WSR5672@TNTECH.BITNET (11/15/89)
Does anyone know of a good public domain Ada compiler?
I need something would run on UNIX (Sun workstations) or ULTRIX (Vax 750).
Has GNU made any plans to produce one?
Scott Redmon
Tenn. Tech. Universitykarl@grebyn.com (Karl Nyberg) (11/16/89)
Date: Wed, 15 Nov 89 09:53 CST From: <haven!VMA.CC.CMU.EDU!WSR5672%TNTECH.BITNET> Does anyone know of a good public domain Ada compiler? Depends on what you mean by GOOD... I understand there's some stuff from Florida State that may be available for reproduction costs (Ted - is this true?). Also, I think certain flavors of the AL/S may be licensed for government use (and certain non-distribution use) in a way that is pretty cheap. I need something would run on UNIX (Sun workstations) or ULTRIX (Vax 750). I think the FSU stuff is on the Sun, while the ALS stuff is VAX/VMS based. Has GNU made any plans to produce one? I asked about this a year or so ago. I seem to recall the response was along the lines that there was more important stuff to be done, namely the operating system, with the limited resources that they had. It was in the queue of things to be done, but they had no schedule or people interested in doing the work. I think the effort is primarily a matter of taking the existing C (and C++) compilers, the Ada grammar, adding some semantic analysis (non-trivial), probably having to rework some of the intermediate language, building specialized runtimes for Ada, and possibly tweaking the code generator(s). Oh yeah, the optimizer. Might actually prove to be a useful collection of projects for a term course in compiler writing. Many compiler vendors also provide significant educational discounts. I seem to recall something on the range of 70-90% off for some vendors. -- Karl --
hudgens@ray.met.fsu.edu (Jim Hudgens) (11/18/89)
I am posting this for Dr. Baker. Please send any responses to him directly, and not to me. -------------------------------------------------------------- This is a response to a recent message from Karl Nyberg on "comp.lang.ada", which apparently was itself in response to a person looking for a public-domain Ada compiler. Since my name and a compiler I worked on here at FSU was mentioned, I feel compelled to fill in some information, though I'm afraid it may not help the solve the original problem much. Below, Karl's remarks are indicated by ">", and parts of what appears to be the original query are indicated by ">>". > Depends on what you mean by GOOD... I understand there's some > stuff from Florida State that may be available for reproduction > costs (Ted - is this true?). I have made copies of the FSU Ada compiler available to people who want to use it for research or education (not for profit), without royalty, but it is not "public domain". FSU (and the authors) claim copyright to this material. (The situation is pretty much similar to UNIX source code.) The compiler source code and reports are also available to government contractors through the NTIS, I believe. I don't claim the FSU compiler is GOOD, though I would rather Karl's had not made the remark above. It was a research and educational project that started in 1980, on which Greg Riccardi and I worked part-time with some student assistants. It suffered from kludges to handle massive language changes in the 1982-84 period, after major compiler design decisions had been made. If I had known what we were in for when we started, I would probably not have done it. We did stick it through, however, and validated the compiler in 1985. As far as I know, our compiler was the first serious effort to produce a cross-compiler for a small embedded computer architecture. The hardest part of doing the compiler was getting it to meet the nit-picking semantic requirements of validation. The second-hardest part was doing cross-development and repeatedly running thousands of tests on the single-board computers (bare 16-bit machines with 48K of RAM), via 1200 baud modems. Another major problem was that we had to start completely from scratch. (That is, we had to develop our own Pascal-based parser-generator and code-generator tools, and our own bare-machine runtime environment.) I suspect the FSU compiler is not likely to be of much use to the person who originated this query, since it is a cross compiler for a very uncommon host/target pair-- CDC Cyber/Zilog Z8002. Moreover, it has not been maintained since it was validated in 1985. There have been significant changes in the language since then, and I am sure it is far from passing the current validation tests. I and some students have ported portions of the compiler, which was written in ETH Pascal 6000 (with peculiarities due to 60-bit words, 6-bit characters, and memory size limitations) so that it will compile and run under Sun/UNIX with the Berkely "pc" compiler, but it is not completely ported, and has not been tested much. We have used the front-end successfully here at FSU for several instructional and research projects. Most recently, one Ph.D. student here (Ted Giering) is now using it to do source-to-source translation of Ada programs with restricted task structures into Ada code modules that can be scheduled deterministically. A M.S. student is currently working on a 68020 code generator. In case anyone is interested in working with the compiler, here is a brief sketch of it: The parser and final code generator make use