1. Forum
  2. Usenet
  3. COMP.LANG.MODULA2

  • Borland Turbo Modula-2 M-Code?

    From Josef Moellers@21:1/5 to All on Mon Nov 26 10:46:17 2018
    Hi,

    I have recently revived an old SB180FX computer (from Steve Ciarcia's
    BYTE articles). Among other things, I have a Borland Modula-2 compiler
    and I was curious what the MCD files look inside.
    Does anyone have any information on this?

    So far, I have found out that
    8dH is a LIB (Load Immediate Byte)
    8eH is a LIW (Load Immediate Word)
    8fH is a LID (Load Immediate Double)
    90H-9fH are LIn (Load Immediate n, where n is a 4-bit quantity)
    FnH seems to be some function call

    Does anyone have more information?

    Josef

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Burrows@21:1/5 to All on Mon Nov 26 04:59:10 2018
    The Turbo Modula-2 manual says that the M-Code used is a variant of the M-Code used by Lilith. If you haven't already got it, you can download ETH Report Nr 40 - The Personal Computer Lilith. Apr 1981. N. Wirth "The Yellow Report" which includes the M-
    Code instruction set and the source code of a sample M-Code interpreter from my site:

    http://www.cfbsoftware.com/modula2

    The Z80 M-Code interpreter is located at the bottom of the TPA at Address 100H and is about 4.5K bytes. You could try disassembling that and comparing the result with the source code in the report perhaps?

    Chris Burrows
    CFB Software

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Josef Moellers@21:1/5 to Chris Burrows on Mon Nov 26 14:21:43 2018
    Thanks.

    On 26.11.18 13:59, Chris Burrows wrote:
    The Turbo Modula-2 manual says that the M-Code used is a variant of the M-Code used by Lilith. If you haven't already got it, you can download ETH Report Nr 40 - The Personal Computer Lilith. Apr 1981. N. Wirth "The Yellow Report" which includes the M-
    Code instruction set and the source code of a sample M-Code interpreter from my site:

    http://www.cfbsoftware.com/modula2

    Yes, I've seen that, but the numbers do not seem to match.

    The Z80 M-Code interpreter is located at the bottom of the TPA at Address 100H and is about 4.5K bytes. You could try disassembling that and comparing the result with the source code in the report perhaps?

    I'm good at disassembling ;-)
    I assume that most of the compiler itself is running on top of the mcode virtual machine, judging from these filenames: COMPILE.MCD, GENZ80.MCD
    I'll see what I can find.

    Thanks again.

    Josef

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Chris Burrows@21:1/5 to Josef Moellers on Mon Nov 26 12:59:17 2018
    On Monday, November 26, 2018 at 11:51:44 PM UTC+10:30, Josef Moellers wrote:
    I assume that most of the compiler itself is running on top of the mcode virtual machine, judging from these filenames: COMPILE.MCD, GENZ80.MCD
    I'll see what I can find.


    That is correct. COMPILE is the compiler and GENZ80 is the native code generator. Refer to Page 258 in the manual, 'Modules in Memory Management' for more information. If you don't already have a copy you can download it from:

    www.cpcwiki.eu/imgs/6/64/Borland_Turbo_Modula_2.pdf

    I'd be interested to hear about any progress that you make.

    Chris.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Josef Moellers@21:1/5 to Chris Burrows on Wed Nov 28 10:14:53 2018
    On 26.11.18 21:59, Chris Burrows wrote:
    On Monday, November 26, 2018 at 11:51:44 PM UTC+10:30, Josef Moellers wrote:
    I assume that most of the compiler itself is running on top of the mcode
    virtual machine, judging from these filenames: COMPILE.MCD, GENZ80.MCD
    I'll see what I can find.


    That is correct. COMPILE is the compiler and GENZ80 is the native code generator. Refer to Page 258 in the manual, 'Modules in Memory Management' for more information. If you don't already have a copy you can download it from:

    www.cpcwiki.eu/imgs/6/64/Borland_Turbo_Modula_2.pdf

    I'd be interested to hear about any progress that you make.

    1st "Erfolgserlebnis" (sense of achievement):

    1) The first 512 bytes in the m2.com are a jump table, unusually coded
    as first 256 low bytes then 256 high bytes.
    2) while most instructions have no arguments, some instructions do 'ave 'em.
    3) DE is the m-code program-counter.
    4) m2.com is mainly running m-code.

    Having found the jump-table, I can now see what each instruction does.

    NB I've used a combination of passive and active analysis:
    A) a (mostly) dis-assembled m2.com (using a self-written disassembler
    which follows jumps and calls and knows a dead-end when it sees one ;-)
    and also allows me to specify entry-points for those cases where the
    address is calculated.
    B) Using Perl/Expect and a small helper to send/receive through the
    serial line, I used a debugger to step through m2.com on my SB180FX.
    This showed a "JP (HL)" at more or less regular intervals, the value of
    the accumulator was the currently decoded instruction.

    Josef

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)