On page 36 of the printed manual, I refer to files EFF86.DOC
and EFF386.DOC, containing opcode mappings for effective
address bytes.  I have renamed these files from the .DOC
extension to the .TXT extension, because they are in fact
plain text files.

On page 46 of the printed manual, there are three PREFETCH
instructions with incorrect mnemonics: there should be an
extra T before the ending digits 0,1,2: i.e., the correct
mnemonics are PREFETCHT0, PREFETCHT1, PREFETCHT2.

On page 45 of the printed manual, the line MOVZX rw,seg
should instead be MOVZX rd,seg .  This is an undocumented
feature of the 386-or-later processor, in which MOV rw,seg
instruction with an O4 override will move the segreg value
into the 32-bit register, with zero-extend.  The O4
override is ignored if the segreg MOV has a memory operand
as its destination.

Chapter 16 of the A86 printed manual, there are numerous
symbols listed as "P6 Instruction".  These are instruction
mnemonics added with the Pentium Pro but not in the Pentium
MMX, and retained in the Pentium 2 and beyond.  I now
refer to these as P2 instructions.

I have added a .DOSSEG command which is ignored in COM mode,
and produces a comment-record in OBJ mode, identical to the
one that MASM produces.  This will presumably tell the
Microsoft linker to arrange program segments in a particular
order.

Also for compatibility, I have relaxed the syntax of the
SEGMENT directive so that a preceding segment name is not
required.  If a segment name is not given, the defualt outer
segment is used.  In COM mode, this is the CODE SEGMENT;
in OBJ mode, this is a segment named _TEXT or modname_TEXT,
as described in Chapter 10.  This new syntax form is used
to implement the .CODE macro, given in the file COMPAT.8.