From: user5857@newsgrouper.org.invalid   
      
   anton@mips.complang.tuwien.ac.at (Anton Ertl) posted:   
      
   > Thomas Koenig writes:   
   > >Anton Ertl schrieb:   
   > >> Thomas Koenig writes:   
   > >   
   > >>>So, kill the 64-bit machines in the scientific marketplace. I'm glad   
   > >>>you agree.   
   > >>   
   > >> Not in the least. Most C programs did not run as-is on I32LP64, and   
   > >> that did not kill these machines, either.   
   > >   
   > >Only those who assumed sizeof(int) = sizeof(char *).   
   >   
   > And lots of others, e.g., those that assumed that longs are 4 bytes in   
   > size.   
   >   
   > >This was   
   > >not true on the PDP-11,   
   >   
   > Can you elaborate on this? What do you think is sizeof(int) on a   
   > PDP-11, and what do you think is sizeof(char *) on a PDP-11?   
      
   sizeof int == 2   
   sizeof char * == 2   
      
   > >and it was a standards violation, anyway.   
   >   
   > That's hilarious. C89 was three years old in 1992. The majority of C   
   > programs available in 1992 were started before ANSI C was released,   
   > and thus contained code from before ANSI C. And like today,   
   > programmers are asked to spend time on other things than fixing things   
   > that are not broken.   
      
   If application vendors were subject to the same recall standards that   
   the auto industry is subject, that might change. {Remember the Pinto}   
      
   > >> And I am sure that C   
   > >> programs were much more relevant for selling these machines than   
   > >> FORTRAN programs.   
   > >   
   > >Based on what data?   
   >   
   > Based on 4 months of internship at HP in 1988 and 1989, in a group   
   > that did sales support, tech support, and courses on HP 9000   
   > workstations and servers and HP/UX (the OS of the HP 9000 machines).   
   > I don't remember hearing about a customer that used FORTRAN.   
      
   C got the OS and compilers up and running, then the people who   
   bought the machine run applications they can compile from their   
   source.   
      
   > Based also on the impressions I got on Usenet. Apart from SPECfp,   
   > Fortran was nowhere to be seen.   
      
   Most FEM, Optics, CFD and larger scale engineering applications are   
   all written in FORTRAN with C-front ends shuffling data/commands   
   back and forth. {Spice, Layout, Design Rule Checking, GDSII, ...}   
      
   > >>C programmers changed the programs to run on   
   > >> I32LP64 (this was called "making them 64-bit-clean"). And until that   
   > >> was done, ILP32 was used.   
   > >   
   > >The problem with 64-bit INTEGERs for Fortran is that they make REAL   
   > >unusable for lots of existing code.   
      
   Nonsense::   
      
   CDC only had Double Precision FP data (60-bit)   
    with 18-bit integers   
   CRAY only had Double Precision FP data (64-bit)   
    with 24-bit integers   
      
   {{Even numerical analysists liked Seymore's 60-bit and 64-bit arithmetic   
   compared to 32-bit IBM and 36-bit Univac FP arithmetic--even with those   
   littered with huge mistakes we would not allow today.}}   
      
   > The size of FORTRAN INTEGERs is something the FORTAN people have to   
   > decide, and I made no statement on that.   
   >   
   > If FORTRAN programs make the assumptions that sizeof(int)==4, maybe   
   > you should tell the FORTRAN programmers something along these lines:   
   > "it is a standards violation, anyway. Only people who like to play   
   > these kind of games are caught."   
      
   FORTRAN programmers think of integer as 1 storage container--even on   
   CDC and CRAY. The integer in memory is 60 or 64 bits, the integer in   
   register is 18-bit or 24-bit. FORTRAN programmers do not have problems   
   with putting 6×6-bit characters in PDP-10 memory container, or 10×6   
   field-data characters in one CDC memory container.   
   >   
   > - anton   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)   
|