bailey@MIST.CS.ORST.EDU (Kirk Bailey) (07/03/88)
In the recent posting from Hubert Matthews I noticed our "C" was missing from
his existance list... Must be time to let folks know what we've been up to:
***BEGIN ADVERTISMENT***
Logical Systems sells a standalone "C" development system which includes a ANSI
style "C" compiler, assembler, linker, librarian, single processor and network
loaders, etc. All tools are provided in "C" source code form (binaries are
also provided for the IBM PC/B004 combo). The "toolset" is CURRENTLY cross-
hosted on a variety of platforms including the APPLE MACII (LS and MPW with
various interface boards), SUN 3 series (B011 interface), the DEC VAX under
ULTRIX and 4.3, the IBM PC AT and equivalents, etc.
The "C" compiler is completely draft ANSI compatible except it doesn't
currently support bitfields, "const", "volatile" or the multibyte international
character set library routines. Performance is VERY good with recent
comparisons putting us in the #1 or #2 spot in terms os code quality,
reliability, etc.
"C" library routines are provided for the T414 and T800 (which are the
currently supported Transputer types), for all the standard I/O, utility, math
and miscellaneous routines. The current release (88.3), doesn't support
some math routines on the T414 (transcendentals). Two additional libraries are
provided to support concurrency on the Transputer using an OCCAM like package
based on the "Jeff Mock" standard and a "fork"/"join" version design to
simplify the porting of UN*X code to arrays of Transputers.
Cost is $995.00 (US) which includes 6 months of updates and telephone
support. Contact:
Kirk Bailey
Logical Systems
P.O. Box 1702
Corvallis, OR 97339
(503) 753-9051
PS: To our existing customers the following features have been added since the
88.2 release:
1. ANSI style 64 bit doubles. We still support the 32 bit only mode an and
also support a "controlled" promotion mode for more control over automatic
expression floating ppoint type promotion.
2. A new internal optimizer has been added to improve code quality.
3. We've implemented 30 or so functions as "intrinsics" which are recognized
by the "C" compiler and which generate inline code. Individual functions may
be disabled using the ANSI "#undef" technique.
4. More I/O and system library routines as well as performance improvements
for some of the old ones ("printf" executes MUCH faster in 88.3).
***END ADVERTISMENT***
Cheers,
Kirk
PPS: Almost forgot, 88.3 also handles structure passing and return. C++ has
been ported to a pre-release of it by one of our customers...joseph@chromo (Joseph Reger) (07/03/88)
In article <8807022315.AA08504@mist.CS.ORST.EDU> bailey@MIST.CS.ORST.EDU (Kirk Bailey) writes: >In the recent posting from Hubert Matthews I noticed our "C" was missing from >his existance list... Must be time to let folks know what we've been up to: > >***BEGIN ADVERTISMENT*** >***END ADVERTISMENT*** > > Kirk Bailey > Logical Systems > P.O. Box 1702 > Corvallis, OR 97339 > (503) 753-9051 > First off, I have no relation to Logical Systems other than being a satisfied customer for the following reason: Some four months ago I posted an inquiry about C compilers for the Transputer. I was in great need, because I had a physics application ( a Monte Carlo simulation ) that had to be run for about a month on our 16 T414-s to get results for the March meeting of the American Physical Society. I called up a number of vendors, but none of them could deliver immediately ( I wanted a stand alone development system, because by this time I really hated TDS ). I got a beta release from Kirk (bought it of course). This came in two days, and after one single call about the network loader I was able to port my whole application, plus a number of UNIX utilities I often use, in one single day. I was extremely pleased with the product, never had any major complaint except for optimization (which is now fixed in the latest release, see Kirk's advert.) I would like to state that I am a physicist, I do mess around with computers a lot, but I have no deep knowledge about soft/hardware, but I was able to do everything I wanted and even more (this is basically "porting" the compiler to my Sun workstation to generate assembly listing of the most crtitical routine and hand optimize it a little). I use Turbo C-s 'make' utility with this C compiler, I have my own (very good) random number generator in the runtime library, that produces *exactly* the same numbers as on UNIX, etc. I have not seen the new optimizer yet, but would like to. My MOnte Carlo simulation (totally in integer) ran about at 75% of the SUN 3/50, on one processor. I somehow feel that one could do better than that. If it gives a significant improvement, I think this product is a steal. Check it out. - Cheers, Joseph, joseph@chromo.ucsc.edu, reger@ucscc.bitnet. PS: I write this as a service to the newsgroup, not to Logical Systems. I beleive this is what is missing from the Transputer discussion, people should share their experience! I would have loved a contribution like this four months ago.
tpm@digsys.engineering.cambridge.ac.UK (tim marsland) (07/06/88)
A satisfied Logical Systems customer writes.. (anyone out there prepared to submit a review of C compilers yet?): > Logical Systems sells a standalone "C" development system .. We currently have the 88.2 release, and are looking forward to 88.3 > .. All tools are provided in "C" source code form .. This has been an enormous advantage to us, both in terms of porting the code to SunOs 3.5/Ultrix 2.2 and in producing our own utilities, but also as a means of understanding more about the way the transputer works. The fact that it is a cross-compiler is another advantage because it means that we don't have to tie up our transputer resources just to run the compiler or linker, or pass #include or library files up and down a transputer link. Compilation rates are no problem on either a 750 or a Sun 3 (generally faster than cc(1)). > Two additional libraries are > provided to support concurrency on the Transputer using an OCCAM like package > based on the "Jeff Mock" standard and a "fork"/"join" version .. These libraries give us the flexibility that occam, despite its mathematical elegance, distressingly lacks. One *can* construct C programs in ``the occam way,'' but one doesn't have to. We think this is a positive advantage, particularly in an educational/research environment. > .. C++ has been ported to a pre-release of it by one of our customers... We are quietly hoping that this will soon become part of the transputer toolset, as we see considerable attraction in the view of processes as concurrent objects.. any chance Kirk? Tim Marsland Richard Prager Cambridge University Engineering Dept., Trumpington Street, Cambridge CB2 1PZ, UK PS: The opinions expressed herein are our own, not necessarily those of Cambridge University Engineering Dept. PPS: but-occam-is-wonderful or similar flames to /dev/null please