Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!usc!snorkelwacker.mit.edu!ai-lab!geech.gnu.ai.mit.edu!petrilli From: petrilli@geech.gnu.ai.mit.edu (Chris Petrilli) Newsgroups: comp.sys.next Subject: Re: Freewhare hurts the NeXT (was Re: Desktop publishing) Message-ID: <14574@life.ai.mit.edu> Date: 3 Apr 91 00:47:00 GMT References: <1991Mar30.063733.27117@news.cs.indiana.edu> <14485@life.ai.mit.edu> <1192@nikhefh.nikhef.nl> Sender: news@ai.mit.edu Distribution: comp Organization: Free Software Foundation Lines: 76 Jos Vermaseren writes: >I did recently some real hackers tests on GNU versus Turbo-C on the Atari. >I have a program FORM (symbolic manipulation program for truely big things) >(available by anonymous ftp from nikhefh.nikhef.nl) which gives exceptionally >compact code on the atari, so much so that the code part of the execuatble >is 10% shorter than what GNU produces. The comparison of one simple (or even not so simple) program is not a basis for conclusion. That's like saying Computer A does 12.4 MIPS, and Computer B does 12.6, therefore Computer B is better, it doesn't mean anything. I suggest you attempt a more "complete" comparison, as I am sure the people around here would be interested in the results. Perhaps try compiling several LARGE programs (I don't know if AKCL could be done on the Atari but that's a good program, and then Emacs v18. :-) so that any simple differences would be avoided. I am not questioning that there is a better compiler, nor that there could be, just that one program doesn't a compiler make. :-) Also, it helps if times are given, 4% on a 4 second program is pretty insignificant compared to 4% on a 12 hour program, and in fact, 4% on a 4 second program could mean absolutely nothing, I mean did you account for different processor loadings? >So I hacked some special startup code so that I could run the atari executable >on my NeXTstation for the real test. (including relocation and redirection >of some routines to the NeXT system calls). >The result was that the executable of the atari is nearly 4% faster than what >the GNU compiler generates. As I pointed out, 4% is not enough difference, especially on only one program. If the compiler consistently generates better code (even 4% See above... glad to see you like gdb, I like it, but it needs alot of work to be perfect (for me anyway, I love Lisp machines). The thing is that the GNU compilers are running on ALOT of machines, I believe it is running on: 68K, SPARC, MIPS, PA-RISC, RS/6000, 386, VAX, and several others. >The problem with high quality free software is that it discourages >professional programmers to work for that platform. There isn't anything >wrong with that as long as enougm good programmers can work on a GNU >grant. Barring that you may wonder how all those other programmers are >going to survive. It is just a social point. I am still happy with GNU. I disagree... free software (of high quality) forces people who wish to charge for software to produce better products. As someone pointed out, if someone wants to charge $4000 for a compiler (like Sun), it better be substantially better than GNU compiler (or any free compiler), and in the case of the Sun compiler, it doesn't even seem to be the equal of the GNU compiler. Not everyone is getting paid for doing work around here... I am writing the GNU C Reference Manual for $0 (free in other words). I use GNU software, and feel that I should help out, not to mention that I WHOLEHARTEDLY believe in what RMS and the FSF are doing. I think professionals see the FSF as a way to ignore the fact that they are not as good as they claim to be (I don't know many "professionals" that are of the quality of the people around here), and simply say "How can we compete, they are giving it away," rather than simply admitting "How can we compete, they are better." I am not saying it's impossible to produce a better compiler, but I suspect that it will come from the FSF for one VERY simple reason, we guide ourselves, we don't have a marketing department to tell us what the public wants, we are the public. Also, we don't have leadership that says "that's not worthwhile," in fact, people are encouraged to produce whatever they want, if someone doesn't want to work on a compiler, they are welcome to write somethign else (including the documentation). Chris -- + Chris Petrilli | Internet: petrilli@gnu.ai.mit.edu + Insert silly disclaimer drivel here.