Mail Archives: djgpp/1992/06/11/13:59:59
>
> On behalf of some disencouraging comments on f2c usage I want to leave
> a word 100% in contrary to these statements:
>
[stuff deleted]
>
> Anyone still saying anything against f2c ?
>
Yes, me. Maybe i've just used the wrong environment (pure DOS), and
outdated version of f2c (i took the DOS-executables from
trickle/simtel20) or just the wrong fortran program ;-)... but i had
nothing but trouble.
First: the program to be compiled was well running under VM/CMS and
could be compiled and ran fine with this rs6k's xlf. Ah yes. And a
smaller version (64x64 matrices instead of 1024x1024) compiled and ran
fine with MS-Fortran.
To be able to use djgcc, i threw out the SIGNAL-related stuff of the
libf77 (or libi77) as djgcc cannot handle them. Afterwards the
libraries compiled with almost no modification.
But when i tried to run the compiled and linked stuff (*this* step
made no problem at all), i had severe trouble with IO-operations:
* Text supposed to appear on the screen appeard in a file called
`fortran.1' or similar.
* an ASCII file which was read in without problems in the environment
mentioned abv., produced a (fatal) read error. Then i replaced the
routine by an appropriate C routine. (i prefer C-style formatted IO
to that !@#$% fortran formatted IO anyway ;-)
* a binary output file (that was larger than 64K -- but this should
make no difference, should it? It is written out in chunks of 8K
each) produced an obscure IO error `illegal error number'.
At this stage i dumped the whole fortran stuff and decided to start
from scratch. Strong optimizations had to be taken anyway and it was
impossible to work with the machine generated C code (even though one
can reach a lot with regular expressins :-)
Yes, i'm not an expert of fortran at all -- i strongly prefer
C (++ or not). ;-)
- Thomas
greve AT rs1 DOT thch DOT uni-bonn DOT de
unt145 AT dbnrhrz1
- Raw text -