VPOPCNTB 15(R8)(R14*1), K4, X16                    // 62827d0c5484300f000000
	VPOPCNTB 15(R8)(R14*2), K4, X16                    // 62827d0c5484700f000000
	VPOPCNTB X14, K4, X14                              // 62527d0c54f6
	VPOPCNTB X19, K4, X14                              // 62327d0c54f3
	VPOPCNTB X8, K4, X14                               // 62527d0c54f0

	SWI $SYS_getcontext
	MOVW.CS $0, R8	// crash on syscall failure
	MOVW.CS R8, (R8)
	RET

TEXT runtime·sigprocmask(SB),NOSPLIT,$0
	MOVW how+0(FP), R0	// arg 1 - how
	MOVW new+4(FP), R1	// arg 2 - set
	MOVW old+8(FP), R2	// arg 3 - oset
	SWI $SYS___sigprocmask14
	MOVW.CS $0, R8	// crash on syscall failure
	MOVW.CS R8, (R8)
	RET

TEXT sigreturn_tramp<>(SB),NOSPLIT|NOFRAME,$0

	ADD     $2, R8          // Done processing 2 entries, i and i+1
	CMP     R8, R16         // Are there at least a couple of more entries left?
	BLE     loopback
	CMP     R8, R4          // Are we at the last element?
	BEQ     loopexit
scalar:	
	ADD     $-1, R8, R10
	SLD     $3, R10
	MOVD    (R6)(R10),R11
	SRD     R9, R11, R11    // x[len(z)-2] >> s
	SLD     $3, R8, R12
	MOVD    (R6)(R12), R12

	MULLW	R6, R7                // b91c0076
	MULLW	R6, R7, R8            // b9040087b91c0086
	MULLW	$8192, R6             // a76c2000
	MULLW	$8192, R6, R7         // 1876a77c2000
	MULLW	$-32769, R8           // c281ffff7fff
	MULLW   $-32769, R8, R9       // 1898c291ffff7fff
	MULLD	$-2147483648, R1      // c21080000000
	MULLD   $-2147483648, R1, R2  // b9040021c22080000000

//	LMUL rreg ',' rreg
//	{
//		outcode(int($1), &$2, 0, &$4);
//	}
	MUL	R19, R8		// 01130018
	MULU	R21, R13	// 01b50019
	MULV	R19, R8		// 0113001c
	MULVU	R21, R13	// 01b5001d

//	LDIV rreg ',' rreg
//	{
//		outcode(int($1), &$2, 0, &$4);
//	}
	DIV	R18, R22	// 02d2001a
	DIVU	R14, R9		// 012e001b
	DIVV	R8, R13		// 01a8001e
	DIVVU	R16, R19	// 0270001f

//	LREM rreg ',' rreg
//	{

TEXT	runtime·racefuncenter(SB), NOSPLIT, $0-8
	MOVD	callpc+0(FP), R8
	BR	racefuncenter<>(SB)

// Common code for racefuncenter
// R11 = caller's return address
TEXT	racefuncenter<>(SB), NOSPLIT, $0-0
	MOVD    runtime·tls_g(SB), R10
	MOVD    0(R10), g
	MOVD    g_racectx(g), R3        // goroutine racectx aka *ThreadState
	MOVD	R8, R4			// caller pc set by caller in R8
	// void __tsan_func_enter(ThreadState *thr, void *pc);

	SHA256ROUND0(3, 0xe9b5dba5, R13, R14, R15, R8, R9, R10, R11, R12)
	SHA256ROUND0(4, 0x3956c25b, R12, R13, R14, R15, R8, R9, R10, R11)
	SHA256ROUND0(5, 0x59f111f1, R11, R12, R13, R14, R15, R8, R9, R10)
	SHA256ROUND0(6, 0x923f82a4, R10, R11, R12, R13, R14, R15, R8, R9)
	SHA256ROUND0(7, 0xab1c5ed5, R9, R10, R11, R12, R13, R14, R15, R8)
	SHA256ROUND0(8, 0xd807aa98, R8, R9, R10, R11, R12, R13, R14, R15)

	// first save R0, because runtime·load_g will clobber it.
	MOVD	R0, R8
	// Set up g register.
	CALL	runtime·load_g(SB)
	MOVD	R8, R0

	CBZ	g, sigtrampnog // g == nil
	MOVD	g_m(g), R6
	CBZ	R6, sigtramp    // g.m == nil
	MOVW	m_ncgo(R6), R7
	CBZW	R7, sigtramp    // g.m.ncgo = 0
	MOVD	m_curg(R6), R8
	CBZ	R8, sigtramp    // g.m.curg == nil
	MOVD	g_syscallsp(R8), R7
	CBZ	R7,	sigtramp    // g.m.curg.syscallsp == 0

	MOVWZ	R21,R7		// crc for next round
	BDNZ 	loop
	ANDCC	$7,R6,R8	// any leftover bytes
	BEQ	done		// none --> done
	MOVD	R8,CTR		// byte count
	PCALIGN $16             // align short loop
short:
	MOVBZ 	0(R5),R8	// get v
	XOR 	R8,R7,R8	// byte(crc)^v -> R8
	RLDIC	$2,R8,$54,R8	// rldicl r8,r8,2,22
	SRD 	$8,R7,R14	// crc>>8
	MOVWZ	(R4)(R8),R10
	ADD	$1,R5
	XOR 	R10,R14,R7	// loop crc in R7
	BDNZ 	short

	VPERMI2B 15(R8)(R14*1), Y15, K2, Y2                // 6292052a7594300f000000
	VPERMI2B 15(R8)(R14*2), Y15, K2, Y2                // 6292052a7594700f000000
	VPERMI2B Y18, Y22, K2, Y2                          // 62b24d2275d2
	VPERMI2B Y24, Y22, K2, Y2                          // 62924d2275d0
	VPERMI2B Y9, Y22, K2, Y2                           // 62d24d2275d1