The SAS 370 C Compiler

[Carl Zeigler]

        This message is a followup to Richard Stevens, Peter DeSilva, and
    Eric Browns' postings on the SAS C compiler.   ( article nos. 243 at hsi.uucp,
    252 at kitty.UUCP, and 3432 at utah-cs.UUCP )

        First a brief intro:  I work for SAS on the IBM-PC product.  While
    I do not work on the 370 Compiler per se, I have followed its development
    very closely.  It is derived from the Lattice compiler under an agreement
    with Lattice.  We also have a cross-compiler and linker under a similar
    arrangement.   
    
    What follows was written by the manager of the 370 C complier project.
    ------------

        The major point is that the 370 C Compiler is not the same as the
    Lattice PC compiler.   Many PC limitations have been removed and many
    features added for the 370.

        1.   3 minor fixes:  Macro definitions and invocations can be spread
    over mutiple source lines.  Unsigned char and unsigned long are implemented.
    The default is non-nesting comments.

        2.  We are aware of the printf and scanf limitations.   We plan
    to enhance these functions as a high priority if Lattice dosen't do it
    first.

        3.  We will do a DOS/VSE version when there is some evidence of
    demand for one.  More likely to be useful is a DOS/VSE runtime library
    ( so you can compile on OS or CMS and ship the object to a DOS/VSE site)
    and we will be doing that later this year.

        4.  Whitesmiths may well be "thinking about CICS".  So are we, and
    we will do a CICS runtime library later this year or early next.  In the
    meantime our general operating system interface, something no other 370
    compiler provides, will allow the user to write a small number of CICS
    routines that the library will call to do all the operating system related
    functions ( get memory, etc ) for execution under CICS.


        Also worth pointing out is that we originally planned to use Whitesmiths
    on the 370 to develop our own products.   We installed it, it was inadequate
    for our needs so we decided to produce our own.  Thus was born the Lattice
    deal.  And finally, how much 370 expertise does Whitesmith's have?  Do they
    have it in house at all ?   [ jcz - Actually they contracted the porting out
    to another firm. ] Contrast this with over 10,000 of our products installed
    on IBM mainframes, a support organization, etc.

    ---------              

    Some comments of my own:

        There are some features that weren't mentioned above such as
    non-constant external variables and re-entrant code ( guess how ).
    A complete UNIX I/O model that allows the development of interfaces to
    any access method transparent to the application.   We translate
    floating constants to their internal form with four bits more
    precision than Whitesmiths.   ( Our product viability is VERY
    dependent on acurrate math. )   As far as support goes, we have perhaps
    the best documentation and support facilities of anyone in the software
    industry, let alone other 370 C compiler vendors.   ( ahem, I will try
    to stay calm.)    The population of our installed product base was mentioned
    above,  it translates  to 350 man years of experience.  We also have about
    80  man years experience in C and UNIX in house.

        Baiscally, one should not review a book one has not read.

	John Carl Zeigler		SAS Institutte Inc.
	...!mcnc!ncsu!jcz		Cary, NC 27511
					( 919 ) 476 - 5322