// Copyright 2016 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.

// Lowering arithmetic
(Add(Ptr|64|32|16|8)  x y) -> (ADD  x y)
(Add64F x y) -> (FADD x y)
(Add32F x y) -> (FADDS x y)

(Sub(Ptr|64|32|16|8)  x y) -> (SUB  x y)
(Sub32F x y) -> (FSUBS x y)
(Sub64F x y) -> (FSUB x y)

(Mod16 x y) -> (Mod32 (SignExt16to32 x) (SignExt16to32 y))
(Mod16u x y) -> (Mod32u (ZeroExt16to32 x) (ZeroExt16to32 y))
(Mod8 x y) -> (Mod32 (SignExt8to32 x) (SignExt8to32 y))
(Mod8u x y) -> (Mod32u (ZeroExt8to32 x) (ZeroExt8to32 y))
(Mod64 x y) -> (SUB x (MULLD y (DIVD x y)))
(Mod64u x y) -> (SUB x (MULLD y (DIVDU x y)))
(Mod32 x y) -> (SUB x (MULLW y (DIVW x y)))
(Mod32u x y) -> (SUB x (MULLW y (DIVWU x y)))

// (x + y) / 2 with x>=y -> (x - y) / 2 + y
(Avg64u <t> x y) -> (ADD (SRDconst <t> (SUB <t> x y) [1]) y)

(Mul64  x y) -> (MULLD  x y)
(Mul(32|16|8)  x y) -> (MULLW  x y)
(Mul64uhilo x y) -> (LoweredMuluhilo x y)

(Div64  x y) -> (DIVD  x y)
(Div64u x y) -> (DIVDU x y)
(Div32  x y) -> (DIVW  x y)
(Div32u x y) -> (DIVWU x y)
(Div16  x y) -> (DIVW  (SignExt16to32 x) (SignExt16to32 y))
(Div16u x y) -> (DIVWU (ZeroExt16to32 x) (ZeroExt16to32 y))
(Div8   x y) -> (DIVW  (SignExt8to32 x) (SignExt8to32 y))
(Div8u  x y) -> (DIVWU (ZeroExt8to32 x) (ZeroExt8to32 y))

(Hmul(64|64u|32|32u)  x y) -> (MULH(D|DU|W|WU)  x y)

(Mul32F x y) -> (FMULS x y)
(Mul64F x y) -> (FMUL x y)

(Div32F x y) -> (FDIVS x y)
(Div64F x y) -> (FDIV x y)

// Lowering float <-> int
(Cvt32to32F x) -> (FCFIDS (MTVSRD (SignExt32to64 x)))
(Cvt32to64F x) -> (FCFID (MTVSRD (SignExt32to64 x)))
(Cvt64to32F x) -> (FCFIDS (MTVSRD x))
(Cvt64to64F x) -> (FCFID (MTVSRD x))

(Cvt32Fto32 x) -> (MFVSRD (FCTIWZ x))
(Cvt32Fto64 x) -> (MFVSRD (FCTIDZ x))
(Cvt64Fto32 x) -> (MFVSRD (FCTIWZ x))
(Cvt64Fto64 x) -> (MFVSRD (FCTIDZ x))

(Cvt32Fto64F x) -> x // Note x will have the wrong type for patterns dependent on Float32/Float64
(Cvt64Fto32F x) -> (FRSP x)

(Round(32|64)F x) -> (LoweredRound(32|64)F x)

(Sqrt x) -> (FSQRT x)
(Floor x) -> (FFLOOR x)
(Ceil x) -> (FCEIL x)
(Trunc x) -> (FTRUNC x)
(Round x) -> (FROUND x)
(Copysign x y) -> (FCPSGN y x)
(Abs x) -> (FABS x)

// Lowering constants
(Const(64|32|16|8)  [val]) -> (MOVDconst [val])
(Const(32|64)F [val]) -> (FMOV(S|D)const [val])
(ConstNil) -> (MOVDconst [0])
(ConstBool [b]) -> (MOVDconst [b])

// Constant folding
(FABS (FMOVDconst [x])) -> (FMOVDconst [auxFrom64F(math.Abs(auxTo64F(x)))])
(FSQRT (FMOVDconst [x])) -> (FMOVDconst [auxFrom64F(math.Sqrt(auxTo64F(x)))])
(FFLOOR (FMOVDconst [x])) -> (FMOVDconst [auxFrom64F(math.Floor(auxTo64F(x)))])
(FCEIL (FMOVDconst [x])) -> (FMOVDconst [auxFrom64F(math.Ceil(auxTo64F(x)))])
(FTRUNC (FMOVDconst [x])) -> (FMOVDconst [auxFrom64F(math.Trunc(auxTo64F(x)))])

// Rotate generation with const shift
(ADD (SLDconst x [c]) (SRDconst x [d])) && d == 64-c -> (ROTLconst [c] x)
( OR (SLDconst x [c]) (SRDconst x [d])) && d == 64-c -> (ROTLconst [c] x)
(XOR (SLDconst x [c]) (SRDconst x [d])) && d == 64-c -> (ROTLconst [c] x)

(ADD (SLWconst x [c]) (SRWconst x [d])) && d == 32-c -> (ROTLWconst [c] x)
( OR (SLWconst x [c]) (SRWconst x [d])) && d == 32-c -> (ROTLWconst [c] x)
(XOR (SLWconst x [c]) (SRWconst x [d])) && d == 32-c -> (ROTLWconst [c] x)

// Rotate generation with non-const shift
// these match patterns from math/bits/RotateLeft[32|64], but there could be others
(ADD (SLD x (ANDconst <typ.Int64> [63] y)) (SRD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))) -> (ROTL x y)
( OR (SLD x (ANDconst <typ.Int64> [63] y)) (SRD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))) -> (ROTL x y)
(XOR (SLD x (ANDconst <typ.Int64> [63] y)) (SRD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))) -> (ROTL x y)

(ADD (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))) -> (ROTLW x y)
( OR (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))) -> (ROTLW x y)
(XOR (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))) -> (ROTLW x y)

(Lsh64x64  x (Const64 [c])) && uint64(c) < 64 -> (SLDconst x [c])
(Rsh64x64  x (Const64 [c])) && uint64(c) < 64 -> (SRADconst x [c])
(Rsh64Ux64 x (Const64 [c])) && uint64(c) < 64 -> (SRDconst x [c])
(Lsh32x64  x (Const64 [c])) && uint64(c) < 32 -> (SLWconst x [c])
(Rsh32x64  x (Const64 [c])) && uint64(c) < 32 -> (SRAWconst x [c])
(Rsh32Ux64 x (Const64 [c])) && uint64(c) < 32 -> (SRWconst x [c])
(Lsh16x64  x (Const64 [c])) && uint64(c) < 16 -> (SLWconst x [c])
(Rsh16x64  x (Const64 [c])) && uint64(c) < 16 -> (SRAWconst (SignExt16to32 x) [c])
(Rsh16Ux64 x (Const64 [c])) && uint64(c) < 16 -> (SRWconst (ZeroExt16to32 x) [c])
(Lsh8x64   x (Const64 [c])) && uint64(c) < 8  -> (SLWconst x [c])
(Rsh8x64   x (Const64 [c])) && uint64(c) < 8  -> (SRAWconst (SignExt8to32  x) [c])
(Rsh8Ux64  x (Const64 [c])) && uint64(c) < 8  -> (SRWconst (ZeroExt8to32  x) [c])

(Lsh64x32  x (Const64 [c])) && uint32(c) < 64 -> (SLDconst x [c])
(Rsh64x32  x (Const64 [c])) && uint32(c) < 64 -> (SRADconst x [c])
(Rsh64Ux32 x (Const64 [c])) && uint32(c) < 64 -> (SRDconst x [c])
(Lsh32x32  x (Const64 [c])) && uint32(c) < 32 -> (SLWconst x [c])
(Rsh32x32  x (Const64 [c])) && uint32(c) < 32 -> (SRAWconst x [c])
(Rsh32Ux32 x (Const64 [c])) && uint32(c) < 32 -> (SRWconst x [c])
(Lsh16x32  x (Const64 [c])) && uint32(c) < 16 -> (SLWconst x [c])
(Rsh16x32  x (Const64 [c])) && uint32(c) < 16 -> (SRAWconst (SignExt16to32 x) [c])
(Rsh16Ux32 x (Const64 [c])) && uint32(c) < 16 -> (SRWconst (ZeroExt16to32 x) [c])
(Lsh8x32   x (Const64 [c])) && uint32(c) < 8  -> (SLWconst x [c])
(Rsh8x32   x (Const64 [c])) && uint32(c) < 8  -> (SRAWconst (SignExt8to32  x) [c])
(Rsh8Ux32  x (Const64 [c])) && uint32(c) < 8  -> (SRWconst (ZeroExt8to32  x) [c])

// large constant shifts
(Lsh64x64  _ (Const64 [c])) && uint64(c) >= 64 -> (MOVDconst [0])
(Rsh64Ux64 _ (Const64 [c])) && uint64(c) >= 64 -> (MOVDconst [0])
(Lsh32x64  _ (Const64 [c])) && uint64(c) >= 32 -> (MOVDconst [0])
(Rsh32Ux64 _ (Const64 [c])) && uint64(c) >= 32 -> (MOVDconst [0])
(Lsh16x64  _ (Const64 [c])) && uint64(c) >= 16 -> (MOVDconst [0])
(Rsh16Ux64 _ (Const64 [c])) && uint64(c) >= 16 -> (MOVDconst [0])
(Lsh8x64   _ (Const64 [c])) && uint64(c) >= 8  -> (MOVDconst [0])
(Rsh8Ux64  _ (Const64 [c])) && uint64(c) >= 8  -> (MOVDconst [0])

// large constant signed right shift, we leave the sign bit
(Rsh64x64 x (Const64 [c])) && uint64(c) >= 64 -> (SRADconst x [63])
(Rsh32x64 x (Const64 [c])) && uint64(c) >= 32 -> (SRAWconst x [63])
(Rsh16x64 x (Const64 [c])) && uint64(c) >= 16 -> (SRAWconst (SignExt16to32 x) [63])
(Rsh8x64  x (Const64 [c])) && uint64(c) >= 8  -> (SRAWconst (SignExt8to32  x) [63])

// constant shifts
(Lsh64x64  x (MOVDconst [c])) && uint64(c) < 64 -> (SLDconst x [c])
(Rsh64x64  x (MOVDconst [c])) && uint64(c) < 64 -> (SRADconst x [c])
(Rsh64Ux64 x (MOVDconst [c])) && uint64(c) < 64 -> (SRDconst x [c])
(Lsh32x64  x (MOVDconst [c])) && uint64(c) < 32 -> (SLWconst x [c])
(Rsh32x64  x (MOVDconst [c])) && uint64(c) < 32 -> (SRAWconst x [c])
(Rsh32Ux64 x (MOVDconst [c])) && uint64(c) < 32 -> (SRWconst x [c])
(Lsh16x64  x (MOVDconst [c])) && uint64(c) < 16 -> (SLWconst x [c])
(Rsh16x64  x (MOVDconst [c])) && uint64(c) < 16 -> (SRAWconst (SignExt16to32 x) [c])
(Rsh16Ux64 x (MOVDconst [c])) && uint64(c) < 16 -> (SRWconst (ZeroExt16to32 x) [c])
(Lsh8x64   x (MOVDconst [c])) && uint64(c) < 8  -> (SLWconst x [c])
(Rsh8x64   x (MOVDconst [c])) && uint64(c) < 8  -> (SRAWconst (SignExt8to32  x) [c])
(Rsh8Ux64  x (MOVDconst [c])) && uint64(c) < 8  -> (SRWconst (ZeroExt8to32  x) [c])

