(j3.2006) (SC22WG5.5770) RE: RE: Other languages with unit support
Van Snyder
Van.Snyder
Sun Jul 10 17:10:45 EDT 2016
On Sun, 2016-07-10 at 20:51 +0000, Whitlock, Stan wrote:
> This still doesn't address the problems inherent in using
> preprocessors or directives that aren't integrated with the processor.
> Using source debuggers becomes more tedious and expensive because line
> numbers are different. It's not obvious the decrease in cost inherent
> in increased reliability is more than the increase in cost imposed by
> the inability to use debuggers effectively.
>
> Stan> These are not concerns that the Fortran Standard ever has chosen
> to address even though they have been around since the first
> debuggers. Some vendors and third parties have some good interactive
> debugging facilities but those are and will remain outside the
> standard.
> End Stan>
I didn't say the Fortran standard should address this red herring. I
simply said that because of this problem, preprocessors aren't the ideal
solution that others insist.
There's also the problem when using a preprocessor that error messages
from the processor refer to its line numbers, not the line numbers in
the original source. That increases labor cost, and might offset the
benefits from using preprocessors.
If features provided by preprocessors, that result in different line
numbers from the point of view of the processor, were in the processor
instead of a preprocessor, this wouldn't be an issue.
Those who insist preprocessors and directives are good enough still
haven't answered my question "Why did we need to do DO CONCURRENT and
the goofy locality specs, if it's the same as something in OpenMP?"
Would Cray have rolled over and played dead if somebody had insisted it
is possible to do coarrays using preprocessors, with no standard for the
preprocessors?
More information about the J3
mailing list