(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

	HasOSXSAVE          bool // OS supports XSAVE/XRESTOR for saving/restoring XMM registers.
	HasPCLMULQDQ        bool // PCLMULQDQ instruction - most often used for AES-GCM
	HasPOPCNT           bool // Hamming weight instruction POPCNT.
	HasRDRAND           bool // RDRAND instruction (on-chip random number generator)
	HasRDSEED           bool // RDSEED instruction (on-chip random number generator)

	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

		{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,

	NEG	R1                    // b9030011
	NEG	R1, R2                // b9030021
	NEGW	R1                    // b9130011
	NEGW	R1, R2                // b9130021
	FLOGR	R2, R2                // b9830022
	POPCNT	R3, R4                // b9e10043

	AND	R1, R2                // b9800021
	AND	R1, R2, R3            // b9e42031
	AND	$-2, R1               // a517fffe
	AND	$-65536, R1           // c01bffff0000

		//
		// Counts the number of ones in each byte of arg0
		// and places the result into the corresponding byte
		// of the result.
		{
			name:         "POPCNT",
			argLength:    1,
			reg:          gp11,
			asm:          "POPCNT",
			typ:          "UInt64",
			clobberFlags: true,
		},

		// unsigned multiplication (64x64 → 128)
		//

	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:

(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)

				{0, 1073733624}, // R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29
			},
		},
	},
	{
		name:   "POPCNTW",
		argLen: 1,
		asm:    ppc64.APOPCNTW,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 1073733630}, // SP SB R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29
			},

    auto output_type = op.getType().clone(
        rewriter.getIntegerType(/*width=*/8, /*isSigned=*/false));
    auto pop_cnt = rewriter.create<TF::PopulationCountOp>(
        op.getLoc(), output_type, op.getOperand());
    auto cast_or_pop_cnt =
        rewriter.createOrFold<TF::CastOp>(op.getLoc(), op.getType(), pop_cnt);
    rewriter.replaceOp(op, {cast_or_pop_cnt});
    return success();
  }
};