(Lsh64x32  x (MOVDconst [c])) && uint32(c) < 64 -> (SLDconst x [c])
(Rsh64x32  x (MOVDconst [c])) && uint32(c) < 64 -> (SRADconst x [c])
(Rsh64Ux32 x (MOVDconst [c])) && uint32(c) < 64 -> (SRDconst x [c])
(Lsh32x32  x (MOVDconst [c])) && uint32(c) < 32 -> (SLWconst x [c])
(Rsh32x32  x (MOVDconst [c])) && uint32(c) < 32 -> (SRAWconst x [c])
(Rsh32Ux32 x (MOVDconst [c])) && uint32(c) < 32 -> (SRWconst x [c])
(Lsh16x32  x (MOVDconst [c])) && uint32(c) < 16 -> (SLWconst x [c])
(Rsh16x32  x (MOVDconst [c])) && uint32(c) < 16 -> (SRAWconst (SignExt16to32 x) [c])
(Rsh16Ux32 x (MOVDconst [c])) && uint32(c) < 16 -> (SRWconst (ZeroExt16to32 x) [c])
(Lsh8x32   x (MOVDconst [c])) && uint32(c) < 8  -> (SLWconst x [c])
(Rsh8x32   x (MOVDconst [c])) && uint32(c) < 8  -> (SRAWconst (SignExt8to32  x) [c])
(Rsh8Ux32  x (MOVDconst [c])) && uint32(c) < 8  -> (SRWconst (ZeroExt8to32  x) [c])

// Lower bounded shifts first. No need to check shift value.
(Lsh64x(64|32|16|8)  x y) && shiftIsBounded(v) -> (SLD x y)
(Lsh32x(64|32|16|8)  x y) && shiftIsBounded(v) -> (SLW x y)
(Lsh16x(64|32|16|8)  x y) && shiftIsBounded(v) -> (SLW x y)
(Lsh8x(64|32|16|8)   x y) && shiftIsBounded(v) -> (SLW x y)
(Rsh64Ux(64|32|16|8) x y) && shiftIsBounded(v) -> (SRD x y)
(Rsh32Ux(64|32|16|8) x y) && shiftIsBounded(v) -> (SRW x y)
(Rsh16Ux(64|32|16|8) x y) && shiftIsBounded(v) -> (SRW (MOVHZreg x) y)
(Rsh8Ux(64|32|16|8)  x y) && shiftIsBounded(v) -> (SRW (MOVBZreg x) y)
(Rsh64x(64|32|16|8)  x y) && shiftIsBounded(v) -> (SRAD x y)
(Rsh32x(64|32|16|8)  x y) && shiftIsBounded(v) -> (SRAW x y)
(Rsh16x(64|32|16|8)  x y) && shiftIsBounded(v) -> (SRAW (MOVHreg x) y)
(Rsh8x(64|32|16|8)   x y) && shiftIsBounded(v) -> (SRAW (MOVBreg x) y)

// non-constant rotates
// These are subexpressions found in statements that can become rotates
// In these cases the shift count is known to be < 64 so the more complicated expressions
// with Mask & Carry is not needed
(Lsh64x64 x (AND y (MOVDconst [63]))) -> (SLD x (ANDconst <typ.Int64> [63] y))
(Lsh64x64 x (ANDconst <typ.Int64> [63] y)) -> (SLD x (ANDconst <typ.Int64> [63] y))
(Rsh64Ux64 x (AND y (MOVDconst [63]))) -> (SRD x (ANDconst <typ.Int64> [63] y))
(Rsh64Ux64 x (ANDconst <typ.UInt> [63] y)) -> (SRD x (ANDconst <typ.UInt> [63] y))
(Rsh64Ux64 x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y))) -> (SRD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))
(Rsh64Ux64 x (SUB <typ.UInt> (MOVDconst [64]) (AND <typ.UInt> y (MOVDconst [63])))) -> (SRD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))
(Rsh64x64 x (AND y (MOVDconst [63]))) -> (SRAD x (ANDconst <typ.Int64> [63] y))
(Rsh64x64 x (ANDconst <typ.UInt> [63] y)) -> (SRAD x (ANDconst <typ.UInt> [63] y))
(Rsh64x64 x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y))) -> (SRAD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))
(Rsh64x64 x (SUB <typ.UInt> (MOVDconst [64]) (AND <typ.UInt> y (MOVDconst [63])))) -> (SRAD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))

(Rsh64x64 x y)  -> (SRAD x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] y))))
(Rsh64Ux64 x y) -> (SRD  x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] y))))
(Lsh64x64 x y)  -> (SLD  x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] y))))

(Lsh32x64 x (AND y (MOVDconst [31]))) -> (SLW x (ANDconst <typ.Int32> [31] y))
(Lsh32x64 x (ANDconst <typ.Int32> [31] y)) -> (SLW x (ANDconst <typ.Int32> [31] y))

(Rsh32Ux64 x (AND y (MOVDconst [31]))) -> (SRW x (ANDconst <typ.Int32> [31] y))
(Rsh32Ux64 x (ANDconst <typ.UInt> [31] y)) -> (SRW x (ANDconst <typ.UInt> [31] y))
(Rsh32Ux64 x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y))) -> (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))
(Rsh32Ux64 x (SUB <typ.UInt> (MOVDconst [32]) (AND <typ.UInt> y (MOVDconst [31])))) -> (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))

(Rsh32x64 x (AND y (MOVDconst [31]))) -> (SRAW x (ANDconst <typ.Int32> [31] y))
(Rsh32x64 x (ANDconst <typ.UInt> [31] y)) -> (SRAW x (ANDconst <typ.UInt> [31] y))
(Rsh32x64 x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y))) -> (SRAW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))
(Rsh32x64 x (SUB <typ.UInt> (MOVDconst [32]) (AND <typ.UInt> y (MOVDconst [31])))) -> (SRAW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))

(Rsh32x64 x y)  -> (SRAW x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] y))))
(Rsh32Ux64 x y) -> (SRW  x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] y))))
(Lsh32x64 x y)  -> (SLW  x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] y))))

(Rsh16x64 x y)  -> (SRAW (SignExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] y))))
(Rsh16Ux64 x y) -> (SRW  (ZeroExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] y))))
(Lsh16x64 x y)  -> (SLW  x                 (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] y))))

(Rsh8x64 x y)  -> (SRAW (SignExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] y))))
(Rsh8Ux64 x y) -> (SRW  (ZeroExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] y))))
(Lsh8x64 x y)  -> (SLW  x                (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] y))))

(Rsh64x32 x y)  -> (SRAD x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt32to64 y)))))
(Rsh64Ux32 x y) -> (SRD x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt32to64 y)))))
(Lsh64x32 x y)  -> (SLD x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt32to64 y)))))

(Rsh32x32 x y)  -> (SRAW x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt32to64 y)))))
(Rsh32Ux32 x y) -> (SRW x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt32to64 y)))))
(Lsh32x32 x y)  -> (SLW x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt32to64 y)))))

(Rsh16x32 x y)  -> (SRAW (SignExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt32to64 y)))))
(Rsh16Ux32 x y) -> (SRW  (ZeroExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt32to64 y)))))
(Lsh16x32 x y)  -> (SLW  x                 (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt32to64 y)))))

(Rsh8x32 x y)  -> (SRAW (SignExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt32to64 y)))))
(Rsh8Ux32 x y) -> (SRW  (ZeroExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt32to64 y)))))
(Lsh8x32 x y)  -> (SLW  x                (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt32to64 y)))))


(Rsh64x16 x y)  -> (SRAD x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt16to64 y)))))
(Rsh64Ux16 x y) -> (SRD x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt16to64 y)))))
(Lsh64x16 x y)  -> (SLD x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt16to64 y)))))

(Rsh32x16 x y)  -> (SRAW x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt16to64 y)))))
(Rsh32Ux16 x y) -> (SRW x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt16to64 y)))))
(Lsh32x16 x y)  -> (SLW x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt16to64 y)))))

(Rsh16x16 x y)  -> (SRAW (SignExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt16to64 y)))))
(Rsh16Ux16 x y) -> (SRW  (ZeroExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt16to64 y)))))
(Lsh16x16 x y)  -> (SLW  x                 (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt16to64 y)))))

(Rsh8x16 x y)  -> (SRAW (SignExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt16to64 y)))))
(Rsh8Ux16 x y) -> (SRW  (ZeroExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt16to64 y)))))
(Lsh8x16 x y)  -> (SLW  x                (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt16to64 y)))))


(Rsh64x8 x y)  -> (SRAD x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt8to64 y)))))
(Rsh64Ux8 x y) -> (SRD x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt8to64 y)))))
(Lsh64x8 x y)  -> (SLD x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-64] (ZeroExt8to64 y)))))

(Rsh32x8 x y)  -> (SRAW x (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt8to64 y)))))
(Rsh32Ux8 x y) -> (SRW x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt8to64 y)))))
(Lsh32x8 x y)  -> (SLW x  (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-32] (ZeroExt8to64 y)))))

(Rsh16x8 x y)  -> (SRAW (SignExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt8to64 y)))))
(Rsh16Ux8 x y) -> (SRW  (ZeroExt16to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt8to64 y)))))
(Lsh16x8 x y)  -> (SLW  x                 (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-16] (ZeroExt8to64 y)))))

(Rsh8x8 x y)  -> (SRAW (SignExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt8to64 y)))))
(Rsh8Ux8 x y) -> (SRW  (ZeroExt8to32 x) (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt8to64 y)))))
(Lsh8x8 x y)  -> (SLW  x                (ORN y <typ.Int64> (MaskIfNotCarry (ADDconstForCarry [-8] (ZeroExt8to64 y)))))

// Cleaning up shift ops when input is masked
(MaskIfNotCarry (ADDconstForCarry [c] (ANDconst [d] _))) && c < 0 && d > 0 && c + d < 0 -> (MOVDconst [-1])
(ORN x (MOVDconst [-1])) -> x

// Potentially useful optimizing rewrites.
// (ADDconstForCarry [k] c), k < 0 && (c < 0 || k+c >= 0) -> CarrySet
// (ADDconstForCarry [k] c), K < 0 && (c >= 0 && k+c < 0) -> CarryClear
// (MaskIfNotCarry CarrySet) -> 0
// (MaskIfNotCarry CarrySet) -> -1

(Addr {sym} base) -> (MOVDaddr {sym} base)
(LocalAddr {sym} base _) -> (MOVDaddr {sym} base)
(OffPtr [off] ptr) -> (ADD (MOVDconst <typ.Int64> [off]) ptr)

