RE: [sv-ac] Issue #1550

Hello Dmitry,
 
the definition 
sampled(expression, clocking event) = (clocking event) ? expression :
$past(expression, clocking event). The straightforward synthesis without
a system clock is not clear in general case.
 
has a problem because there is no way to detect clocking event in a
ternary expression.
 
What about something like this:
 
$sampled(expression, @(clocking_event)) ::= 
   var type(expression) sample_expr;
   always @(clocking_event) sample_expr = $sample(expression);
 
and use sample_expr in place of the original $sampled.
 
Would this work?
 
ed
 


________________________________

	From: owner-sv-ac@eda.org [mailto:owner-sv-ac@eda.org] On Behalf
Of Korchemny, Dmitry
	Sent: Tuesday, October 03, 2006 6:08 AM
	To: sv-ac@eda.org
	Subject: [sv-ac] Issue #1550
	
	

	Hi all,

	 

	During the discussion of #1550 a question was raised whether we
need to keep a second argument of $sampled function. The current
definition is $sampled(expression [, clocking event])

	 

	If we want to keep it then its semantics should be:

	 

	$sampled(expression, clocking event) = (clocking event) ?
expression : $past(expression, clocking event). The straightforward
synthesis without a system clock is not clear in general case.

	 

	E.g., always a = $sampled(b, @(posedge clk))

	or

	disable iff ($sampled(reset, @(posedge clk))

	 

	 Therefore we have three options:

	1.	Allow only one argument form, which is not backward
compatible 
	2.	Allow two argument form in trivial cases only, e.g.,
always @(posedge clk) a = $sampled(b, @(posedge clk)), which is
redundant 
	3.	Allow two argument form in trivial cases or when a
system clock is present (I have a proposal about a system clock in one
of my presentations) 

	 

	What do you think?

	 

	Thanks,

	Dmitry