Re: GPL vs. sweatshops (was Re: Bias)

Am Don, 2002-06-13 um 13.24 schrieb Roger Fujii:

> Care to name any?   Only one that comes to mind is the Intel C compiler and that's
> a special case.

I have searched for Fortran compilers a lot in the past, and seen that
most vendors who sell Fortran also sell C/C++. Lahey/Fujitsu comes to
mind, for example, and the Portland Group. Mostly high-performance
stuff, of course.

> Fortran isn't necessarily a good example, given that I don't think even IBM
> has accepted F95 (I know they vetoed F90).   Is there active work on F?? on
> gcc?  (I know g77 didn't have any active interest a little bit ago).

Whatever IBM does, both 90 and 95 are standards. g77 is an excellent
Fortran 77 compiler, but has no support for F90/95. The g95 project is
yet in a pre-pre-alpha state.

> gcc *never* had as much diversity.  

Of course not. It is only _one_ compiler. But there used to be a
diversity among C compilers, and I think there still is. Just that
most Linux people see very little need to check out alternatives to gcc.

You are also overlooking the fact that all the fancy products you're
talking about are much more than just a compiler. They should be rather
compared to IDEs. AFAIK there are a couple of different IDEs for C/C++
programming for Linux, both free and non-free. Of course, almost all of
them use gcc as the compiler and gdb as the debugger.

> I am not going to put forth the effort required to become qualified to work on it.  I
> have
> looked at the code before, and the operations of the compiler is non-trival (side
> effects
> of macros, proper sequencing of setups...).  Besides, I don't *want* to work on a
> compiler.

Bad luck for gcc. If you are right and gcc generates faulty code for
Sparc64, that would be the ideal chance for a commercial vendor to offer
a superior product. Don't tell me they cannot compete because they can't
build on the gcc engine. There are enough other engines out there that
they could license.

> Only one I know that fits this description is IA64.  
I think similar things have happened for embedded processors.

> Besides, I don't see how
> this affects my statement.  Once you have the compiler working, there is little
> commercial incentive in improving it.  

Sure. Better optimization may give you +20% performance, at the cost of
having to port your code. You might as well wait three months and buy
20% faster hardware. Unless the compiler has really advanced features 
(supporting massive parallelism, for example), the number of customers 
is of course limited. That has nothing to do with gcc being GPL'd. An
X11-licensed gcc would have the same effect on the market (if it had the
same quality as gcc).

> What non-x86 processor do you see a huge
> amount of work optimization work on for the gcc?

If gcc is so bad on these platforms, your whole argument is void.
Entrepreneurs would have ideal chances with commercial products.

Martin

-- 
Martin Wilck                Phone: +49 5251 8 15113
Fujitsu Siemens Computers   Fax:   +49 5251 8 20409
Heinz-Nixdorf-Ring 1	    mailto:[email protected]
D-33106 Paderborn           http://www.fujitsu-siemens.com/primergy