// align destination to 4 bytes
	AND	$3, R1, R6
	BEQ	R6, f_dest_aligned
	SUBU	R1, R0, R6
	AND	$3, R6
	MOVWHI	0(R2), R7
	SUBU	R6, R3
	MOVWLO	3(R2), R7
	ADDU	R6, R2
	MOVWHI	R7, 0(R1)
	ADDU	R6, R1

f_dest_aligned:
	AND	$31, R3, R7
	AND	$3, R3, R6
	SUBU	R7, R5, R7	// end pointer for 32-byte chunks
	SUBU	R6, R5, R6	// end pointer for 4-byte chunks

	// if source is not aligned, use unaligned reads

	BNE	R3, small_zero

ptr_align:
	AND	$3, R1, R3
	BEQ	R3, setup
	SUBU	R1, R0, R3
	AND	$3, R3		// R3 contains number of bytes needed to align ptr
	MOVWHI	R0, 0(R1)	// MOVWHI will write zeros up to next word boundary
	SUBU	R3, R2
	ADDU	R3, R1

setup:
	AND	$31, R2, R6
	AND	$3, R2, R5
	SUBU	R6, R4, R6	// end pointer for 32-byte chunks
	SUBU	R5, R4, R5	// end pointer for 4-byte chunks

large:
	BEQ	R1, R6, words

	MOVW	p+4(FP), R5
	MOVW	n+8(FP), R6
	MOVW	$SYS_write, R2
	SYSCALL
	BEQ	R7, 2(PC)
	SUBU	R2, R0, R2	// caller expects negative errno
	MOVW	R2, ret+12(FP)
	RET

TEXT runtime·read(SB),NOSPLIT,$0-16
	MOVW	fd+0(FP), R4
	MOVW	p+4(FP), R5
	MOVW	n+8(FP), R6
	MOVW	$SYS_read, R2
	SYSCALL
	BEQ	R7, 2(PC)
	SUBU	R2, R0, R2	// caller expects negative errno
	MOVW	R2, ret+12(FP)
	RET

	MOVW	b_len+4(FP), R2
	MOVBU	c+12(FP), R3	// byte to find
	ADDU	$1, R1, R4	// store base+1 for later
	ADDU	R1, R2	// end

loop:
	BEQ	R1, R2, notfound
	MOVBU	(R1), R5
	ADDU	$1, R1
	BNE	R3, R5, loop

	SUBU	R4, R1	// R1 will be one beyond the position we want so remove (base+1)
	MOVW	R1, ret+16(FP)
	RET

notfound:
	MOVW	$-1, R1
	MOVW	R1, ret+16(FP)
	RET

TEXT ·IndexByteString(SB),NOSPLIT,$0-16

	 *  first arg.
	 */

	// Space for 9 caller-saved GPR + LR + 6 caller-saved FPR.
	// O32 ABI allows us to smash 16 bytes argument area of caller frame.
#ifndef GOMIPS_softfloat
	SUBU	$(4*14+8*6-16), R29
#else
	SUBU	$(4*14-16), R29	// For soft-float, no FPR.
#endif
	MOVW	R4, (4*1)(R29)	// fn unsafe.Pointer
	MOVW	R5, (4*2)(R29)	// a unsafe.Pointer
	MOVW	R7, (4*3)(R29)	// ctxt uintptr
	MOVW	R16, (4*4)(R29)

	MOVBU	(R3), R6
	ADDU	$1, R3
	MOVBU	(R4), R7
	ADDU	$1, R4
	BEQ	R6, R7 , loop

	SGTU	R6, R7, R8
	MOVW	$-1, R6
	CMOVZ	R8, R6, R8
	JMP	cmp_ret
samebytes:
	SGTU	R1, R2, R6
	SGTU	R2, R1, R7
	SUBU	R7, R6, R8
cmp_ret:
	MOVW	R8, ret+24(FP)
	RET

TEXT runtime·cmpstring(SB),NOSPLIT,$0-20
	MOVW	a_base+0(FP), R3
	MOVW	a_len+4(FP), R1
	MOVW	b_base+8(FP), R4
	MOVW	b_len+12(FP), R2
	BEQ	R3, R4, samebytes

	}
	return x + int64(y)
}

// subU returns x-y. Requires that x-y does not underflow an int64.
func subU(x int64, y uint64) int64 {
	if y >= 1<<63 {
		if x < 0 {
			base.Fatalf("subU underflowed %d - %d", x, y)
		}
		x -= 1<<63 - 1
		x -= 1
		y -= 1 << 63
	}
	if subWillUnderflow(x, int64(y)) {
		base.Fatalf("subU underflowed %d - %d", x, y)
	}
	return x - int64(y)
}

	"REM",
	"REMU",
	"RFE",
	"ROTR",
	"ROTRV",
	"SC",
	"SCV",
	"SEB",
	"SEH",
	"SGT",
	"SGTU",
	"SLL",
	"SQRTD",
	"SQRTF",
	"SRA",
	"SRL",
	"SUB",
	"SUBD",
	"SUBF",
	"SUBU",
	"SUBW",
	"SYNC",
	"SYSCALL",
	"TEQ",
	"TLBP",
	"TLBR",
	"TLBWI",
	"TLBWR",
	"TNE",
	"WORD",
	"WSBH",
	"XOR",
	"MOVV",
	"MOVVL",
	"MOVVR",
	"SLLV",
	"SRAV",

//		outcode(int($1), &$2, int($4), &$6);
//	}
	SUB	R6, R26, R27	// 0346d822
	SUBU	R6, R26, R27	// 0346d823
	SUBV	R16, R17, R26	// 0230d02e
	SUBVU	R16, R17, R26	// 0230d02f

//	LSUBW imm ',' sreg ',' rreg
//	{
//		outcode(int($1), &$2, int($4), &$6);
//	}
	SUB	$-3126, R17, R22	// 22360c36
	SUB	$3126, R17, R22		// 2236f3ca
	SUBU	$16384, R17, R12	// 262cc000
	SUBV	$-6122, R10, R9		// 614917ea

	"NOOP",
	"NOR",
	"OR",
	"REM",
	"REMU",
	"RFE",
	"SC",
	"SCV",
	"SGT",
	"SGTU",
	"SLL",
	"SQRTD",
	"SQRTF",
	"SRA",
	"SRL",
	"ROTR",
	"SUB",
	"SUBD",
	"SUBF",
	"SUBU",
	"SUBW",
	"DBAR",
	"SYSCALL",
	"TEQ",
	"TNE",
	"WORD",
	"XOR",
	"MASKEQZ",
	"MASKNEZ",
	"MOVV",
	"MOVVL",
	"MOVVR",
	"SLLV",
	"SRAV",
	"SRLV",
	"ROTRV",
	"DIVV",