// add current hash values with previously saved
	PADDD		abefSave, state0
	PADDD		cdghSave, state1

	// advance data pointer; loop until buffer empty
	ADDQ		$64, dataPtr
	CMPQ		numBytes, dataPtr
	JNE		roundLoop

	// write hash values back in the correct order
	PSHUFD		$0x1b, state0, state0		// FEBA
	PSHUFD		$0xb1, state1, state1		// DCHG
	VMOVDQA		state0, m4
	PBLENDW		$0xf0, state1, state0		// DCBA

		// TODO: have these return flags instead of bool.  The current system generates:
		//    CMPXCHGQ ...
		//    SETEQ AX
		//    CMPB  AX, $0
		//    JNE ...
		// instead of just
		//    CMPXCHGQ ...
		//    JEQ ...
		// but we can't do that because memory-using ops can't generate flags yet
		// (flagalloc wants to move flag-generating instructions around).

	JCS 2(PC)
	//TODO: LJMPQ* (BX)                     // 48ff2b
	JCS 2(PC)
	//TODO: LJMPQ* (R11)                    // 49ff2b
	//TODO: JNE .+$0x11223344               // 480f8544332211 or 0f8544332211
	//TODO: JNE .+$0x11                     // 7511
	//TODO: JNO .+$0x11223344               // 480f8144332211 or 0f8144332211
	//TODO: JNO .+$0x11                     // 7111