| Re: Praise for Gfortran (finally) |
|
 |
|
 |
|
 |
|
 |
Group: comp.lang.fortran · Group Profile
Author: glen herrmannsfeldtglen herrmannsfeldt Date: Sep 20, 2008 12:58
James Van Buskirk wrote:
(snip)
> The Intel compilers for 64-bit Windows at least used to be cross-
> compilers in that they ran under 32-bit mode, making it in principle
> possible to overrun compiler resources for a program that would not
> exceed runtime resources.
One would hope that the code for most programs would fit
in 2GB even
integer x(10000000000)
x=1
write(*,*) x(987654)
end
will compile in 32bit g95, but won't run.
integer x(1000000000)
data x/1000000000*1/
write(*,*) x(987654)
end
the compiler runs for over seven minutes before it dies with
an internal compiler error. (g95 0.91)
Someone last year had a very complicated DATA statement
that would run out of compiler resources much earlier.
> There are gfortran builds for 64-bit Windows that run on 32-bit
> Windows, Cygwin, or even LINUX. I think they are all built on
> LINUX because the build environment assumes some kind of UNIX
> command shell, so the versions that don't run under LINUX or 64-bit
> Windows are triple crosses.
There are versions of unix-like command shells, that
run on cygwin and on Win-32.
-- glen
| 