// TODO: optimize these cases?
(Ctz32NonZero x) -> (Ctz32 x)
(Ctz64NonZero x) -> (Ctz64 x)

(Ctz64 x) -> (POPCNTD (ANDN <typ.Int64> (ADDconst <typ.Int64> [-1] x) x))
(Ctz32 x) -> (POPCNTW (MOVWZreg (ANDN <typ.Int> (ADDconst <typ.Int> [-1] x) x)))
(Ctz16 x) -> (POPCNTW (MOVHZreg (ANDN <typ.Int16> (ADDconst <typ.Int16> [-1] x) x)))
(Ctz8 x) -> (POPCNTB (MOVBZreg (ANDN <typ.UInt8> (ADDconst <typ.UInt8> [-1] x) x)))

(BitLen64 x) -> (SUB (MOVDconst [64]) (CNTLZD <typ.Int> x))
(BitLen32 x) -> (SUB (MOVDconst [32]) (CNTLZW <typ.Int> x))

(PopCount64 x) -> (POPCNTD x)
(PopCount32 x) -> (POPCNTW (MOVWZreg x))
(PopCount16 x) -> (POPCNTW (MOVHZreg x))
(PopCount8 x) -> (POPCNTB (MOVBZreg x))

(And(64|32|16|8) x y) -> (AND x y)
(Or(64|32|16|8) x y) -> (OR x y)
(Xor(64|32|16|8) x y) -> (XOR x y)

(Neg(64|32|16|8)  x) -> (NEG x)
(Neg64F x) -> (FNEG x)
(Neg32F x) -> (FNEG x)

(Com(64|32|16|8) x) -> (NOR x x)

// Lowering boolean ops
(AndB x y) -> (AND x y)
(OrB x y) -> (OR x y)
(Not x) -> (XORconst [1] x)

// Use ANDN for AND x NOT y
(AND x (NOR y y)) -> (ANDN x y)

// Lowering comparisons
(EqB x y)  -> (ANDconst [1] (EQV x y))
// Sign extension dependence on operand sign sets up for sign/zero-extension elision later
(Eq8 x y) && isSigned(x.Type) && isSigned(y.Type) -> (Equal (CMPW (SignExt8to32 x) (SignExt8to32 y)))
(Eq16 x y) && isSigned(x.Type) && isSigned(y.Type) -> (Equal (CMPW (SignExt16to32 x) (SignExt16to32 y)))
(Eq8 x y)  -> (Equal (CMPW (ZeroExt8to32 x) (ZeroExt8to32 y)))
(Eq16 x y) -> (Equal (CMPW (ZeroExt16to32 x) (ZeroExt16to32 y)))
(Eq32 x y) -> (Equal (CMPW x y))
(Eq64 x y) -> (Equal (CMP x y))
(Eq32F x y) -> (Equal (FCMPU x y))
(Eq64F x y) -> (Equal (FCMPU x y))
(EqPtr x y) -> (Equal (CMP x y))

(NeqB x y)  -> (XOR x y)
// Like Eq8 and Eq16, prefer sign extension likely to enable later elision.
(Neq8 x y) && isSigned(x.Type) && isSigned(y.Type) -> (NotEqual (CMPW (SignExt8to32 x) (SignExt8to32 y)))
(Neq16 x y) && isSigned(x.Type) && isSigned(y.Type) -> (NotEqual (CMPW (SignExt16to32 x) (SignExt16to32 y)))
(Neq8 x y)  -> (NotEqual (CMPW (ZeroExt8to32 x) (ZeroExt8to32 y)))
(Neq16 x y) -> (NotEqual (CMPW (ZeroExt16to32 x) (ZeroExt16to32 y)))
(Neq32 x y) -> (NotEqual (CMPW x y))
(Neq64 x y) -> (NotEqual (CMP x y))
(Neq32F x y) -> (NotEqual (FCMPU x y))
(Neq64F x y) -> (NotEqual (FCMPU x y))
(NeqPtr x y) -> (NotEqual (CMP x y))

(Less8 x y)  -> (LessThan (CMPW (SignExt8to32 x) (SignExt8to32 y)))
(Less16 x y) -> (LessThan (CMPW (SignExt16to32 x) (SignExt16to32 y)))
(Less32 x y) -> (LessThan (CMPW x y))
(Less64 x y) -> (LessThan (CMP x y))
(Less32F x y) -> (FLessThan (FCMPU x y))
(Less64F x y) -> (FLessThan (FCMPU x y))

(Less8U x y)  -> (LessThan (CMPWU (ZeroExt8to32 x) (ZeroExt8to32 y)))
(Less16U x y) -> (LessThan (CMPWU (ZeroExt16to32 x) (ZeroExt16to32 y)))
(Less32U x y) -> (LessThan (CMPWU x y))
(Less64U x y) -> (LessThan (CMPU x y))

(Leq8 x y)  -> (LessEqual (CMPW (SignExt8to32 x) (SignExt8to32 y)))
(Leq16 x y) -> (LessEqual (CMPW (SignExt16to32 x) (SignExt16to32 y)))
(Leq32 x y) -> (LessEqual (CMPW x y))
(Leq64 x y) -> (LessEqual (CMP x y))
(Leq32F x y) -> (FLessEqual (FCMPU x y))
(Leq64F x y) -> (FLessEqual (FCMPU x y))

(Leq8U x y)  -> (LessEqual (CMPWU (ZeroExt8to32 x) (ZeroExt8to32 y)))
(Leq16U x y) -> (LessEqual (CMPWU (ZeroExt16to32 x) (ZeroExt16to32 y)))
(Leq32U x y) -> (LessEqual (CMPWU x y))
(Leq64U x y) -> (LessEqual (CMPU x y))

(Greater8 x y)  -> (GreaterThan (CMPW (SignExt8to32 x) (SignExt8to32 y)))
(Greater16 x y) -> (GreaterThan (CMPW (SignExt16to32 x) (SignExt16to32 y)))
(Greater32 x y) -> (GreaterThan (CMPW x y))
(Greater64 x y) -> (GreaterThan (CMP x y))
(Greater(32|64)F x y) -> (FGreaterThan (FCMPU x y))

(Greater8U x y)  -> (GreaterThan (CMPWU (ZeroExt8to32 x) (ZeroExt8to32 y)))
(Greater16U x y) -> (GreaterThan (CMPWU (ZeroExt16to32 x) (ZeroExt16to32 y)))
(Greater32U x y) -> (GreaterThan (CMPWU x y))
(Greater64U x y) -> (GreaterThan (CMPU x y))

(Geq8 x y)  -> (GreaterEqual (CMPW (SignExt8to32 x) (SignExt8to32 y)))
(Geq16 x y) -> (GreaterEqual (CMPW (SignExt16to32 x) (SignExt16to32 y)))
(Geq32 x y) -> (GreaterEqual (CMPW x y))
(Geq64 x y) -> (GreaterEqual (CMP x y))
(Geq(32|64)F x y) -> (FGreaterEqual (FCMPU x y))

(Geq8U x y)  -> (GreaterEqual (CMPWU (ZeroExt8to32 x) (ZeroExt8to32 y)))
(Geq16U x y) -> (GreaterEqual (CMPWU (ZeroExt16to32 x) (ZeroExt16to32 y)))
(Geq32U x y) -> (GreaterEqual (CMPWU x y))
(Geq64U x y) -> (GreaterEqual (CMPU x y))

// Absorb pseudo-ops into blocks.
(If (Equal cc) yes no) -> (EQ cc yes no)
(If (NotEqual cc) yes no) -> (NE cc yes no)
(If (LessThan cc) yes no) -> (LT cc yes no)
(If (LessEqual cc) yes no) -> (LE cc yes no)
(If (GreaterThan cc) yes no) -> (GT cc yes no)
(If (GreaterEqual cc) yes no) -> (GE cc yes no)
(If (FLessThan cc) yes no) -> (FLT cc yes no)
(If (FLessEqual cc) yes no) -> (FLE cc yes no)
(If (FGreaterThan cc) yes no) -> (FGT cc yes no)
(If (FGreaterEqual cc) yes no) -> (FGE cc yes no)

(If cond yes no) -> (NE (CMPWconst [0] cond) yes no)

// Absorb boolean tests into block
(NE (CMPWconst [0] (Equal cc)) yes no) -> (EQ cc yes no)
(NE (CMPWconst [0] (NotEqual cc)) yes no) -> (NE cc yes no)
(NE (CMPWconst [0] (LessThan cc)) yes no) -> (LT cc yes no)
(NE (CMPWconst [0] (LessEqual cc)) yes no) -> (LE cc yes no)
(NE (CMPWconst [0] (GreaterThan cc)) yes no) -> (GT cc yes no)
(NE (CMPWconst [0] (GreaterEqual cc)) yes no) -> (GE cc yes no)
(NE (CMPWconst [0] (FLessThan cc)) yes no) -> (FLT cc yes no)
(NE (CMPWconst [0] (FLessEqual cc)) yes no) -> (FLE cc yes no)
(NE (CMPWconst [0] (FGreaterThan cc)) yes no) -> (FGT cc yes no)
(NE (CMPWconst [0] (FGreaterEqual cc)) yes no) -> (FGE cc yes no)

// Elide compares of bit tests // TODO need to make both CC and result of ANDCC available.
(EQ (CMPconst [0] (ANDconst [c] x)) yes no) -> (EQ (ANDCCconst [c] x) yes no)
(NE (CMPconst [0] (ANDconst [c] x)) yes no) -> (NE (ANDCCconst [c] x) yes no)
(EQ (CMPWconst [0] (ANDconst [c] x)) yes no) -> (EQ (ANDCCconst [c] x) yes no)
(NE (CMPWconst [0] (ANDconst [c] x)) yes no) -> (NE (ANDCCconst [c] x) yes no)

// absorb flag constants into branches
(EQ (FlagEQ) yes no) -> (First nil yes no)
(EQ (FlagLT) yes no) -> (First nil no yes)
(EQ (FlagGT) yes no) -> (First nil no yes)

(NE (FlagEQ) yes no) -> (First nil no yes)
(NE (FlagLT) yes no) -> (First nil yes no)
(NE (FlagGT) yes no) -> (First nil yes no)

(LT (FlagEQ) yes no) -> (First nil no yes)
(LT (FlagLT) yes no) -> (First nil yes no)
(LT (FlagGT) yes no) -> (First nil no yes)

(LE (FlagEQ) yes no) -> (First nil yes no)
(LE (FlagLT) yes no) -> (First nil yes no)
(LE (FlagGT) yes no) -> (First nil no yes)

(GT (FlagEQ) yes no) -> (First nil no yes)
(GT (FlagLT) yes no) -> (First nil no yes)
(GT (FlagGT) yes no) -> (First nil yes no)

(GE (FlagEQ) yes no) -> (First nil yes no)
(GE (FlagLT) yes no) -> (First nil no yes)
(GE (FlagGT) yes no) -> (First nil yes no)

