(Ctz64NonZero x) && buildcfg.GOAMD64 <  3 => (Select0 (BSFQ x))
(Ctz32NonZero x) && buildcfg.GOAMD64 <  3 => (BSFL x)
(Ctz16NonZero x) && buildcfg.GOAMD64 <  3 => (BSFL x)
(Ctz8NonZero  x) && buildcfg.GOAMD64 <  3 => (BSFL x)

// BitLen64 of a 64 bit value x requires checking whether x == 0, since BSRQ is undefined when x == 0.
// However, for zero-extended values, we can cheat a bit, and calculate

(Ctz32 x) => (CNTTZW (MOVWZreg 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) => (SUBFCconst [64] (CNTLZD <typ.Int> x))
(BitLen32 x) => (SUBFCconst [32] (CNTLZW <typ.Int> x))

(PopCount64 ...) => (POPCNTD ...)
(PopCount(32|16|8) x) => (POPCNT(W|W|B) (MOV(W|H|B)Zreg x))

		v.AddArg2(v0, v1)
		return true
	}
	return false
}
func rewriteValuedec64_OpBitLen64(v *Value) bool {
	v_0 := v.Args[0]
	b := v.Block
	typ := &b.Func.Config.Types
	// match: (BitLen64 x)
	// result: (Add32 <typ.Int> (BitLen32 <typ.Int> (Int64Hi x)) (BitLen32 <typ.Int> (Or32 <typ.UInt32> (Int64Lo x) (Zeromask (Int64Hi x)))))
	for {
		x := v_0
		v.reset(OpAdd32)
		v.Type = typ.Int

(Ctz64 <t> x) => (SUB (MOVDconst [64]) (FLOGR (AND <t> (SUBconst <t> [1] x) (NOT <t> x))))
(Ctz32 <t> x) => (SUB (MOVDconst [64]) (FLOGR (MOVWZreg (ANDW <t> (SUBWconst <t> [1] x) (NOTW <t> x)))))

(BitLen64 x) => (SUB (MOVDconst [64]) (FLOGR x))

// POPCNT treats the input register as a vector of 8 bytes, producing
// a population count for each individual byte. For inputs larger than