TEXT ·IndexByte<ABIInternal>(SB),NOSPLIT,$0-40
	// R4 = b_base
	// R5 = b_len
	// R6 = b_cap (unused)
	// R7 = byte to find
	AND	$0xff, R7
	MOVV	R4, R6		// store base for later
	ADDV	R4, R5		// end
	ADDV	$-1, R4

	PCALIGN	$16
loop:
	ADDV	$1, R4
	BEQ	R4, R5, notfound
	MOVBU	(R4), R8
	BNE	R7, R8, loop

	SUBV	R6, R4		// remove base
	RET

notfound:
	MOVV	$-1, R4
	RET

	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

 *
 * Calling into the go tool chain, where all registers are caller save.
 * Called from standard s390x C ABI, where r6-r13, r15, and f8-f15 are
 * callee-save, so they must be saved explicitly.
 */
.globl crosscall_s390x
crosscall_s390x:
	/* save r6-r15 in the register save area of the calling function */
	stmg    %r6, %r15, 48(%r15)

	/* allocate 64 bytes of stack space to save f8-f15 */
	lay     %r15, -64(%r15)

	MOVV	tid+8(FP), R5
	MOVV	sig+16(FP), R6
	MOVV	$SYS_tgkill, R2
	SYSCALL
	RET

TEXT runtime·setitimer(SB),NOSPLIT|NOFRAME,$0-24
	MOVW	mode+0(FP), R4
	MOVV	new+8(FP), R5
	MOVV	old+16(FP), R6
	MOVV	$SYS_setitimer, R2
	SYSCALL
	RET

TEXT runtime·timer_create(SB),NOSPLIT,$0-28
	MOVW	clockid+0(FP), R4
	MOVV	sevp+8(FP), R5
	MOVV	timerid+16(FP), R6
	MOVV	$SYS_timer_create, R2
	SYSCALL

	VMOVB	$0, W0

	SGTU	$128, R2, R3
	BEQ	R3, msa_large

	AND	$15, R2, R5
	XOR	R2, R5, R6
	ADDVU	R1, R6

msa_small:
	VMOVB	W0, (R1)
	ADDVU	$16, R1
	SGTU	R6, R1, R3
	BNE	R3, R0, msa_small
	BEQ	R5, R0, done
	VMOVB	W0, -16(R4)
	JMP	done

msa_large:
	AND	$127, R2, R5
	XOR	R2, R5, R6
	ADDVU	R1, R6

msa_large_loop:
	VMOVB	W0, (R1)
	VMOVB	W0, 16(R1)
	VMOVB	W0, 32(R1)
	VMOVB	W0, 48(R1)

	RET

// memequal_varlen(a, b unsafe.Pointer) bool
TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT,$40-17
	BEQ	R4, R5, eq
	MOVV	8(REGCTXT), R6    // compiler stores size at offset 8 in the closure
	MOVV	R4, 8(R3)
	MOVV	R5, 16(R3)
	MOVV	R6, 24(R3)
	JAL	runtime·memequal(SB)
	MOVBU	32(R3), R4
	RET
eq:
	MOVV	$1, R4

TEXT runtime·memclrNoHeapPointers<ABIInternal>(SB),NOSPLIT,$0-16
	ADDV	R4, R5, R6

	// if less than 8 bytes, do one byte at a time
	SGTU	$8, R5, R8
	BNE	R8, out

	// do one byte at a time until 8-aligned
	AND	$7, R4, R8
	BEQ	R8, words
	MOVB	R0, (R4)
	ADDV	$1, R4
	JMP	-4(PC)

words:
	// do 8 bytes at a time if there is room
	ADDV	$-7, R6, R5

	PCALIGN	$16
	SGTU	R5, R4, R8
	BEQ	R8, out
	MOVV	R0, (R4)

	MOVW	$0, R4	// child tid ptr
	MOVW	$0, R5

	// Copy mp, gp, fn off parent stack for use by child.
	MOVW	$-16(R1), R1
	MOVW	mp+8(FP), R6
	MOVW	R6, 0(R1)
	MOVW	gp+12(FP), R6
	MOVW	R6, 4(R1)
	MOVW	fn+16(FP), R6
	MOVW	R6, 8(R1)
	MOVW	$1234, R6
	MOVW	R6, 12(R1)

	MOVW	$SYS_clone, R7
	SWI	$0

	// In parent, return.
	CMP	$0, R0
	BEQ	3(PC)
	MOVW	R0, ret+20(FP)
	RET