// absorb InvertFlags into branches
(LT (InvertFlags cmp) yes no) -> (GT cmp yes no)
(GT (InvertFlags cmp) yes no) -> (LT cmp yes no)
(LE (InvertFlags cmp) yes no) -> (GE cmp yes no)
(GE (InvertFlags cmp) yes no) -> (LE cmp yes no)
(EQ (InvertFlags cmp) yes no) -> (EQ cmp yes no)
(NE (InvertFlags cmp) yes no) -> (NE cmp yes no)

// constant comparisons
(CMPWconst (MOVDconst [x]) [y]) && int32(x)==int32(y) -> (FlagEQ)
(CMPWconst (MOVDconst [x]) [y]) && int32(x)<int32(y)  -> (FlagLT)
(CMPWconst (MOVDconst [x]) [y]) && int32(x)>int32(y)  -> (FlagGT)

(CMPconst (MOVDconst [x]) [y]) && x==y -> (FlagEQ)
(CMPconst (MOVDconst [x]) [y]) && x<y  -> (FlagLT)
(CMPconst (MOVDconst [x]) [y]) && x>y  -> (FlagGT)

(CMPWUconst (MOVDconst [x]) [y]) && int32(x)==int32(y)  -> (FlagEQ)
(CMPWUconst (MOVDconst [x]) [y]) && uint32(x)<uint32(y) -> (FlagLT)
(CMPWUconst (MOVDconst [x]) [y]) && uint32(x)>uint32(y) -> (FlagGT)

(CMPUconst (MOVDconst [x]) [y]) && x==y  -> (FlagEQ)
(CMPUconst (MOVDconst [x]) [y]) && uint64(x)<uint64(y) -> (FlagLT)
(CMPUconst (MOVDconst [x]) [y]) && uint64(x)>uint64(y) -> (FlagGT)

// other known comparisons
//(CMPconst (MOVBUreg _) [c]) && 0xff < c -> (FlagLT)
//(CMPconst (MOVHUreg _) [c]) && 0xffff < c -> (FlagLT)
//(CMPconst (ANDconst _ [m]) [n]) && 0 <= int32(m) && int32(m) < int32(n) -> (FlagLT)
//(CMPconst (SRLconst _ [c]) [n]) && 0 <= n && 0 < c && c <= 32 && (1<<uint32(32-c)) <= uint32(n) -> (FlagLT)

// absorb flag constants into boolean values
(Equal (FlagEQ)) -> (MOVDconst [1])
(Equal (FlagLT)) -> (MOVDconst [0])
(Equal (FlagGT)) -> (MOVDconst [0])

(NotEqual (FlagEQ)) -> (MOVDconst [0])
(NotEqual (FlagLT)) -> (MOVDconst [1])
(NotEqual (FlagGT)) -> (MOVDconst [1])

(LessThan (FlagEQ)) -> (MOVDconst [0])
(LessThan (FlagLT)) -> (MOVDconst [1])
(LessThan (FlagGT)) -> (MOVDconst [0])

(LessEqual (FlagEQ)) -> (MOVDconst [1])
(LessEqual (FlagLT)) -> (MOVDconst [1])
(LessEqual (FlagGT)) -> (MOVDconst [0])

(GreaterThan (FlagEQ)) -> (MOVDconst [0])
(GreaterThan (FlagLT)) -> (MOVDconst [0])
(GreaterThan (FlagGT)) -> (MOVDconst [1])

(GreaterEqual (FlagEQ)) -> (MOVDconst [1])
(GreaterEqual (FlagLT)) -> (MOVDconst [0])
(GreaterEqual (FlagGT)) -> (MOVDconst [1])

// absorb InvertFlags into boolean values
(Equal (InvertFlags x)) -> (Equal x)
(NotEqual (InvertFlags x)) -> (NotEqual x)
(LessThan (InvertFlags x)) -> (GreaterThan x)
(GreaterThan (InvertFlags x)) -> (LessThan x)
(LessEqual (InvertFlags x)) -> (GreaterEqual x)
(GreaterEqual (InvertFlags x)) -> (LessEqual x)

// Elide compares of bit tests // TODO need to make both CC and result of ANDCC available.
((EQ|NE|LT|LE|GT|GE) (CMPconst [0] (ANDconst [c] x)) yes no) -> ((EQ|NE|LT|LE|GT|GE) (ANDCCconst [c] x) yes no)
((EQ|NE|LT|LE|GT|GE) (CMPWconst [0] (ANDconst [c] x)) yes no) -> ((EQ|NE|LT|LE|GT|GE) (ANDCCconst [c] x) yes no)
((EQ|NE|LT|LE|GT|GE) (CMPconst [0] z:(AND x y)) yes no) && z.Uses == 1 -> ((EQ|NE|LT|LE|GT|GE) (ANDCC x y) yes no)
((EQ|NE|LT|LE|GT|GE) (CMPconst [0] z:(OR x y)) yes no) && z.Uses == 1 -> ((EQ|NE|LT|LE|GT|GE) (ORCC x y) yes no)
((EQ|NE|LT|LE|GT|GE) (CMPconst [0] z:(XOR x y)) yes no) && z.Uses == 1 -> ((EQ|NE|LT|LE|GT|GE) (XORCC x y) yes no)

// Lowering loads
(Load <t> ptr mem) && (is64BitInt(t) || isPtr(t)) -> (MOVDload ptr mem)
(Load <t> ptr mem) && is32BitInt(t) && isSigned(t) -> (MOVWload ptr mem)
(Load <t> ptr mem) && is32BitInt(t) && !isSigned(t) -> (MOVWZload ptr mem)
(Load <t> ptr mem) && is16BitInt(t) && isSigned(t) -> (MOVHload ptr mem)
(Load <t> ptr mem) && is16BitInt(t) && !isSigned(t) -> (MOVHZload ptr mem)
(Load <t> ptr mem) && t.IsBoolean() -> (MOVBZload ptr mem)
(Load <t> ptr mem) && is8BitInt(t) && isSigned(t) -> (MOVBreg (MOVBZload ptr mem)) // PPC has no signed-byte load.
(Load <t> ptr mem) && is8BitInt(t) && !isSigned(t) -> (MOVBZload ptr mem)

(Load <t> ptr mem) && is32BitFloat(t) -> (FMOVSload ptr mem)
(Load <t> ptr mem) && is64BitFloat(t) -> (FMOVDload ptr mem)

(Store {t} ptr val mem) && t.(*types.Type).Size() == 8 && is64BitFloat(val.Type) -> (FMOVDstore ptr val mem)
(Store {t} ptr val mem) && t.(*types.Type).Size() == 8 && is32BitFloat(val.Type) -> (FMOVDstore ptr val mem) // glitch from (Cvt32Fto64F x) -> x -- type is wrong
(Store {t} ptr val mem) && t.(*types.Type).Size() == 4 && is32BitFloat(val.Type) -> (FMOVSstore ptr val mem)
(Store {t} ptr val mem) && t.(*types.Type).Size() == 8 && (is64BitInt(val.Type) || isPtr(val.Type)) -> (MOVDstore ptr val mem)
(Store {t} ptr val mem) && t.(*types.Type).Size() == 4 && is32BitInt(val.Type) -> (MOVWstore ptr val mem)
(Store {t} ptr val mem) && t.(*types.Type).Size() == 2 -> (MOVHstore ptr val mem)
(Store {t} ptr val mem) && t.(*types.Type).Size() == 1 -> (MOVBstore ptr val mem)

// Using Zero instead of LoweredZero allows the
// target address to be folded where possible.
(Zero [0] _ mem) -> mem
(Zero [1] destptr mem) -> (MOVBstorezero destptr mem)
(Zero [2] destptr mem) ->
	(MOVHstorezero destptr mem)
(Zero [3] destptr mem) ->
	(MOVBstorezero [2] destptr
		(MOVHstorezero destptr mem))
(Zero [4] destptr mem) ->
	(MOVWstorezero destptr mem)
(Zero [5] destptr mem) ->
	(MOVBstorezero [4] destptr
        	(MOVWstorezero destptr mem))
(Zero [6] destptr mem) ->
	(MOVHstorezero [4] destptr
		(MOVWstorezero destptr mem))
(Zero [7] destptr mem) ->
	(MOVBstorezero [6] destptr
		(MOVHstorezero [4] destptr
			(MOVWstorezero destptr mem)))

// MOVD for store with DS must have offsets that are multiple of 4
(Zero [8] {t} destptr mem) && t.(*types.Type).Alignment()%4 == 0 ->
        (MOVDstorezero destptr mem)
(Zero [8] destptr mem) ->
        (MOVWstorezero [4] destptr
                (MOVWstorezero [0] destptr mem))
// Handle these cases only if aligned properly, otherwise use general case below
(Zero [12] {t} destptr mem) && t.(*types.Type).Alignment()%4 == 0 ->
        (MOVWstorezero [8] destptr
                (MOVDstorezero [0] destptr mem))
(Zero [16] {t} destptr mem) && t.(*types.Type).Alignment()%4 == 0 ->
       (MOVDstorezero [8] destptr
                (MOVDstorezero [0] destptr mem))
(Zero [24] {t} destptr mem) && t.(*types.Type).Alignment()%4 == 0 ->
       (MOVDstorezero [16] destptr
               (MOVDstorezero [8] destptr
                       (MOVDstorezero [0] destptr mem)))
(Zero [32] {t} destptr mem) && t.(*types.Type).Alignment()%4 == 0 ->
       (MOVDstorezero [24] destptr
               (MOVDstorezero [16] destptr
                       (MOVDstorezero [8] destptr
                               (MOVDstorezero [0] destptr mem))))

// Handle cases not handled above
(Zero [s] ptr mem) -> (LoweredZero [s] ptr mem)

// moves
// Only the MOVD and MOVW instructions require 4 byte
// alignment in the offset field.  The other MOVx instructions
// allow any alignment.
(Move [0] _ _ mem) -> mem
(Move [1] dst src mem) -> (MOVBstore dst (MOVBZload src mem) mem)
(Move [2] dst src mem) ->
        (MOVHstore dst (MOVHZload src mem) mem)
(Move [4] dst src mem) ->
	(MOVWstore dst (MOVWZload src mem) mem)
// MOVD for load and store must have offsets that are multiple of 4
(Move [8] {t} dst src mem) && t.(*types.Type).Alignment()%4 == 0 ->
	(MOVDstore dst (MOVDload src mem) mem)
(Move [8] dst src mem) ->
	(MOVWstore [4] dst (MOVWZload [4] src mem)
		(MOVWstore dst (MOVWZload src mem) mem))
(Move [3] dst src mem) ->
        (MOVBstore [2] dst (MOVBZload [2] src mem)
                (MOVHstore dst (MOVHload src mem) mem))
(Move [5] dst src mem) ->
        (MOVBstore [4] dst (MOVBZload [4] src mem)
                (MOVWstore dst (MOVWZload src mem) mem))
(Move [6] dst src mem) ->
        (MOVHstore [4] dst (MOVHZload [4] src mem)
                (MOVWstore dst (MOVWZload src mem) mem))
(Move [7] dst src mem) ->
        (MOVBstore [6] dst (MOVBZload [6] src mem)
                (MOVHstore [4] dst (MOVHZload [4] src mem)
                        (MOVWstore dst (MOVWZload src mem) mem)))

// Large move uses a loop. Since the address is computed and the
// offset is zero, any alignment can be used.
(Move [s] dst src mem) && s > 8 ->
        (LoweredMove [s] dst src mem)

// Calls
// Lowering calls
(StaticCall [argwid] {target} mem) -> (CALLstatic [argwid] {target} mem)
(ClosureCall [argwid] entry closure mem) -> (CALLclosure [argwid] entry closure mem)
(InterCall [argwid] entry mem) -> (CALLinter [argwid] entry mem)

// Miscellaneous
(GetClosurePtr) -> (LoweredGetClosurePtr)
(GetCallerSP) -> (LoweredGetCallerSP)
(GetCallerPC) -> (LoweredGetCallerPC)
(IsNonNil ptr) -> (NotEqual (CMPconst [0] ptr))
(IsInBounds idx len) -> (LessThan (CMPU idx len))
(IsSliceInBounds idx len) -> (LessEqual (CMPU idx len))
(NilCheck ptr mem) -> (LoweredNilCheck ptr mem)

// Write barrier.
(WB {fn} destptr srcptr mem) -> (LoweredWB {fn} destptr srcptr mem)

// Optimizations
// Note that PPC "logical" immediates come in 0:15 and 16:31 unsigned immediate forms,
// so ORconst, XORconst easily expand into a pair.

