Open Watcom 1.0 Released

JoshRendlesham writes "The Open Watcom C/C++ and FORTRAN 1.0 compilers have been officially released. The source, and binaries for Win32 and OS/2 systems, are available. This release also means that outside developers can join and contribute to the project." Or if you prefer, gcc is up to 3.2.2.

Comments

DOS days

[Graspee_Leemoor]

Back in the days of DOS, if you were a developer, the Watcom C compiler was *the* thing to pirate.

graspee

Re:Stop duplication of effort

[jacquesm]

gcc could do with some competition, yes its stable, no it is not at the leading edge of performance any more, processor optimization is at least a generation behind what's commonly available and ignores some architectures completely.

I'm looking forward to someone benchmarking gcc vs watcom to see how they do.

Watcom was great. How about today?

[CresentCityRon]

In the late 80s (?) Watcom products were really great. They were beating on everything for the Intel platform.

I received the email yesterday about Watcom's "release" to open source. In that email it says that Sybase felt there was no commercial value in the product anymore so they released it. My question is "Has Sybase been keeping this thing up? Is it useful today?" Or is this a scam to try to give life to a dying patient? I mean perhaps people working on this might be better off working on gcc or something.

Thanks!

Re:cool ! that's great news

[Animats]

QNX uses the GNU/GCC toolchain now, and ship their premium product with the Dinkumware C++ library.

Incidentally, if someone can tell me how to prevent loader crashes in "ld" under QNX when there's an undefined symbol in a trivial program that includes "", I'd appreciate it. Nobody in the QNX newsgroups seems to know.

Performance comparisons

[golrien]

I was going to ask if there were any performance comparisons around showing how Watcom performed, but then I realised that anyone with half a brain ran something through Google before Slashdot.

Win32 compilers (not including Watcom - and with good reason, it's a bitch to set up on Win32)

as linked from the djgpp FAQ, some info on DOS compilers.

So, hooray! A lesson in using Google before Slashdot mixed with some blatant karma-whoring.

PS. this is good too.

No, actually

[exp(pi*sqrt(163))]

Intel have a long history of claiming that they produce a fast compiler, after all they know the Intel specs. However I have never found this to be true over the last 7 or 8 years (I think it was called Proton years ago). I am not sure I have found any code that is significantly faster compiled with the Intel compiler and have found much that is slower. I haven't tried v6 of their compiler though. Maybe, just maybe, they've now picked up some tricks from the KAI guys.

Incidentally, vectorization in Intel C/C++ is a joke. I put so many hints into my code (aligned variables, processed stuff in suitable sized chunks etc.) and still couldn't trigger the compiler to vectorize. It's much easier to insert SSE instructions yourself.

The Intel compiler has better error reporting than MSVC++. I use it when I don't understand why MSVC++ is barfing on my template code. This is more useful than it sounds!

Re:Watcom was great. How about today?

[robbyjo]

One thing I know is that their optimization routine rocks.

Well, optimization routines can be divided into two parts: One is architecture independent (which involves simplification of AST and stuff) and the other is architecture independent. IIRC, their architecture-independent optimization was really great. It can correctly detect redundant codes and simplify it.

I used to be an ASM programmer as I was a performance freak. When I compile my C/C++ program using Watcom, it almost always produced near optimized (i.e. the "gold-standard") asm code. I knew this when I dumped out the assembler code.

I knew that their arch-independent optimization is really good because when you add things such as calculation of busy expression (i.e. expression that you used over and over) and stuff, it correctly cache the calculation before hand. So, you will save a tremendous time, especially if you do it in a loop. The problem was (again, IIRC) that was not perfect and some of the expressions are left undetected. But, that's probably a bug.

IMHO, arch-independent optimization play a lot greater role than the arch-dependent one (ok, some of you may not agree with me). Things like peephole optimization is great, but is of limited usefulness once you apply the correct transformation of the AST and other internal structures.

This is also partly why Intel optimizing compiler is also great. I heard that some of the folks are doing partial evaluation on the code -- which can greatly help speeding up the result. The idea was: If you use a particular routine (like function) only with a handful of value range, it will automatically create a specialized and optimized function for you exploiting the nature of the input values. For example: You probably have seen the routine that calculates (-1)^n used in a routine that calculates x^y. The optimizing compiler thus should be able to generate: return (n && 1 == 0) ? 1 : -1; instead of the looping. This only involves some (expensive) static analyses computations. I have yet to see this in other compilers.

Therefore, this release is really really good thing. I hope that GNU compiler teams would pickup some of their good stuff.