|
From: Cary R. <cy...@ya...> - 2009-02-23 20:24:12
|
--- On Mon, 2/23/09, Stephen Williams <st...@ic...> wrote:
> You have to look for all the places where the elaborator
> does
> timescale conversions. Sometimes the expression is
> non-constant
> and different code it invoked to do the scaling, but it may
> not
> be all that hard to find all the spots.
I believe all delays including variable delays should be
processed in elaborate_delay_expr(). The first half does
constant delays and the second part adds mul or div as
needed to scale non-constant delays. There were a few
places that used to just pass the raw variable, but that
is incorrect and I think I fixed all/most of them to
us this routine to do the correct scaling.
Cary
|