// Include very-large constants in the const-const case.
(AND (MOVDconst [c]) (MOVDconst [d])) -> (MOVDconst [c&d])
(OR (MOVDconst [c]) (MOVDconst [d])) -> (MOVDconst [c|d])
(XOR (MOVDconst [c]) (MOVDconst [d])) -> (MOVDconst [c^d])

// Discover consts
(AND x (MOVDconst [c])) && isU16Bit(c) -> (ANDconst [c] x)
(XOR x (MOVDconst [c])) && isU32Bit(c) -> (XORconst [c] x)
(OR x (MOVDconst [c])) && isU32Bit(c) -> (ORconst [c] x)

// Simplify consts
(ANDconst [c] (ANDconst [d] x)) -> (ANDconst [c&d] x)
(ORconst [c] (ORconst [d] x)) -> (ORconst [c|d] x)
(XORconst [c] (XORconst [d] x)) -> (XORconst [c^d] x)
(ANDconst [-1] x) -> x
(ANDconst [0] _) -> (MOVDconst [0])
(XORconst [0] x) -> x
(ORconst [-1] _) -> (MOVDconst [-1])
(ORconst [0] x) -> x

// zero-extend of small and -> small and
(MOVBZreg y:(ANDconst [c] _)) && uint64(c) <= 0xFF -> y
(MOVHZreg y:(ANDconst [c] _)) && uint64(c) <= 0xFFFF -> y
(MOVWZreg y:(ANDconst [c] _)) && uint64(c) <= 0xFFFFFFFF -> y
(MOVWZreg y:(AND (MOVDconst [c]) _)) && uint64(c) <= 0xFFFFFFFF -> y

// sign extend of small-positive and -> small-positive-and
(MOVBreg y:(ANDconst [c] _)) && uint64(c) <= 0x7F -> y
(MOVHreg y:(ANDconst [c] _)) && uint64(c) <= 0x7FFF -> y
(MOVWreg y:(ANDconst [c] _)) && uint64(c) <= 0xFFFF -> y // 0xFFFF is largest immediate constant, when regarded as 32-bit is > 0
(MOVWreg y:(AND (MOVDconst [c]) _)) && uint64(c) <= 0x7FFFFFFF -> y

// small and of zero-extend -> either zero-extend or small and
(ANDconst [c] y:(MOVBZreg _)) && c&0xFF == 0xFF -> y
(ANDconst [0xFF] y:(MOVBreg _)) -> y
(ANDconst [c] y:(MOVHZreg _))  && c&0xFFFF == 0xFFFF -> y
(ANDconst [0xFFFF] y:(MOVHreg _)) -> y

(AND (MOVDconst [c]) y:(MOVWZreg _))  && c&0xFFFFFFFF == 0xFFFFFFFF -> y
(AND (MOVDconst [0xFFFFFFFF]) y:(MOVWreg x)) -> (MOVWZreg x)
// normal case
(ANDconst [c] (MOV(B|BZ)reg x)) -> (ANDconst [c&0xFF] x)
(ANDconst [c] (MOV(H|HZ)reg x)) -> (ANDconst [c&0xFFFF] x)
(ANDconst [c] (MOV(W|WZ)reg x)) -> (ANDconst [c&0xFFFFFFFF] x)

// Eliminate unnecessary sign/zero extend following right shift
(MOV(B|H|W)Zreg (SRWconst [c] (MOVBZreg x))) -> (SRWconst [c] (MOVBZreg x))
(MOV(H|W)Zreg (SRWconst [c] (MOVHZreg x))) -> (SRWconst [c] (MOVHZreg x))
(MOVWZreg (SRWconst [c] (MOVWZreg x))) -> (SRWconst [c] (MOVWZreg x))
(MOV(B|H|W)reg (SRAWconst [c] (MOVBreg x))) -> (SRAWconst [c] (MOVBreg x))
(MOV(H|W)reg (SRAWconst [c] (MOVHreg x))) -> (SRAWconst [c] (MOVHreg x))
(MOVWreg (SRAWconst [c] (MOVWreg x))) -> (SRAWconst [c] (MOVWreg x))

(MOVWZreg (SRWconst [c] x)) && sizeof(x.Type) <= 32 -> (SRWconst [c] x)
(MOVHZreg (SRWconst [c] x)) && sizeof(x.Type) <= 16 -> (SRWconst [c] x)
(MOVBZreg (SRWconst [c] x)) && sizeof(x.Type) == 8 -> (SRWconst [c] x)
(MOVWreg (SRAWconst [c] x)) && sizeof(x.Type) <= 32 -> (SRAWconst [c] x)
(MOVHreg (SRAWconst [c] x)) && sizeof(x.Type) <= 16 -> (SRAWconst [c] x)
(MOVBreg (SRAWconst [c] x)) && sizeof(x.Type) == 8 -> (SRAWconst [c] x)

// initial right shift will handle sign/zero extend
(MOVBZreg (SRDconst [c] x)) && c>=56 -> (SRDconst [c] x)
(MOVBreg (SRDconst [c] x)) && c>56 -> (SRDconst [c] x)
(MOVBreg (SRDconst [c] x)) && c==56 -> (SRADconst [c] x)
(MOVBZreg (SRWconst [c] x)) && c>=24 -> (SRWconst [c] x)
(MOVBreg (SRWconst [c] x)) && c>24 -> (SRWconst [c] x)
(MOVBreg (SRWconst [c] x)) && c==24 -> (SRAWconst [c] x)

(MOVHZreg (SRDconst [c] x)) && c>=48 -> (SRDconst [c] x)
(MOVHreg (SRDconst [c] x)) && c>48 -> (SRDconst [c] x)
(MOVHreg (SRDconst [c] x)) && c==48 -> (SRADconst [c] x)
(MOVHZreg (SRWconst [c] x)) && c>=16 -> (SRWconst [c] x)
(MOVHreg (SRWconst [c] x)) && c>16 -> (SRWconst [c] x)
(MOVHreg (SRWconst [c] x)) && c==16 -> (SRAWconst [c] x)

(MOVWZreg (SRDconst [c] x)) && c>=32 -> (SRDconst [c] x)
(MOVWreg (SRDconst [c] x)) && c>32 -> (SRDconst [c] x)
(MOVWreg (SRDconst [c] x)) && c==32 -> (SRADconst [c] x)

// Various redundant zero/sign extension combinations.
(MOVBZreg y:(MOVBZreg _)) -> y  // repeat
(MOVBreg y:(MOVBreg _)) -> y // repeat
(MOVBreg (MOVBZreg x)) -> (MOVBreg x)
(MOVBZreg (MOVBreg x)) -> (MOVBZreg x)

// H - there are more combinations than these

(MOVHZreg y:(MOVHZreg _)) -> y // repeat
(MOVHZreg y:(MOVBZreg _)) -> y // wide of narrow
(MOVHZreg y:(MOVHBRload _ _)) -> y

(MOVHreg y:(MOVHreg _)) -> y // repeat
(MOVHreg y:(MOVBreg _)) -> y // wide of narrow

(MOVHreg y:(MOVHZreg x)) -> (MOVHreg x)
(MOVHZreg y:(MOVHreg x)) -> (MOVHZreg x)

// W - there are more combinations than these

(MOVWZreg y:(MOVWZreg _)) -> y // repeat
(MOVWZreg y:(MOVHZreg _)) -> y // wide of narrow
(MOVWZreg y:(MOVBZreg _)) -> y // wide of narrow
(MOVWZreg y:(MOVHBRload _ _)) -> y
(MOVWZreg y:(MOVWBRload _ _)) -> y

(MOVWreg y:(MOVWreg _)) -> y // repeat
(MOVWreg y:(MOVHreg _)) -> y // wide of narrow
(MOVWreg y:(MOVBreg _)) -> y // wide of narrow

(MOVWreg y:(MOVWZreg x)) -> (MOVWreg x)
(MOVWZreg y:(MOVWreg x)) -> (MOVWZreg x)

// Arithmetic constant ops

(ADD x (MOVDconst [c])) && is32Bit(c) -> (ADDconst [c] x)
(ADDconst [c] (ADDconst [d] x)) && is32Bit(c+d) -> (ADDconst [c+d] x)
(ADDconst [0] x) -> x
(SUB x (MOVDconst [c])) && is32Bit(-c) -> (ADDconst [-c] x)
// TODO deal with subtract-from-const

(ADDconst [c] (MOVDaddr [d] {sym} x)) -> (MOVDaddr [c+d] {sym} x)

// Use register moves instead of stores and loads to move int<->float values
// Common with math Float64bits, Float64frombits
(MOVDload [off] {sym} ptr (FMOVDstore [off] {sym} ptr x _)) -> (MFVSRD x)
(FMOVDload [off] {sym} ptr (MOVDstore [off] {sym} ptr x _)) -> (MTVSRD x)

(FMOVDstore [off] {sym} ptr (MTVSRD x) mem) -> (MOVDstore [off] {sym} ptr x mem)
(MOVDstore [off] {sym} ptr (MFVSRD x) mem) -> (FMOVDstore [off] {sym} ptr x mem)

(MTVSRD (MOVDconst [c])) -> (FMOVDconst [c])
(MFVSRD (FMOVDconst [c])) -> (MOVDconst [c])

(MTVSRD x:(MOVDload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) -> @x.Block (FMOVDload [off] {sym} ptr mem)
(MFVSRD x:(FMOVDload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) -> @x.Block (MOVDload [off] {sym} ptr mem)

