SUBQ    $128, inl
	LEAQ    128(inp), inp

	MOVQ $8, itr1
	MOVQ $2, itr2

	CMPQ inl, $128
	JBE  sealAVX2Tail128
	CMPQ inl, $256
	JBE  sealAVX2Tail256
	CMPQ inl, $384
	JBE  sealAVX2Tail384
	CMPQ inl, $512
	JBE  sealAVX2Tail512

	// We have 448 bytes to hash, but main loop hashes 512 bytes at a time - perform some rounds, before the main loop

		// 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).