On Tue, 07 Apr 2015 08:40:22 +0200
Tristan Gingold <[email protected]> wrote:

> Note that the current (disabled) implementation might not be
> very intuitive.  For example, if you select time resolution as us,
> you have:
> 1 ms = 1000 us,
> 1 s = 1000_000 us,

Yep.

> ...
> and 1 ps = 0 us, 1 fs = 0 us, 1 ns = 0 us.

Reasonable.

> 
> That looks obvious.  But the consequences aren't:
> 10 ns = 0 us

Cool.

> 100 ns = 0 us

Cool.

> 1000 ns = 0 us

Slightly curious at first glance but understandable.

> ...
> ie it is as if: 1000 ns = 1000 * 1 ns = 0 us

It seems [to me] like a reasonable limitation that the model shouldn't use time 
units that are more detailed than the resolution time unit.

On Tue, 07 Apr 2015 08:50:15 +0200
[email protected] wrote:

> Le 2015-04-07 08:40, Tristan Gingold a écrit :
> > ie it is as if: 1000 ns = 1000 * 1 ns = 0 us
> This case is easy to check or assert, isn't it ?

Agreed, it might be a nice feature if the compiler could issue a 
warning/reminder if the model uses time units that are more detailed than the 
resolution time unit. Alternatively, it might be sufficient just to note that 
detail in the man page.


_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss

Reply via email to