On 31 May 2013, at 9:37 AM, Jon Taylor <[email protected]> wrote:
> Thanks for the tip. What should I use as a value to enable an infinite
> stop-delta? Zero?
>
The purpose of delta simulation cycles is to emulate concurrent execution
leading to signal updates not to enable synthetic programming.
A look through the LRM(s) will show that there is nothing that implies a limit
to the number of delta cycles. You will find that ghdl does impose a limit
based on the expressible range in a natural (half the range of an integer). The
comparison is >=, so setting --stop-delta to 0 won't work. The delta counter
gets set to zero after time next is incremented.
grt-processes.adb, 986:
if Current_Delta >= Stop_Delta then
Error ("simulation stopped by --stop-delta");
exit;
end if;
The reasons for placing limits in the first place would be to prevent code from
never completing (which you are trying to overcome) or to prevent running out
of memory by dynamic elaboration (subprogram calls which can be called
recursively).
Historically all early VHDL simulators had delta cycle maximums (see Re: What
is delta simulation time?). I'm surprised ghdl's is as high as it is, if I
recall correctly the original Modelsim had it's set to 99. It was common to
have the ability to report delta cycle statistics to allow users to learn to
program better VHDL models, using a delta cycle count for description
optimization feedback.
A solution might be to use a foreign subprogram call containing the unbound
loop or programming in another language or implementation. If you have the
capacity to compile ghdl you could alter it to selective disable stop-delta,
the feature doesn't otherwise appear to be widely needed.
Despite any hype to the contrary VHDL is not particularly suitable to general
purpose programming and an infinite number of delta cycles doesn't represent
hardware described in VHDL's signal object paradigm using discrete simulation
time advancement and if you were to recast your code as a signal driven
hardware model you'd still be up against time'HIGH.
You could note that IEEE Std 1076-1987 doesn't contain the word delta. There
is no expectation someone would write zero simulation time hardware models.
_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss
