A6:
	MOVD   (R8)(R1*1), R6
	MULHDU R9, R6
	MOVD   (R2)(R1*1), R10
	ADDC   R10, R11        // add to low order bits
	ADDE   R0, R6
	ADDC   R4, R11
	ADDE   R0, R6
	MOVD   R6, R4
	MOVD   R11, (R2)(R1*1)

	MOVD   (8)(R8)(R1*1), R6
	MULHDU R9, R6
	MOVD   (8)(R2)(R1*1), R10
	ADDC   R10, R11           // add to low order bits
	ADDE   R0, R6
	ADDC   R4, R11
	ADDE   R0, R6
	MOVD   R6, R4

	LDADDALW	R5, (R6), RSP                            // ERROR "illegal combination"
	LDADDALH	R5, (R6), RSP                            // ERROR "illegal combination"
	LDADDALB	R5, (R6), RSP                            // ERROR "illegal combination"
	LDADDD	R5, (R6), RSP                                    // ERROR "illegal combination"
	LDADDW	R5, (R6), RSP                                    // ERROR "illegal combination"

	MOVD.W	-8(R2), R6
	LSR	R4, R6, R5	// return value
	LSL	R3, R6, R8	// x[i] << s
	SUB	$1, R1
one:	TBZ	$0, R1, two
	MOVD.W	-8(R2), R6
	LSR	R4, R6, R7
	ORR	R8, R7
	LSL	R3, R6, R8
	SUB	$1, R1
	MOVD.W	R7, -8(R0)
two:
	TBZ	$1, R1, loop
	LDP.W	-16(R2), (R6, R7)
	LSR	R4, R7, R10
	ORR	R8, R10
	LSL	R3, R7
	LSR	R4, R6, R9
	ORR	R7, R9
	LSL	R3, R6, R8
	SUB	$2, R1

	LMG  0(R8), R5, R6
	MOVD $0, 0(R9)          // R9 address of SAVSTACK_ASYNC
	LE_CALL                 // balr R7, R6 (return #1)
	NOPH
	MOVD R3, ret+32(FP)
	CMP  R3, $-1            // compare result to -1
	BNE  done

	// retrieve errno and errno2
	MOVD  zosLibVec<>(SB), R8
	ADD   $(__errno), R8
	LMG   0(R8), R5, R6
	LE_CALL                   // balr R7, R6 __errno (return #3)
	NOPH

	MOVD	savedm-8(SP), R6
	CMPBNE	R6, $0, droppedm

	// Skip dropm to reuse it in the next call, when a pthread key has been created.
	MOVD	_cgo_pthread_key_created(SB), R6
	// It means cgo is disabled when _cgo_pthread_key_created is a nil pointer, need dropm.
	CMPBEQ	R6, $0, dropm
	MOVD	(R6), R6
	CMPBNE	R6, $0, droppedm

dropm:
	MOVD	$runtime·dropm(SB), R3
	BL	(R3)

	// g is not nil. Check further.
	MOVD	g_m(R9), R6
	CMP	$0, R6
	BEQ	sigtramp    // g.m == nil
	MOVW	m_ncgo(R6), R7
	CMPW	$0, R7
	BEQ	sigtramp    // g.m.ncgo = 0
	MOVD	m_curg(R6), R7
	CMP	$0, R7
	BEQ	sigtramp    // g.m.curg == nil
	MOVD	g_syscallsp(R7), R7
	CMP	$0, R7
	BEQ	sigtramp    // g.m.curg.syscallsp == 0
	MOVD	m_cgoCallers(R6), R7 // R7 is the fifth arg in C calling convention.
	CMP	$0, R7

	SUBW	R3, R4, R5            // b9f93054
	SUBW	$8192, R1             // c21500002000
	SUBW	$8192, R1, R2         // 1821c22500002000
	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

	SYSCALL
	MOVW	R4, errno+16(FP)
	RET

// func usleep(usec uint32)
TEXT runtime·usleep(SB),NOSPLIT,$16-4
	MOVWU	usec+0(FP), R7
	MOVV	$1000, R6
	MULVU	R6, R7, R7
	MOVV	$1000000000, R6

	DIVVU	R6, R7, R5	// ts->tv_sec
	REMVU	R6, R7, R4	// ts->tv_nsec
	MOVV	R5, 8(R3)
	MOVV	R4, 16(R3)

	// nanosleep(&ts, 0)
	ADDV	$8, R3, R4
	MOVV	R0, R5
	MOVV	$SYS_nanosleep, R11
	SYSCALL
	RET

	CMP     R11, $0
	BEQ     done

	MOVD    0(R8), R20
	ADD     $-1, R11
	MULLD   R9, R20, R6       // R6 = z0 = Low-order(x[i]*y)
	MULHDU  R9, R20, R7       // R7 = z1 = High-order(x[i]*y)
	ADDC    R4, R6            // R6 = z0 + r
	ADDZE   R7, R4            // R4 = z1 + CA
	CMP     R11, $0
	MOVD    R6, 0(R10)        // z[i]
	BEQ     done

	// We will read 4 elements per iteration

	MOVD	callpc+8(FP), R5
	MOVD	pc+16(FP), R6
	// void __tsan_write_pc(ThreadState *thr, void *addr, void *callpc, void *pc);
	MOVD	$__tsan_write_pc(SB), R8
	BR	racecalladdr<>(SB)

// func runtime·RaceReadRange(addr, size uintptr)
// Called from instrumented Go code.
TEXT	runtime·racereadrange<ABIInternal>(SB), NOSPLIT, $0-16
	MOVD	R4, R5 // size
	MOVD	R3, R4 // addr
	MOVD	LR, R6