[sv-bc] RE: [sv-ec] Case @* - altera

Hi, Daniel.

You are correct. A strict interpretation of the LRM makes the implicit sensitivity list include i.

However, in this particular example, the question is whether there would be an initial change in i that triggers the always procedures. That depends on whether the initial procedure executes before or after the always procedures are initiated. The LRM does not define an order.

Your example also demonstrates a "gotcha" of always @* that can cause simulations to enter an inifinite loop. This is documented in my paper at http://www.scribd.com/doc/14210222/Return-of-the-SystemVerilog-Gotchasbresticker. Some of the simulation vendors have also documented this behavior. I have seen this occur more than once.

This also occured in OVL and was reported in Mantis 2174 and 2426, the latter filed by you.

SV has solved this gotcha in always_comb by excluding such variables from the implicit sensitivity list. This is one of the advantages of always_comb over always @*.

As a result of this gotcha, some vendors have started excluding such variables from the implicit sensitivity list of always @* anyway, like always_comb, as this is the desirable behavior.

Regards,
Shalom

________________________________
From: owner-sv-ec@server.eda.org [mailto:owner-sv-ec@server.eda.org] On Behalf Of Daniel Mlynek
Sent: Tuesday, May 19, 2009 9:56 AM
To: sv-ec@server.eda.org
Subject: [sv-ec] Case @* - altera

Altera in their libs is using code patterns like below (it was simplified in order to show the root of the problem):
module top;

integer i;

initial i = 0;

always @(*)
    for (i = 0; i < 10; i++)
    begin
        $display("1 ", i);
    end

always @(*)
    for (i = 0; i < 10; i++)
    begin
        $display("2", i);
    end

endmodule

The intention is that alwayses in above code  should be in-sensitive for changes of "i" - so nothing should be printed. I've analysed the LRM in context of above code and IMHO according to LRM alwayses should be sensivitve to i changes - as this is global variable readed in in both those process. There is no such rule in LRM that for loop index variable should be exluded from implicit sensitivity list.

Could you please take a look on above and point me how should it work and where is it described in LRM - or the author intention was wrong and code should be rewritten using local index variables in for loop?


DANiel

--
This message has been scanned for viruses and
dangerous content by MailScanner<http://www.mailscanner.info/>, and is
believed to be clean.
---------------------------------------------------------------------
Intel Israel (74) Limited

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.