// Fold offsets for stores.
(MOVDstore [off1] {sym} (ADDconst [off2] x) val mem) && is16Bit(off1+off2) -> (MOVDstore [off1+off2] {sym} x val mem)
(MOVWstore [off1] {sym} (ADDconst [off2] x) val mem) && is16Bit(off1+off2) -> (MOVWstore [off1+off2] {sym} x val mem)
(MOVHstore [off1] {sym} (ADDconst [off2] x) val mem) && is16Bit(off1+off2) -> (MOVHstore [off1+off2] {sym} x val mem)
(MOVBstore [off1] {sym} (ADDconst [off2] x) val mem) && is16Bit(off1+off2) -> (MOVBstore [off1+off2] {sym} x val mem)

(FMOVSstore [off1] {sym} (ADDconst [off2] ptr) val mem) && is16Bit(off1+off2) -> (FMOVSstore [off1+off2] {sym} ptr val mem)
(FMOVDstore [off1] {sym} (ADDconst [off2] ptr) val mem) && is16Bit(off1+off2) -> (FMOVDstore [off1+off2] {sym} ptr val mem)

// Fold address into load/store.
// The assembler needs to generate several instructions and use
// temp register for accessing global, and each time it will reload
// the temp register. So don't fold address of global, unless there
// is only one use.
(MOVBstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVBstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
(MOVHstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVHstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
(MOVWstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVWstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
(MOVDstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVDstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)

(FMOVSstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (FMOVSstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
(FMOVDstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (FMOVDstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)

(MOVBZload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVBZload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(MOVHload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVHload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(MOVHZload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVHZload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(MOVWload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVWload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(MOVWZload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVWZload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(MOVDload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (MOVDload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(FMOVSload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (FMOVSload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
(FMOVDload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
	&& (ptr.Op != OpSB || p.Uses == 1) ->
        (FMOVDload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)

// Fold offsets for loads.
(FMOVSload [off1] {sym} (ADDconst [off2] ptr) mem) && is16Bit(off1+off2) -> (FMOVSload [off1+off2] {sym} ptr mem)
(FMOVDload [off1] {sym} (ADDconst [off2] ptr) mem) && is16Bit(off1+off2) -> (FMOVDload [off1+off2] {sym} ptr mem)

(MOVDload [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) -> (MOVDload [off1+off2] {sym} x mem)
(MOVWload [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) -> (MOVWload [off1+off2] {sym} x mem)
(MOVWZload [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) -> (MOVWZload [off1+off2] {sym} x mem)
(MOVHload [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) -> (MOVHload [off1+off2] {sym} x mem)
(MOVHZload [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) -> (MOVHZload [off1+off2] {sym} x mem)
(MOVBZload [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) -> (MOVBZload [off1+off2] {sym} x mem)

// Determine load + addressing that can be done as a register indexed load
(MOV(D|W|WZ|H|HZ|BZ)load [0] {sym} p:(ADD ptr idx) mem) && sym == nil && p.Uses == 1 -> (MOV(D|W|WZ|H|HZ|BZ)loadidx ptr idx mem)

// Determine indexed loads with constant values that can be done without index
(MOV(D|W|WZ|H|HZ|BZ)loadidx ptr (MOVDconst [c]) mem) && is16Bit(c) -> (MOV(D|W|WZ|H|HZ|BZ)load [c] ptr mem)
(MOV(D|W|WZ|H|HZ|BZ)loadidx (MOVDconst [c]) ptr mem) && is16Bit(c) -> (MOV(D|W|WZ|H|HZ|BZ)load [c] ptr mem)


// Store of zero -> storezero
(MOVDstore [off] {sym} ptr (MOVDconst [0]) mem) -> (MOVDstorezero [off] {sym} ptr mem)
(MOVWstore [off] {sym} ptr (MOVDconst [0]) mem) -> (MOVWstorezero [off] {sym} ptr mem)
(MOVHstore [off] {sym} ptr (MOVDconst [0]) mem) -> (MOVHstorezero [off] {sym} ptr mem)
(MOVBstore [off] {sym} ptr (MOVDconst [0]) mem) -> (MOVBstorezero [off] {sym} ptr mem)

// Fold offsets for storezero
(MOVDstorezero [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) ->
    (MOVDstorezero [off1+off2] {sym} x mem)
(MOVWstorezero [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) ->
    (MOVWstorezero [off1+off2] {sym} x mem)
(MOVHstorezero [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) ->
    (MOVHstorezero [off1+off2] {sym} x mem)
(MOVBstorezero [off1] {sym} (ADDconst [off2] x) mem) && is16Bit(off1+off2) ->
    (MOVBstorezero [off1+off2] {sym} x mem)

// Stores with addressing that can be done as indexed stores
(MOV(D|W|H|B)store [off] {sym} p:(ADD ptr idx) val mem) && off == 0 && sym == nil && p.Uses == 1 -> (MOV(D|W|H|B)storeidx ptr idx val mem)

// Stores with constant index values can be done without indexed instructions
(MOV(D|W|H|B)storeidx ptr (MOVDconst [c]) val mem) && is16Bit(c) -> (MOV(D|W|H|B)store [c] ptr val mem)
(MOV(D|W|H|B)storeidx (MOVDconst [c]) ptr val mem) && is16Bit(c) -> (MOV(D|W|H|B)store [c] ptr val mem)

// Fold symbols into storezero
(MOVDstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
	&& (x.Op != OpSB || p.Uses == 1) ->
    (MOVDstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
(MOVWstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
	&& (x.Op != OpSB || p.Uses == 1) ->
    (MOVWstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
(MOVHstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
	&& (x.Op != OpSB || p.Uses == 1) ->
    (MOVHstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
(MOVBstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
	&& (x.Op != OpSB || p.Uses == 1) ->
    (MOVBstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)

// atomic intrinsics
(AtomicLoad(32|64|Ptr)  ptr mem) -> (LoweredAtomicLoad(32|64|Ptr) [1] ptr mem)
(AtomicLoadAcq32        ptr mem) -> (LoweredAtomicLoad32 [0] ptr mem)

(AtomicStore(32|64)      ptr val mem) -> (LoweredAtomicStore(32|64) [1] ptr val mem)
(AtomicStoreRel32        ptr val mem) -> (LoweredAtomicStore32 [0] ptr val mem)
//(AtomicStorePtrNoWB ptr val mem) -> (STLR  ptr val mem)

(AtomicExchange(32|64) ptr val mem) -> (LoweredAtomicExchange(32|64) ptr val mem)

(AtomicAdd(32|64) ptr val mem) -> (LoweredAtomicAdd(32|64) ptr val mem)

(AtomicCompareAndSwap(32|64) ptr old new_ mem) -> (LoweredAtomicCas(32|64) [1] ptr old new_ mem)
(AtomicCompareAndSwapRel32   ptr old new_ mem) -> (LoweredAtomicCas32 [0] ptr old new_ mem)

(AtomicAnd8 ptr val mem) -> (LoweredAtomicAnd8 ptr val mem)
(AtomicOr8  ptr val mem) -> (LoweredAtomicOr8  ptr val mem)

// Lowering extension
// Note: we always extend to 64 bits even though some ops don't need that many result bits.
(SignExt8to(16|32|64)  x) -> (MOVBreg x)
(SignExt16to(32|64) x) -> (MOVHreg x)
(SignExt32to64 x) -> (MOVWreg x)

(ZeroExt8to(16|32|64)  x) -> (MOVBZreg x)
(ZeroExt16to(32|64) x) -> (MOVHZreg x)
(ZeroExt32to64 x) -> (MOVWZreg x)

(Trunc(16|32|64)to8 <t> x) && isSigned(t) -> (MOVBreg x)
(Trunc(16|32|64)to8  x) -> (MOVBZreg x)
(Trunc(32|64)to16 <t> x) && isSigned(t) -> (MOVHreg x)
(Trunc(32|64)to16 x) -> (MOVHZreg x)
(Trunc64to32 <t> x) && isSigned(t) -> (MOVWreg x)
(Trunc64to32 x) -> (MOVWZreg x)

(Slicemask <t> x) -> (SRADconst (NEG <t> x) [63])

// Note that MOV??reg returns a 64-bit int, x is not necessarily that wide
// This may interact with other patterns in the future. (Compare with arm64)
(MOV(B|H|W)Zreg x:(MOVBZload _ _)) -> x
(MOV(B|H|W)Zreg x:(MOVBZloadidx _ _ _)) -> x
(MOV(H|W)Zreg x:(MOVHZload _ _)) -> x
(MOV(H|W)Zreg x:(MOVHZloadidx _ _ _)) -> x
(MOV(H|W)reg x:(MOVHload _ _)) -> x
(MOV(H|W)reg x:(MOVHloadidx _ _ _)) -> x
(MOVWZreg x:(MOVWZload _ _)) -> x
(MOVWZreg x:(MOVWZloadidx _ _ _)) -> x
(MOVWreg x:(MOVWload _ _)) -> x
(MOVWreg x:(MOVWloadidx _ _ _)) -> x

// don't extend if argument is already extended
(MOVBreg x:(Arg <t>)) && is8BitInt(t) && isSigned(t) -> x
(MOVBZreg x:(Arg <t>)) && is8BitInt(t) && !isSigned(t) -> x
(MOVHreg x:(Arg <t>)) && (is8BitInt(t) || is16BitInt(t)) && isSigned(t) -> x
(MOVHZreg x:(Arg <t>)) && (is8BitInt(t) || is16BitInt(t)) && !isSigned(t) -> x
(MOVWreg x:(Arg <t>)) && (is8BitInt(t) || is16BitInt(t) || is32BitInt(t)) && isSigned(t) -> x
(MOVWZreg x:(Arg <t>)) && (is8BitInt(t) || is16BitInt(t) || is32BitInt(t)) && !isSigned(t) -> x

(MOVBZreg (MOVDconst [c]))  -> (MOVDconst [int64(uint8(c))])
(MOVBreg (MOVDconst [c]))  -> (MOVDconst [int64(int8(c))])
(MOVHZreg (MOVDconst [c]))  -> (MOVDconst [int64(uint16(c))])
(MOVHreg (MOVDconst [c]))  -> (MOVDconst [int64(int16(c))])
(MOVWreg (MOVDconst [c])) -> (MOVDconst [int64(int32(c))])
(MOVWZreg (MOVDconst [c])) -> (MOVDconst [int64(uint32(c))])


