Re: [Iverilog-devel] $urandom( seed ) discrepancy

[Cary R. <cy...@ya...>]

--- On Thu, 6/18/09, Trevor Williams <pha...@gm...> wrote:


> It appears that Icarus treats the $urandom( seed
> ) like the $random( seed ) system call in that the provided
> seed is treated as an inout -- the value is used to seed the
> RNG and the function returns the new seed value back into
> the provided variable.  However, after seeing a discrepancy
> between Icarus and VCS in this regard, I consulted some
> online documentation which confirms that $urandom( seed )
> treats the seed parameter as an input only.  $urandom is a
> SystemVerilog built-in system call and SystemVerilog threads
> keep track of their own random seed value; therefore, the
> seed value does not need to be returned and, indeed,
> isn't.  Is there a reason why Icarus implements this
> functionality differently than the SystemVerilog standard?
>  If so and there is no plan to change it, it should
> definitely be documented as such in the wiki.  This would
> make any testbenches that use this functionality --
> originally developed for an Icarus platform -- to not work
> as expected in a SystemVerilog compliant platform.
> 
> Any thoughts?

My thought is that I did not have enough documentation when
I wrote the code and made it like $random. Given that neither
Steve or myself have a copy of the SV standard things like
this are bound to crop up. So, yes this is a bug and needs
to be fixed, but I will need more information before I can
actually figure out what needs to be done.

Cary