RE: [sv-bc] E-mail Ballot: Respond by Sun Sep 16 8am PDT

Steven,

Regarding 933, the BNF restricts the casting type to be a
constant_primary, and does not allow a general constant_expression.

My recollection was that we agreed talking about primaries in the text
would be unnecessarily confusing to most readers.

-- Brad

-----Original Message-----
From: Steven Sharp [mailto:sharp@cadence.com] 
Sent: Sunday, September 16, 2007 11:44 AM
To: shalom.bresticker@intel.com; Brad.Pierce@synopsys.COM;
cliffc@sunburst-design.com; Surrendra.Dudani@synopsys.COM;
Mark.Hartoog@synopsys.COM; fm@cadence.com; Karen.Pieper@synopsys.COM;
Dave_Rich@mentor.com; sharp@cadence.com; gordon_vreugdenhil@mentor.com;
stuart@sutherland-hdl.com; alex_gran@mentor.com;
Don.Mills@microchip.com; hmcdvi@msn.com; thomas.r.alsop@intel.com;
doug_warmke@mentor.com; Logie.Ramachandran@synopsys.COM;
will_cummings@mindspring.com; matthew.r.maidment@intel.com
Cc: sv-bc@eda-stds.org
Subject: Re: [sv-bc] E-mail Ballot: Respond by Sun Sep 16 8am PDT

I voted NO on 933 and 1438, with reasons given.


>SVDB  933 ___Yes   _X_No  
>http://www.eda.org/svdb/view.php?id=933

Didn't we agree that this needed to be a constant primary, not just a
constant expression?  Allowing an arbitrary expression gets us into
issues of the precedence of the ' with respect to operators.  I.e. is

	1+P'(x - 2)
	
supposed to mean

	(1+P)'(x - 2)
	
or

	1 + (P'(x - 2))
	

>SVDB 1175 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1175
>
>SVDB 1302 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1302
>
>SVDB 1354 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1354
>
>SVDB 1417 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1417
>
>SVDB 1438 ___Yes   _X_No  
>http://www.eda.org/svdb/view.php?id=1438

The new text says that the real number is converted to an integer.  That
implies truncation to 32 bits before assignment to the data object.
This is incorrect.  Values too large to fit into 32 bits can still be
converted successfully to wider integral data types.  The text should
say that it is converted to a signed integral value, not an integer.  If
something more specific is desired, it could say that it is converted to
a signed integral value of the same width as the final type.


>SVDB 1468 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1468
>
>SVDB 1473 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1468

A comment on Dave's concern about a keyword starting with a number:
This is not actually a keyword, since it is not an identifier and
therefore does not need to be reserved.   It is a unique syntax that can
be distinguished from anything else.  It is a bit unusual, but I don't
see any actual problem with it. 

I don't know that the meaning of 1step has been defined for all of the
places where delay_value can appear, but this change is an improvement.


>SVDB 1554 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1554
>
>SVDB 1767 _X_Yes   ___No  
>http://www.eda.org/svdb/view.php?id=1767

Steven Sharp
sharp@cadence.com


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