Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!seismo!rutgers!cbmvax!higgin From: higgin@cbmvax.cbm.UUCP (Paul Higginbottom GUEST) Newsgroups: comp.sys.amiga Subject: Re: C compiler questions (Mostly Manx). Message-ID: <1010@cbmvax.cbmvax.cbm.UUCP> Date: Wed, 19-Nov-86 12:05:23 EST Article-I.D.: cbmvax.1010 Posted: Wed Nov 19 12:05:23 1986 Date-Received: Thu, 20-Nov-86 02:26:52 EST References: <727@ulowell.UUCP> <2118@pur-phy.UUCP> Reply-To: higgin@cbmvax.UUCP (Paul Higginbottom GUEST) Distribution: na Organization: PDH, Inc. Lines: 97 Keywords: C compilers, floating point In article <2118@pur-phy.UUCP> ng@pur-phy.UUCP (Giordano) writes: >I am planning to purchase a C compiler... >...I am heavily leaning towards Manx, based on its superior compile/link/run >times as compared to Lattice. > >My questions are: > >(1) Manx Developers versus Commercial versions. Is it worth it for someone >like me to pay the extra money for the Commercial version? I'm not sure about this, but I don't know if the Developer (a misnomer if ever I heard one) version comes with a debugger, or even "make". VERIFY THIS before buying - those two are pretty essential, not to mention a nice editor (z) if you're into "vi" style editing as opposed to "emacs" variants. I have the Commercial version and DON'T consider it to be filled with all kinds of things I don't need - I use almost all of it, so I personally would lean heavily in favor of the Commercial version. >(2) Programs with large (usually) data segments. How is Manx for these >applications. No problem. Manx supports different memory "models" and small data AND small code are not a problem usually for even those huge mungo programs in the public domain like Hack. The reason: small data means no more than 64K of INITIALIZED GLOBAL DATA - pretty damned unlikely! If you want to do float stuff[BIZILLIONS]; The compiler won't care because the storage won't be initialized with values (although the startup code automatically makes globals zero if they do not have an initial value). Small code means no more than 32K functions! ALSO p.d.u! >I have read a lot on the net about using Manx with 16 versus >32 bit ints. Does one use different libraries for the two cases, and if >so, do they both come with the Developers package? Yes, there are two different libraries to link with, and (a guess) I'm sure they must both come with the Developer version (but call Manx 1-800-221-0440). >(3) 1.2 compatibility. I spoke with Manx last week, and they said they >expect to release the 1.2 compatible version in a "month or so". I would >rather not wait that long if I don't have to. Can the current version >be made compatible with 1.2 (I believe that I read that it can), and if >so how, and are there any remaining bugs? "1.2 compatibility" means that those NEW routines in 1.2 are accessible through Manx's NEW link libraries (for the new version). Since 1.2 is compatible (from a programmer's point of view) with 1.1 you can CERTAINLY use the currently sold version to write programs under 1.2 - you just won't be able to access the new routines until you receive your FREE upgrade from Manx (so waiting for the new version is not worth it - in the interim you'll get yourself comfortable with the compiler and tools). Re: bugs - do you mean in 1.2 :-) or the "to be released in a month or so" version of Manx? The former - I've heard there is; the latter - compilers are never perfect but the guy who works in this one has worked incredibly hard to fix all known problems and it's a great piece of work. I'm writing applications which contain 60 source files, and produce executables of 60K+ as well as a run-time library and other tools with NO PROBLEMS. We've run into one or two glitches in the compiler, but the "to be released" version seems to fix these (I'm a Beta tester). >What else will be in the next release? The author has spent a lot of time improving the math (the first version [3.20] released only supported single precision math internally) which now supports single and double precision (MUCH faster than Lattice) using either Amiga's math or their own, AND it supports the 68881/010/020 chips. I have no idea how well this works though. The new release is basically an upgrade - better math, fixed bugs in the compiler, linker, assembler, etc, and support for 1.2 enhancements. >By the way, what version of Manx is the current one? 3.20 >(4) Floating point. The benchmarks I have seen give Manx a large edge in >floating point execution speed (which for my small scale crunching would >be important). There has been some discussion on the net about >Manx handling floating point operations differently from Lattice (I believe >FFP versus IEEE are the terms used). What exactly do these mean, especially >concerning accuracy? Read above. >Thanks for the answers these questions and any other comments on Manx versus >Lattice which anyone may have. > >nick. You're welcome, Paul. Disclaimer: I work for myself, and my opinions are my own.