johnl@ima.UUCP (04/09/88)
I have heard some not too great comments about the commercially
available C compilers on IBM 370s. I would like to find out
the truth regarding this matter. So, if anyone has any experience,
comments, suggestions, I would like to hear them. Also, any
recommendations for good C compilers. Thanks.
Yogesh Gupta | If you think my company will let me
Cullinet Software, Inc. | speak for them, you must be joking.
Ph: (408) 434-6636
[From decvax!decwrl!ames!cullsj!gupta (Yogesh Gupta)]
--
Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU
Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima
Please send responses to the originator of the message -- I cannot forward
mail accidentally sent back to compilers. Meta-mail to ima!compilers-requestam@computer-lab.cambridge.ac.uk (Alan Mycroft) (04/30/88)
In article <941@ima.ISC.COM> you write: >I have heard some not too great comments about the commercially >available C compilers on IBM 370s. I would like to find out >the truth regarding this matter. So, if anyone has any experience, >comments, suggestions, I would like to hear them. Also, any >recommendations for good C compilers. Thanks. >Yogesh Gupta We [a couple of academics wearing startup-company hats] wrote a re-targetable, optimising ANSI (draft) C compiler, oriented at the many new risc machines (e.g. register colouring etc). Somewhat as a joke, but also as a donation to the university, we re-targetted it to our ibm 3084 running MVS/XA (however, all the library except a public domain I/O library and 150 lines of assembler is written in C). The compiler produces either or both ibm object modules or assemblable listing. Much to our amusement we got dhrystone figures over 23K with a lame procedure calling standard, and can compile the 0.6mb of source in less than 1 minute cpu. Let me know if you are further interested, there follows the usual sort of advert we put out on getting requests: \documentstyle[a4,12pt]{article} \begin{verbatim} NorCroft C ---------- NorCroft is a trading name of a startup company (Codemist Ltd) formed by three academics: Prof J.P. Fitch, Dept of Computer Science, Bath University, UK Dr. A. Mycroft, Computer Lab, Cambridge University, UK Email: am@cl.cam.ac.uk, phone +44 223 334621 (office), +44 223 871088 (home 1800-2200 GMT) Dr. A.C. Norman, Computer Lab, Cambridge University, UK Email: acn1@phx.cam.ac.uk, phone +44 223 338550 (office), +44 223 247367 (home 1800-2200 GMT) FAX: +44 223 334748 (marked attention Dr. Mycroft/Norman Computer Lab). The NorCroft C compiler is a significantly optimising re-targetable C compiler to the ANSI draft standard. It includes many 'lint'-like features and warnings for common errors (e.g. printf format/argument checks, '=' and '==' confused, /* inside comment, etc.). The optimiser includes register allocation by colouring, loop invariant optimisation and tail-recursion removal. Rapid re-targetability is encouraged by the use of a machine (parameterised, but otherwise) independent code which easily translates into many register-oriented target architectures including many RISC machines. Ports to CISC machines (e.g. NS32000) use peephole optimisation to map a load-store architecture onto all of the CISC addressing modes while preserving the ability to freely allocate temporaries and variables into machine registers. Standard NorCroft compilers can produce both assembler- and object-form output. The compiler itself is written in C and includes many self-checks. NorCroft has holds and/or has licensed the machine independent parts to various organisations for back-ends for Acorn ARM, Inmos Transputer, IBM 370, AMD 29000, Clipper, Motorola 68000-series, National Semiconductor 32000-series and various proprietory CPU's. A port (re-target) generally takes several weeks to the compiler being able to bootstrap itself (of course exploiting every machine idiom takes longer). In addition a small minority of our clients have been involved in producing stack-oriented machine code directly from the optimised parse tree instead of using the general register-oriented optimiser. Various rights are for sale, including source code. \end{verbatim} \end{document} [From Alan Mycroft <am@computer-lab.cambridge.ac.uk>] -- Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU Plausible paths are { ihnp4 | decvax | cbosgd | harvard | yale | bbn}!ima Please send responses to the originator of the message -- I cannot forward mail accidentally sent back to compilers. Meta-mail to ima!compilers-request