of (ambiguous) S-attributed grammars, and a set of Pascal LALR(1) parser and attribute-processor generating tools I cooked up locally. The final code generation is from a fairly machine-independent form of prefix code, based on the work of Ganapathi and Fischer. The semantic analysis is written in Pascal, in a LISP-like style making heavy use of linked lists and high-level operators (with functions and procedures as parameters). The intermediate structures used in semantic analysis are software-paged in and out from a direct-access disk file, and manipulated locally via Pascal pointers. The "middle" of the compiler traverses the graph-like linked structures built by the semantic analysis phase and translates them to prefix-code. The runtime system is written in Z8002 assembly code, based on a design originally coded in Ada. > Also, I think certain flavors of the AL/S may be licensed for > government use (and certain non-distribution use) in a way that is > pretty cheap. Yes, if you are a government contractor (and most universities are), you ought to be able to get the original AL/S, or maybe one of the improved variants. It is reputed to be a great time and space hog, but the code is written in Ada and reputed to be readable and modifiable. (I have not looked at it myself, so I don't know these things for certain.) >> I need something would run on UNIX (Sun workstations) or ULTRIX >> (Vax 750). > I think the FSU stuff is on the Sun, while the ALS stuff is > VAX/VMS based. What I recall about the AL/S agrees with Karl. I don't think the AL/S is available in a UNIX/ULTRIX hosted version, but again I don't know for certain. The FSU compiler that was validated ran under CDC NOS, but as I have explained above we have ported parts of it to the Sun. I think I remember hearing that DARPA is funding David Fisher's company, Incremental Systems, in Pittsburgh to do something they may conceive of as a "public domain" Ada compier or front-end. I don't know any of the details. >> Has GNU made any plans to produce one? > I asked about this a year or so ago. I seem to recall the > response was along the lines that there was more important stuff > to be done, namely the operating system, with the limited > resources that they had. It was in the queue of things to be > done, but they had no schedule or people interested in doing the > work. I think the effort is primarily a matter of taking the > existing C (and C++) compilers, the Ada grammar, adding some > semantic analysis (non-trivial), probably having to rework some of > the intermediate language, building specialized runtimes for Ada, > and possibly tweaking the code generator(s). Oh yeah, the > optimizer. Might actually prove to be a useful collection of > projects for a term course in compiler writing. I am sorry to hear that GNU is not interested in doing one. I think a GOOD public-domain Ada compiler would be a big boost to Ada, encouraging both use of Ada and development of Ada software tools. One of the big problems with Ada for software tools is the overloading. One can't even do an accurate cross-reference listing without doing overload resolution. The separate compilation and other complelx visibility rules make things even worse. In short, you can't do decent tools that process Ada source code without repeating most of what is in the front end of an Ada compiler. I have been interested in helping out with such a public-domain compiler for some time, but don't know where the funding would come from. Using the Gnu-C code generator and tools may be a good idea, but I have some reservations. To be really useful, the compiler should be written in Ada. Moreover, from what I learned and what I have heard from others who attempted to use existing code generators for Ada, it would require some changes to the optimizer, code generator, and intermediate language -- hence, ideally, cooperation with those maintaining Gnu-C. The semantic analysis is definitely "non-trivial", if you want to be able to validate and also want it to run in a reasonable amount of time. This is closely connected with the design of the semantic database you need for separate compilation. In short, it is a big job if you want to produce a good quality result. Based on my experience teaching first and second-term compiler courses, I don't think an Ada compiler produced as a course project would be any good. I'd rather bet on a couple of experienced people working full-time for about a year. > Many compiler vendors also provide significant educational > discounts. I seem to recall something on the range of 70-90% off > for some vendors. Yes, I have seen educational and quantity commercial discounts similar to those Karl reports. Site licenses are also available. We have received very reasonable prices from those vendors we have contacted (Verdix, RR, Meridian), and I expect other vendors are competitive. In fact, it is clearly more effective for us here at FSU to buy our compilers from commercial vendors and concentrate our time on research, than to port and maintain our own compiler. --Ted Baker Prof. Ted Baker, B-173 Dept. of Computer Science Florida State University Tallahassee, FL 32306-4019 (904) 644-5452 tbaker@ajpo.sei.cmu.edu baker@cs.fsu.edu ------------------------------------------------------------------- -- Jim Hudgens -- Nothing to disclaim. Dept of Meteorology, Florida State Univ. hudgens@ray.met.fsu.edu