Xref: utzoo comp.sys.ibm.pc:18429 comp.sources.wanted:4892 comp.sys.intel:507
Path: utzoo!attcan!uunet!mcvax!unido!uklirb!kirchner
From: kirchner@uklirb.UUCP (Reinhard Kirchner)
Newsgroups: comp.sys.ibm.pc,comp.sources.wanted,comp.sys.intel
Subject: Re: Looking for Z80 to 8086 translator
Message-ID: <2658@uklirb.UUCP>
Date: 26 Aug 88 14:31:25 GMT
References: <135@ambone.UUCP>
Organization: University of Kaiserslautern, W-Germany
Lines: 35

From article <135@ambone.UUCP>, by leif@ambone.UUCP (Leif Andrew Rump):
> hartvig@freja.dk (Hartvig Ekner) writes:
> 
> It should be quite easy because the original 8088 was made to look
> like 8080 meaning the source-assembler-format is the same so you
> actually only need to substitude text to convert it! (Intel has some
> papers that explain how to do it). The Z80 is a superset of the 8080
> as 8088 is a superset of 8080, so if you can get your code from Z80 to
> 8080 using the same method and from 8080 to 8088 your problems is
> solved! Except for on thing! Z80 has more commands than 8080 and 8080
> has some special features that isn't included in 8088 but the most of
> the job should be faily easy... :-)
> 
The thing is not as easy as it looks. Years ago ( >=5 ) we got such a
translator from intel to run on CP/M. With it came a handbook which
showed how the special z80-opcodes are translated. We found out
that most of them resulted in rather terrible 8086-code. This
comes partly from the different use of the flags, so the 8086-code
contained lots of
     save flags
     emulate 780-instr
     load flags
     save flags
     emulate z80-instr
     load flags
        .
        .
The code was mostly correct, but very long and slow, so we had to rework
it to make it usable.

BTW: The code we translated was interpreter, runtime-system and arithmetic
routines of PASCAL-SC, which is now available for the IBM-PC from
Whiley in US and Teubner in Europe. ( Very modest price -:) )

R. Kirchner  ( speaking for himself, what else )