\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,-`MOV[BHW]`,-`REV16W`,-`REVW`
 	binary.BigEndian.PutUint64(b[idx:], sink64)
 }

 func store_be32(b []byte) {
 	// amd64:`BSWAPL`,-`SHR.`
-	// arm64:`MOVW`,`REVW`,-`MOV[BH]`
+	// arm64:`MOVW`,`REVW`,-`MOV[BH]`,-`REV16W`
 	binary.BigEndian.PutUint32(b, sink32)
 }

 func store_be32_idx(b []byte, idx int) {
 	// amd64:`BSWAPL`,-`SHR.`
-	// arm64:`MOVW`,`REVW`,-`MOV[BH]`
+	// arm64:`REVW`,`MOVW\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,-`MOV[BH]`,-`REV16W`
 	binary.BigEndian.PutUint32(b[idx:], sink32)
 }

@@ -227,7 +231,7 @@ func store_be16(b []byte) {

 func store_be16_idx(b []byte, idx int) {
 	// amd64:`ROLW\s\$8`,-`SHR.`
-	// arm64:`MOVH`,`REV16W`,-`MOVB`
+	// arm64:`MOVH\sR[0-9]+,\s\(R[0-9]+\)\(R[0-9]+\)`,`REV16W`,-`MOVB`
 	binary.BigEndian.PutUint16(b[idx:], sink16)
 }