AVSRW
	AVSR
	AVSRO
	AVSLD
	AVSRD
	AVSA
	AVSRAB
	AVSRAH
	AVSRAW
	AVSRAD
	AVSOI
	AVSLDOI
	AVCLZ
	AVCLZB
	AVCLZH
	AVCLZW
	AVCLZD
	AVPOPCNT
	AVPOPCNTB
	AVPOPCNTH
	AVPOPCNTW
	AVPOPCNTD
	AVCMPEQ
	AVCMPEQUB
	AVCMPEQUBCC
	AVCMPEQUH
	AVCMPEQUHCC
	AVCMPEQUW
	AVCMPEQUWCC
	AVCMPEQUD
	AVCMPEQUDCC
	AVCMPGT
	AVCMPGTUB
	AVCMPGTUBCC

	AVPMOVZXDQ
	AVPMOVZXWD
	AVPMOVZXWQ
	AVPMULDQ
	AVPMULHRSW
	AVPMULHUW
	AVPMULHW
	AVPMULLD
	AVPMULLQ
	AVPMULLW
	AVPMULTISHIFTQB
	AVPMULUDQ
	AVPOPCNTB
	AVPOPCNTD
	AVPOPCNTQ
	AVPOPCNTW
	AVPOR
	AVPORD
	AVPORQ
	AVPROLD
	AVPROLQ
	AVPROLVD
	AVPROLVQ
	AVPRORD
	AVPRORQ
	AVPRORVD
	AVPRORVQ
	AVPSADBW
	AVPSCATTERDD

	case APOPW, APOPL, APOPQ,
		ALEAL, ALEAQ,
		AIMUL3W, AIMUL3L, AIMUL3Q,
		APEXTRB, APEXTRW, APEXTRD, APEXTRQ, AVPEXTRB, AVPEXTRW, AVPEXTRD, AVPEXTRQ, AEXTRACTPS,
		ABSFW, ABSFL, ABSFQ, ABSRW, ABSRL, ABSRQ, APOPCNTW, APOPCNTL, APOPCNTQ, ALZCNTW, ALZCNTL, ALZCNTQ,
		ASHLXL, ASHLXQ, ASHRXL, ASHRXQ, ASARXL, ASARXQ:
		// These instructions are pure writes to To. They don't use its old value.
		m &^= readTo
	case AXORL, AXORQ:

	AMOVDBR

	// conditional moves
	AMOVDEQ
	AMOVDGE
	AMOVDGT
	AMOVDLE
	AMOVDLT
	AMOVDNE
	ALOCR
	ALOCGR

	// find leftmost one
	AFLOGR

	// population count
	APOPCNT

	// integer bitwise
	AAND
	AANDW
	AOR
	AORW
	AXOR
	AXORW
	ASLW
	ASLD
	ASRW
	ASRAW
	ASRD
	ASRAD
	ARLL
	ARLLG
	ARNSBG
	ARXSBG
	AROSBG
	ARNSBGT

(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
// a single byte we therefore need to sum the individual bytes produced
// by the POPCNT instruction. For example, the following instruction
// sequence could be used to calculate the population count of a 4-byte

		{name: "POPCNTD", argLength: 1, reg: gp11, asm: "POPCNTD"}, // number of set bits in arg0
		{name: "POPCNTW", argLength: 1, reg: gp11, asm: "POPCNTW"}, // number of set bits in each word of arg0 placed in corresponding word
		{name: "POPCNTB", argLength: 1, reg: gp11, asm: "POPCNTB"}, // number of set bits in each byte of arg0 placed in corresponding byte

		{name: "I64Rotl", asm: "I64Rotl", argLength: 2, reg: gp21, typ: "Int64"},     // rotl(arg0, arg1)
		{name: "I64Popcnt", asm: "I64Popcnt", argLength: 1, reg: gp11, typ: "Int64"}, // popcnt(arg0)
	}

	archs = append(archs, arch{
		name:            "Wasm",
		pkg:             "cmd/internal/obj/wasm",
		genfile:         "../../wasm/ssa.go",
		ops:             WasmOps,
		blocks:          nil,

(Ctz64NonZero ...) => (Ctz64 ...)

(Ctz64 x) && buildcfg.GOPPC64<=8 => (POPCNTD (ANDN <typ.Int64> (ADDconst <typ.Int64> [-1] x) x))
(Ctz64 x) => (CNTTZD x)
(Ctz32 x) && buildcfg.GOPPC64<=8 => (POPCNTW (MOVWZreg (ANDN <typ.Int> (ADDconst <typ.Int> [-1] x) x)))
(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)))

	POPCNTW (BX), DX                        // 66f30fb813
	POPCNTW (R11), DX                       // 66f3410fb813
	POPCNTW DX, DX                          // 66f30fb8d2
	POPCNTW R11, DX                         // 66f3410fb8d3
	POPCNTW (BX), R11                       // 66f3440fb81b
	POPCNTW (R11), R11                      // 66f3450fb81b
	POPCNTW DX, R11                         // 66f3440fb8da

(Bswap(64|32) ...) => (BSWAP(Q|L) ...)
(Bswap16 x) => (ROLWconst [8] x)

(PopCount(64|32) ...) => (POPCNT(Q|L) ...)
(PopCount16 x) => (POPCNTL (MOVWQZX <typ.UInt32> x))
(PopCount8 x) => (POPCNTL (MOVBQZX <typ.UInt32> x))

(Sqrt ...) => (SQRTSD ...)
(Sqrt32 ...) => (SQRTSS ...)

(RoundToEven x) => (ROUNDSD [0] x)
(Floor x)       => (ROUNDSD [1] x)