@H(Differences between IMP80 and IOPT



The following is a collection of differences noted between the currently
supported IMP compiler on EMAS, IMP80, and the portable compiler, currently
known as IOPT.  The latter will be the only IMP compiler available on EMAS 3.

N.B. Perhaps users of the old IMP9 compiler (there are probably still quite
a few) should be warned in good time that IMP9 will not be available on EMAS 3.


 1) In IMP80 access to a byte beyond the currently defined length of a string
    by use of CHARNO is permitted.  This is not true in IOPT.  However,
    in IOPT the length of a string constant or ~constant~ ~string~ can be found
    (but not changed!) by use of LENGTH(str), and the form   CHARNO(str,n),
    where str is a string constant or ~constant~ ~string~ and n is an integer
    constant or ~constant~ ~integer~, is also permitted on the right hand side
    of an expression.  This is not true of IMP80.

 2) The keyword ~system~ is permitted in IOPT, but has a different meaning.
    At present, use of ~system~ in a ~spec~ statement produces a warning message,
    but the correct (i.e. IMP80-meaning) code is generated (the entry point is
    S#procname, where procname is the name, in upper case, of the procedure
    specified).  It is advisable for users to remove all occurrences of the
    keyword ~system~ from their programs, by use of ~alias~; for example:

                   ~system~ ~routine~ ~spec~ itos(~integer~ n)
    should become
                   ~external~ ~routine~ ~spec~ itos ~alias~ "S#ITOS"(~integer~ n)

    Note that the entry point name (the part in double quotes) MUST be given
    in upper case, whatever case is used elsewhere in the program.

    Programs in which system procedures are defined (rather than merely
    specified) MUST be converted using ~alias~, as above.  This is because such
    use of the keyword ~system~ is valid in IOPT, but does not mean the same as
    in IMP80.  No warning message is given (perhaps it should be?), and the
    first indication of trouble occurs when entry points are found to be
    missing.

    The best advice is to get rid of the keyword ~system~ from all IMP programs.

 3) 2900 assembler code is handled differently in some cases.  Noted thus far:

    a) string descriptors are not loaded into the descriptor register by the
       instructions *LD_S, *LSD_S.  Assembler programmers who need to use
       string descriptors can construct them in (locally declared) ~long~
       ~integer~s, using ADDR.  They will then operate correctly in IMP80 and
       in IOPT.

 4) In IMP80, the array map ARRAY can take the name of an existing array as
    its second parameter, the format of that array then being used by ARRAY.
    This is not permitted in IOPT: the second parameter of ARRAY must be an
    explicitly declared ~array~ ~format~.

@P()
 5) There is no unassigned checking in IOPT on EMAS 2900, although there will
    be on EMAS 3.

 6) The main advantage of IOPT on EMAS 2900 is that it produces faster programs
    than IMP80 when PARM(OPT) is used, although it takes approximately 2.25
    times as long to compile a program.  The background compilation command
    BIOPT, corresponding to BIMP80, is available.  It is recommended that IMP80
    be used when developing a program, but then when optimising IOPT be used to
    get the speed benefit, and to verify that the program will run on EMAS 3.


                                                                 John Murison
                                                                     May 1984