MOVL	(0*4)(SI),	DI
	MOVL	DX,		BP

	ROUND1(AX,BX,CX,DX, 1,0xd76aa478, 7);
	ROUND1(DX,AX,BX,CX, 2,0xe8c7b756,12);
	ROUND1(CX,DX,AX,BX, 3,0x242070db,17);
	ROUND1(BX,CX,DX,AX, 4,0xc1bdceee,22);
	ROUND1(AX,BX,CX,DX, 5,0xf57c0faf, 7);
	ROUND1(DX,AX,BX,CX, 6,0x4787c62a,12);
	ROUND1(CX,DX,AX,BX, 7,0xa8304613,17);
	ROUND1(BX,CX,DX,AX, 8,0xfd469501,22);
	ROUND1(AX,BX,CX,DX, 9,0x698098d8, 7);
	ROUND1(DX,AX,BX,CX,10,0x8b44f7af,12);

	ROUND1(R7,R4,R5,R6, 6,0x4787c62a,12);
	ROUND1(R6,R7,R4,R5, 7,0xa8304613,17);
	ROUND1(R5,R6,R7,R4, 8,0xfd469501,22);
	ROUND1(R4,R5,R6,R7, 9,0x698098d8, 7);
	ROUND1(R7,R4,R5,R6,10,0x8b44f7af,12);
	ROUND1(R6,R7,R4,R5,11,0xffff5bb1,17);
	ROUND1(R5,R6,R7,R4,12,0x895cd7be,22);
	ROUND1(R4,R5,R6,R7,13,0x6b901122, 7);
	ROUND1(R7,R4,R5,R6,14,0xfd987193,12);
	ROUND1(R6,R7,R4,R5,15,0xa679438e,17);
	ROUND1(R5,R6,R7,R4, 0,0x49b40821,22);

	MOVM.IB [Ra,Rb,Rc,Rd,Re], (R13)

	MOVW	$w_buf, Rw
	MOVW	$0x5A827999, Rconst
	MOVW	$3, Rctr
loop1:	ROUND1(Ra, Rb, Rc, Rd, Re)
	ROUND1(Re, Ra, Rb, Rc, Rd)
	ROUND1(Rd, Re, Ra, Rb, Rc)
	ROUND1(Rc, Rd, Re, Ra, Rb)
	ROUND1(Rb, Rc, Rd, Re, Ra)
	SUB.S	$1, Rctr
	BNE	loop1

	ROUND1(Ra, Rb, Rc, Rd, Re)
	ROUND1x(Re, Ra, Rb, Rc, Rd)
	ROUND1x(Rd, Re, Ra, Rb, Rc)
	ROUND1x(Rc, Rd, Re, Ra, Rb)

	ROUND1(R5,R2,R3,R4, 6,0x4787c62a,12);
	ROUND1(R4,R5,R2,R3, 7,0xa8304613,17);
	ROUND1(R3,R4,R5,R2, 8,0xfd469501,22);
	ROUND1(R2,R3,R4,R5, 9,0x698098d8, 7);
	ROUND1(R5,R2,R3,R4,10,0x8b44f7af,12);
	ROUND1(R4,R5,R2,R3,11,0xffff5bb1,17);
	ROUND1(R3,R4,R5,R2,12,0x895cd7be,22);
	ROUND1(R2,R3,R4,R5,13,0x6b901122, 7);
	ROUND1(R5,R2,R3,R4,14,0xfd987193,12);
	ROUND1(R4,R5,R2,R3,15,0xa679438e,17);
	ROUND1(R3,R4,R5,R2, 0,0x49b40821,22);

	ROUND1(Rb, Rc, Rd, Ra,  3, 22, Rc3)

	MOVM.IA.W (Rtable), [Rc0,Rc1,Rc2,Rc3]
	ROUND1(Ra, Rb, Rc, Rd,  4,	7, Rc0)
	ROUND1(Rd, Ra, Rb, Rc,  5, 12, Rc1)
	ROUND1(Rc, Rd, Ra, Rb,  6, 17, Rc2)
	ROUND1(Rb, Rc, Rd, Ra,  7, 22, Rc3)