// Lose widening ops fed to stores
(MOVBstore [off] {sym} ptr (MOV(B|BZ|H|HZ|W|WZ)reg x) mem) -> (MOVBstore [off] {sym} ptr x mem)
(MOVHstore [off] {sym} ptr (MOV(H|HZ|W|WZ)reg x) mem) -> (MOVHstore [off] {sym} ptr x mem)
(MOVWstore [off] {sym} ptr (MOV(W|WZ)reg x) mem) -> (MOVWstore [off] {sym} ptr x mem)
(MOVBstore [off] {sym} ptr (SRWconst (MOV(H|HZ)reg x) [c]) mem) && c <= 8 -> (MOVBstore [off] {sym} ptr (SRWconst <typ.UInt32> x [c]) mem)
(MOVBstore [off] {sym} ptr (SRWconst (MOV(W|WZ)reg x) [c]) mem) && c <= 24 -> (MOVBstore [off] {sym} ptr (SRWconst <typ.UInt32> x [c]) mem)
(MOVBstoreidx [off] {sym} ptr idx (MOV(B|BZ|H|HZ|W|WZ)reg x) mem) -> (MOVBstoreidx [off] {sym} ptr idx x mem)
(MOVHstoreidx [off] {sym} ptr idx (MOV(H|HZ|W|WZ)reg x) mem) -> (MOVHstoreidx [off] {sym} ptr idx x mem)
(MOVWstoreidx [off] {sym} ptr idx (MOV(W|WZ)reg x) mem) -> (MOVWstoreidx [off] {sym} ptr idx x mem)
(MOVBstoreidx [off] {sym} ptr idx (SRWconst (MOV(H|HZ)reg x) [c]) mem) && c <= 8 -> (MOVBstoreidx [off] {sym} ptr idx (SRWconst <typ.UInt32> x [c]) mem)
(MOVBstoreidx [off] {sym} ptr idx (SRWconst (MOV(W|WZ)reg x) [c]) mem) && c <= 24 -> (MOVBstoreidx [off] {sym} ptr idx (SRWconst <typ.UInt32> x [c]) mem)
(MOVHBRstore {sym} ptr (MOV(H|HZ|W|WZ)reg x) mem) -> (MOVHBRstore {sym} ptr x mem)
(MOVWBRstore {sym} ptr (MOV(W|WZ)reg x) mem) -> (MOVWBRstore {sym} ptr x mem)

// Lose W-widening ops fed to compare-W
(CMPW x (MOVWreg y)) -> (CMPW x y)
(CMPW (MOVWreg x) y) -> (CMPW x y)
(CMPWU x (MOVWZreg y)) -> (CMPWU x y)
(CMPWU (MOVWZreg x) y) -> (CMPWU x y)

(CMP x (MOVDconst [c])) && is16Bit(c) -> (CMPconst x [c])
(CMP (MOVDconst [c]) y) && is16Bit(c) -> (InvertFlags (CMPconst y [c]))
(CMPW x (MOVDconst [c])) && is16Bit(c) -> (CMPWconst x [c])
(CMPW (MOVDconst [c]) y) && is16Bit(c) -> (InvertFlags (CMPWconst y [c]))

(CMPU x (MOVDconst [c])) && isU16Bit(c) -> (CMPUconst x [c])
(CMPU (MOVDconst [c]) y) && isU16Bit(c) -> (InvertFlags (CMPUconst y [c]))
(CMPWU x (MOVDconst [c])) && isU16Bit(c) -> (CMPWUconst x [c])
(CMPWU (MOVDconst [c]) y) && isU16Bit(c) -> (InvertFlags (CMPWUconst y [c]))

// A particular pattern seen in cgo code:
(AND (MOVDconst [c]) x:(MOVBZload _ _)) -> (ANDconst [c&0xFF] x)
(AND x:(MOVBZload _ _) (MOVDconst [c])) -> (ANDconst [c&0xFF] x)

// floating point negative abs
(FNEG (FABS x)) -> (FNABS x)
(FNEG (FNABS x)) -> (FABS x)

// floating-point fused multiply-add/sub
(FADD (FMUL x y) z) -> (FMADD x y z)
(FSUB (FMUL x y) z) -> (FMSUB x y z)
(FADDS (FMULS x y) z) -> (FMADDS x y z)
(FSUBS (FMULS x y) z) -> (FMSUBS x y z)


// The following statements are found in encoding/binary functions UintXX (load) and PutUintXX (store)
// and convert the statements in these functions from multiple single byte loads or stores to
// the single largest possible load or store.
// Some are marked big or little endian based on the order in which the bytes are loaded or stored,
// not on the ordering of the machine. These are intended for little endian machines.
// To implement for big endian machines, most rules would have to be duplicated but the
// resulting rule would be reversed, i. e., MOVHZload on little endian would be MOVHBRload on big endian
// and vice versa.
// b[0] | b[1]<<8 -> load 16-bit Little endian
(OR <t> x0:(MOVBZload [i0] {s} p mem)
	o1:(SL(W|D)const x1:(MOVBZload [i1] {s} p mem) [8]))
	&& !config.BigEndian
	&& i1 == i0+1
	&& x0.Uses ==1 && x1.Uses == 1
	&& o1.Uses == 1
	&& mergePoint(b, x0, x1) != nil
	&& clobber(x0) && clobber(x1) && clobber(o1)
	 -> @mergePoint(b,x0,x1) (MOVHZload <t> {s} [i0] p mem)

// b[0]<<8 | b[1] -> load 16-bit Big endian on Little endian arch.
// Use byte-reverse indexed load for 2 bytes.
(OR <t> x0:(MOVBZload [i1] {s} p mem)
	o1:(SL(W|D)const x1:(MOVBZload [i0] {s} p mem) [8]))
	&& !config.BigEndian
	&& i1 == i0+1
	&& x0.Uses ==1 && x1.Uses == 1
	&& o1.Uses == 1
	&& mergePoint(b, x0, x1) != nil
	&& clobber(x0) && clobber(x1) && clobber(o1)
	  -> @mergePoint(b,x0,x1) (MOVHBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem)

// b[0]<<n+8 | b[1]<<n -> load 16-bit Big endian (where n%8== 0)
// Use byte-reverse indexed load for 2 bytes,
// then shift left to the correct position. Used to match subrules
// from longer rules.
(OR <t> s0:(SL(W|D)const x0:(MOVBZload [i1] {s} p mem) [n1])
	s1:(SL(W|D)const x1:(MOVBZload [i0] {s} p mem) [n2]))
	&& !config.BigEndian
	&& i1 == i0+1
	&& n1%8 == 0
	&& n2 == n1+8
	&& x0.Uses == 1 && x1.Uses == 1
	&& s0.Uses == 1 && s1.Uses == 1
	&& mergePoint(b, x0, x1) != nil
	&& clobber(x0) && clobber(x1) && clobber(s0) && clobber(s1)
	  -> @mergePoint(b,x0,x1) (SLDconst <t> (MOVHBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem) [n1])

// b[0] | b[1]<<8 | b[2]<<16 | b[3]<<24 -> load 32-bit Little endian
// Use byte-reverse indexed load for 4 bytes.
(OR <t> s1:(SL(W|D)const x2:(MOVBZload [i3] {s} p mem) [24])
	o0:(OR <t> s0:(SL(W|D)const x1:(MOVBZload [i2] {s} p mem) [16])
	x0:(MOVHZload [i0] {s} p mem)))
	&& !config.BigEndian
	&& i2 == i0+2
	&& i3 == i0+3
	&& x0.Uses ==1 && x1.Uses == 1 && x2.Uses == 1
	&& o0.Uses == 1
	&& s0.Uses == 1 && s1.Uses == 1
	&& mergePoint(b, x0, x1, x2) != nil
	&& clobber(x0) && clobber(x1) && clobber(x2)
	&& clobber(s0) && clobber(s1)
	&& clobber(o0)
	 -> @mergePoint(b,x0,x1,x2) (MOVWZload <t> {s} [i0] p mem)

// b[0]<<24 | b[1]<<16 | b[2]<<8 | b[3] -> load 32-bit Big endian order on Little endian arch
// Use byte-reverse indexed load for 4 bytes with computed address.
// Could be used to match subrules of a longer rule.
(OR <t> s1:(SL(W|D)const x2:(MOVBZload [i0] {s} p mem) [24])
	o0:(OR <t> s0:(SL(W|D)const x1:(MOVBZload [i1] {s} p mem) [16])
	x0:(MOVHBRload <t> (MOVDaddr <typ.Uintptr> [i2] {s} p) mem)))
	&& !config.BigEndian
	&& i1 == i0+1
	&& i2 == i0+2
	&& x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1
	&& o0.Uses == 1
	&& s0.Uses == 1 && s1.Uses == 1
	&& mergePoint(b, x0, x1, x2) != nil
	&& clobber(x0) && clobber(x1) && clobber(x2)
	&& clobber(s0) && clobber(s1)
	&& clobber(o0)
	  -> @mergePoint(b,x0,x1,x2) (MOVWBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem)

// b[3] | b[2]<<8 | b[1]<<16 | b[0]<<24 -> load 32-bit Big endian order on Little endian arch
// Use byte-reverse indexed load for 4 bytes with computed address.
// Could be used to match subrules of a longer rule.
(OR <t> x0:(MOVBZload [i3] {s} p mem)
	o0:(OR <t> s0:(SL(W|D)const x1:(MOVBZload [i2] {s} p mem) [8])
	s1:(SL(W|D)const x2:(MOVHBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem) [16])))
	&& !config.BigEndian
	&& i2 == i0+2
	&& i3 == i0+3
	&& x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1
	&& o0.Uses == 1
	&& s0.Uses == 1 && s1.Uses == 1
	&& mergePoint(b, x0, x1, x2) != nil
	&& clobber(x0) && clobber(x1) && clobber(x2)
	&& clobber(s0) && clobber(s1)
	&& clobber(o0)
	  -> @mergePoint(b,x0,x1,x2) (MOVWBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem)

// b[0]<<56 | b[1]<<48 | b[2]<<40 | b[3]<<32 -> load 32-bit Big endian order on Little endian arch
// Use byte-reverse indexed load to for 4 bytes with computed address.
// Used to match longer rules.
(OR <t> s2:(SLDconst x2:(MOVBZload [i3] {s} p mem) [32])
	o0:(OR <t> s1:(SLDconst x1:(MOVBZload [i2] {s} p mem) [40])
	s0:(SLDconst x0:(MOVHBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem) [48])))
	&& !config.BigEndian
	&& i2 == i0+2
	&& i3 == i0+3
	&& x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1
	&& o0.Uses == 1
	&& s0.Uses == 1 && s1.Uses == 1 && s2.Uses == 1
	&& mergePoint(b, x0, x1, x2) != nil
	&& clobber(x0) && clobber(x1) && clobber(x2)
	&& clobber(s0) && clobber(s1) && clobber(s2)
	&& clobber(o0)
	  -> @mergePoint(b,x0,x1,x2) (SLDconst <t> (MOVWBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem) [32])

// b[3]<<32 | b[2]<<40 | b[1]<<48 | b[0]<<56 -> load 32-bit Big endian order on Little endian arch
// Use byte-reverse indexed load for 4 bytes with constant address.
// Used to match longer rules.
(OR <t> s2:(SLDconst x2:(MOVBZload [i0] {s} p mem) [56])
        o0:(OR <t> s1:(SLDconst x1:(MOVBZload [i1] {s} p mem) [48])
        s0:(SLDconst x0:(MOVHBRload <t> (MOVDaddr <typ.Uintptr> [i2] {s} p) mem) [32])))
        && !config.BigEndian
        && i1 == i0+1
        && i2 == i0+2
        && x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1
        && o0.Uses == 1
        && s0.Uses == 1 && s1.Uses == 1 && s2.Uses == 1
        && mergePoint(b, x0, x1, x2) != nil
        && clobber(x0) && clobber(x1) && clobber(x2)
        && clobber(s0) && clobber(s1) && clobber(s2)
        && clobber(o0)
          -> @mergePoint(b,x0,x1,x2) (SLDconst <t> (MOVWBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem) [32])