	MOVM.IA.W (Rtable), [Rc0,Rc1,Rc2,Rc3]
	ROUND1(Ra, Rb, Rc, Rd,  8,	7, Rc0)
	ROUND1(Rd, Ra, Rb, Rc,  9, 12, Rc1)
	ROUND1(Rc, Rd, Ra, Rb, 10, 17, Rc2)
	ROUND1(Rb, Rc, Rd, Ra, 11, 22, Rc3)

	ENDIAN_MOVE(12,R6,M03,M15)

	ROUND1(R22,R3,R4,R5,M00,0xd76aa478, 7);
	ROUND1(R5,R22,R3,R4,M01,0xe8c7b756,12);
	ROUND1(R4,R5,R22,R3,M02,0x242070db,17);
	ROUND1(R3,R4,R5,R22,M03,0xc1bdceee,22);

	ENDIAN_MOVE(16,R6,M04,M15)
	ENDIAN_MOVE(20,R6,M05,M15)
	ENDIAN_MOVE(24,R6,M06,M15)
	ENDIAN_MOVE(28,R6,M07,M15)

	ROUND1(R22,R3,R4,R5,M04,0xf57c0faf, 7);
	ROUND1(R5,R22,R3,R4,M05,0x4787c62a,12);

	ROUND1(DX,AX,BX,CX, 6,0x4787c62a,12);
	ROUND1(CX,DX,AX,BX, 7,0xa8304613,17);
	ROUND1(BX,CX,DX,AX, 8,0xfd469501,22);
	ROUND1(AX,BX,CX,DX, 9,0x698098d8, 7);
	ROUND1(DX,AX,BX,CX,10,0x8b44f7af,12);
	ROUND1(CX,DX,AX,BX,11,0xffff5bb1,17);
	ROUND1(BX,CX,DX,AX,12,0x895cd7be,22);
	ROUND1(AX,BX,CX,DX,13,0x6b901122, 7);
	ROUND1(DX,AX,BX,CX,14,0xfd987193,12);
	ROUND1(CX,DX,AX,BX,15,0xa679438e,17);
	ROUND1(BX,CX,DX,AX, 1,0x49b40821,22);

	MOVL	DX,	76(SP)
	MOVL	BP,	80(SP)

	ROUND1(AX, BX, CX, DX, BP, 0)
	ROUND1(BP, AX, BX, CX, DX, 1)
	ROUND1(DX, BP, AX, BX, CX, 2)
	ROUND1(CX, DX, BP, AX, BX, 3)
	ROUND1(BX, CX, DX, BP, AX, 4)
	ROUND1(AX, BX, CX, DX, BP, 5)
	ROUND1(BP, AX, BX, CX, DX, 6)
	ROUND1(DX, BP, AX, BX, CX, 7)
	ROUND1(CX, DX, BP, AX, BX, 8)
	ROUND1(BX, CX, DX, BP, AX, 9)
	ROUND1(AX, BX, CX, DX, BP, 10)
	ROUND1(BP, AX, BX, CX, DX, 11)

	MOVL	CX,	R13
	MOVL	DX,	R14
	MOVL	BP,	R15

	ROUND1(AX, BX, CX, DX, BP, 0)
	ROUND1(BP, AX, BX, CX, DX, 1)
	ROUND1(DX, BP, AX, BX, CX, 2)
	ROUND1(CX, DX, BP, AX, BX, 3)
	ROUND1(BX, CX, DX, BP, AX, 4)
	ROUND1(AX, BX, CX, DX, BP, 5)
	ROUND1(BP, AX, BX, CX, DX, 6)
	ROUND1(DX, BP, AX, BX, CX, 7)
	ROUND1(CX, DX, BP, AX, BX, 8)
	ROUND1(BX, CX, DX, BP, AX, 9)
	ROUND1(AX, BX, CX, DX, BP, 10)
	ROUND1(BP, AX, BX, CX, DX, 11)