// b[0] | b[1]<<8 | b[2]<<16 | b[3]<<24 | b[4] <<32 | b[5]<<40 | b[6]<<48 | b[7]<<56 -> load 64-bit Little endian
// Rules with commutative ops and many operands will result in extremely large functions in rewritePPC64,
// so matching shorter previously defined subrules is important.
// Offset must be multiple of 4 for MOVD
(OR <t> s6:(SLDconst x7:(MOVBZload [i7] {s} p mem) [56])
	o5:(OR <t> s5:(SLDconst x6:(MOVBZload [i6] {s} p mem) [48])
	o4:(OR <t> s4:(SLDconst x5:(MOVBZload [i5] {s} p mem) [40])
	o3:(OR <t> s3:(SLDconst x4:(MOVBZload [i4] {s} p mem) [32])
	x0:(MOVWZload {s} [i0] p mem)))))
	&& !config.BigEndian
	&& i0%4 == 0
	&& i4 == i0+4
	&& i5 == i0+5
	&& i6 == i0+6
	&& i7 == i0+7
	&& x0.Uses == 1 && x4.Uses == 1 && x5.Uses == 1 && x6.Uses ==1 && x7.Uses == 1
	&& o3.Uses == 1 && o4.Uses == 1 && o5.Uses == 1
	&& s3.Uses == 1 && s4.Uses == 1 && s5.Uses == 1 && s6.Uses == 1
	&& mergePoint(b, x0, x4, x5, x6, x7) != nil
	&& clobber(x0) && clobber(x4) && clobber(x5) && clobber(x6) && clobber(x7)
	&& clobber(s3) && clobber(s4) && clobber(s5) && clobber (s6)
	&& clobber(o3) && clobber(o4) && clobber(o5)
	  -> @mergePoint(b,x0,x4,x5,x6,x7) (MOVDload <t> {s} [i0] p mem)

// b[7] | b[6]<<8 | b[5]<<16 | b[4]<<24 | b[3]<<32 | b[2]<<40 | b[1]<<48 | b[0]<<56 load 64-bit Big endian ordered bytes on Little endian arch
// Use byte-reverse indexed load of 8 bytes.
// Rules with commutative ops and many operands can result in extremely large functions in rewritePPC64,
// so matching shorter previously defined subrules is important.
(OR <t> s0:(SLDconst x0:(MOVBZload [i0] {s} p mem) [56])
	o0:(OR <t> s1:(SLDconst x1:(MOVBZload [i1] {s} p mem) [48])
	o1:(OR <t> s2:(SLDconst x2:(MOVBZload [i2] {s} p mem) [40])
	o2:(OR <t> s3:(SLDconst x3:(MOVBZload [i3] {s} p mem) [32])
	x4:(MOVWBRload <t> (MOVDaddr <typ.Uintptr> [i4] p) mem)))))
	&& !config.BigEndian
	&& i1 == i0+1
	&& i2 == i0+2
	&& i3 == i0+3
	&& i4 == i0+4
	&& x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1 && x3.Uses == 1 && x4.Uses == 1
	&& o0.Uses == 1 && o1.Uses == 1 && o2.Uses == 1
	&& s0.Uses == 1 && s1.Uses == 1 && s2.Uses == 1 && s3.Uses == 1
	&& mergePoint(b, x0, x1, x2, x3, x4) != nil
	&& clobber(x0) && clobber(x1) && clobber(x2) && clobber(x3) && clobber(x4)
	&& clobber(o0) && clobber(o1) && clobber(o2)
	&& clobber(s0) && clobber(s1) && clobber(s2) && clobber(s3)
	  -> @mergePoint(b,x0,x1,x2,x3,x4) (MOVDBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem)

// b[0]<<56 | b[1]<<48 | b[2]<<40 | b[3]<<32 | b[4]<<24 | b[5]<<16 | b[6]<<8 | b[7] -> load 64-bit Big endian ordered bytes on Little endian arch
// Use byte-reverse indexed load of 8 bytes.
// Rules with commutative ops and many operands can result in extremely large functions in rewritePPC64,
// so matching shorter previously defined subrules is important.
(OR <t> x7:(MOVBZload [i7] {s} p mem)
	o5:(OR <t> s6:(SLDconst x6:(MOVBZload [i6] {s} p mem) [8])
	o4:(OR <t> s5:(SLDconst x5:(MOVBZload [i5] {s} p mem) [16])
	o3:(OR <t> s4:(SLDconst x4:(MOVBZload [i4] {s} p mem) [24])
	s0:(SL(W|D)const x3:(MOVWBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem) [32])))))
	&& !config.BigEndian
	&& i4 == i0+4
	&& i5 == i0+5
	&& i6 == i0+6
	&& i7 == i0+7
	&& x3.Uses == 1 && x4.Uses == 1 && x5.Uses == 1 && x6.Uses == 1 && x7.Uses == 1
	&& o3.Uses == 1 && o4.Uses == 1 && o5.Uses == 1
	&& s0.Uses == 1 && s4.Uses == 1 && s5.Uses == 1 && s6.Uses == 1
	&& mergePoint(b, x3, x4, x5, x6, x7) != nil
	&& clobber(x3) && clobber(x4) && clobber(x5) && clobber(x6) && clobber(x7)
	&& clobber(o3) && clobber(o4) && clobber(o5)
	&& clobber(s0) && clobber(s4) && clobber(s5) && clobber(s6)
	-> @mergePoint(b,x3,x4,x5,x6,x7) (MOVDBRload <t> (MOVDaddr <typ.Uintptr> [i0] {s} p) mem)

// 2 byte store Little endian as in:
//      b[0] = byte(v >> 16)
//      b[1] = byte(v >> 24)
// Added for use in matching longer rules.
(MOVBstore [i1] {s} p (SR(W|D)const w [24])
        x0:(MOVBstore [i0] {s} p (SR(W|D)const w [16]) mem))
        && !config.BigEndian
        && x0.Uses == 1
        && i1 == i0+1
        && clobber(x0)
          -> (MOVHstore [i0] {s} p (SRWconst <typ.UInt16> w [16]) mem)

// 2 byte store Little endian as in:
//      b[0] = byte(v)
//      b[1] = byte(v >> 8)
(MOVBstore [i1] {s} p (SR(W|D)const w [8])
	x0:(MOVBstore [i0] {s} p w mem))
	&& !config.BigEndian
	&& x0.Uses == 1
	&& i1 == i0+1
	&& clobber(x0)
	  -> (MOVHstore [i0] {s} p w mem)

// 4 byte store Little endian as in:
//     b[0:1] = uint16(v)
//     b[2:3] = uint16(v >> 16)
(MOVHstore [i1] {s} p (SR(W|D)const w [16])
	x0:(MOVHstore [i0] {s} p w mem))
	&& !config.BigEndian
	&& x0.Uses == 1
	&& i1 == i0+2
	&& clobber(x0)
	  -> (MOVWstore [i0] {s} p w mem)

// 4 byte store Big endian as in:
//     b[0] = byte(v >> 24)
//     b[1] = byte(v >> 16)
//     b[2] = byte(v >> 8)
//     b[3] = byte(v)
// Use byte-reverse indexed 4 byte store.
(MOVBstore [i3] {s} p w
	x0:(MOVBstore [i2] {s} p (SRWconst w [8])
	x1:(MOVBstore [i1] {s} p (SRWconst w [16])
	x2:(MOVBstore [i0] {s} p (SRWconst w [24]) mem))))
	&& !config.BigEndian
	&& x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1
	&& i1 == i0+1 && i2 == i0+2 && i3 == i0+3
	&& clobber(x0) && clobber(x1) && clobber(x2)
	  -> (MOVWBRstore (MOVDaddr <typ.Uintptr> [i0] {s} p) w mem)

// The 2 byte store appears after the 4 byte store so that the
// match for the 2 byte store is not done first.
// If the 4 byte store is based on the 2 byte store then there are 
// variations on the MOVDaddr subrule that would require additional
// rules to be written.

// 2 byte store Big endian as in:
//      b[0] = byte(v >> 8)
//      b[1] = byte(v)
(MOVBstore [i1] {s} p w x0:(MOVBstore [i0] {s} p (SRWconst w [8]) mem))
	&& !config.BigEndian
	&& x0.Uses == 1
	&& i1 == i0+1
	&& clobber(x0)
	  -> (MOVHBRstore (MOVDaddr <typ.Uintptr> [i0] {s} p) w mem)

// 8 byte store Little endian as in:
//	b[0] = byte(v)
//	b[1] = byte(v >> 8)
//	b[2] = byte(v >> 16)
//	b[3] = byte(v >> 24)
//	b[4] = byte(v >> 32)
//	b[5] = byte(v >> 40)
//	b[6] = byte(v >> 48)
//	b[7] = byte(v >> 56)
// Built on previously defined rules
// Offset must be multiple of 4 for MOVDstore
(MOVBstore [i7] {s} p (SRDconst w [56])
	x0:(MOVBstore [i6] {s} p (SRDconst w [48])
	x1:(MOVBstore [i5] {s} p (SRDconst w [40])
	x2:(MOVBstore [i4] {s} p (SRDconst w [32])
	x3:(MOVWstore [i0] {s} p w mem)))))
	&& !config.BigEndian
	&& i0%4 == 0
	&& x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1 && x3.Uses == 1
	&& i4 == i0+4 && i5 == i0+5 && i6 == i0+6 && i7 == i0+7
	&& clobber(x0) && clobber(x1) && clobber(x2) && clobber(x3)
	  -> (MOVDstore [i0] {s} p w mem)

// 8 byte store Big endian as in:
//      b[0] = byte(v >> 56)
//      b[1] = byte(v >> 48)
//      b[2] = byte(v >> 40)
//      b[3] = byte(v >> 32)
//      b[4] = byte(v >> 24)
//      b[5] = byte(v >> 16)
//      b[6] = byte(v >> 8)
//      b[7] = byte(v)
// Use byte-reverse indexed 8 byte store.
(MOVBstore [i7] {s} p w
        x0:(MOVBstore [i6] {s} p (SRDconst w [8])
        x1:(MOVBstore [i5] {s} p (SRDconst w [16])
        x2:(MOVBstore [i4] {s} p (SRDconst w [24])
        x3:(MOVBstore [i3] {s} p (SRDconst w [32])
        x4:(MOVBstore [i2] {s} p (SRDconst w [40])
        x5:(MOVBstore [i1] {s} p (SRDconst w [48])
        x6:(MOVBstore [i0] {s} p (SRDconst w [56]) mem))))))))
        && !config.BigEndian
        && x0.Uses == 1 && x1.Uses == 1 && x2.Uses == 1 && x3.Uses == 1 && x4.Uses == 1 && x5.Uses == 1 && x6.Uses == 1
        && i1 == i0+1 && i2 == i0+2 && i3 == i0+3 && i4 == i0+4 && i5 == i0+5 && i6 == i0+6 && i7 == i0+7
        && clobber(x0) && clobber(x1) && clobber(x2) && clobber(x3) && clobber(x4) && clobber(x5) && clobber(x6)
          -> (MOVDBRstore (MOVDaddr <typ.Uintptr> [i0] {s} p) w mem)