Re: Analog function question

From: Kevin Cameron <kevin_at_.....>
Date: Fri Sep 16 2005 - 14:18:56 PDT
Jonathan David wrote:

>While on the one hand I'd like one tool, one language
>to work in.. I think this may really be impractical,
>and instead think we should work towards defining how
>the different languages SHOULD interact.. so that we
>can know what the behavior is if we use a
>SystemVerilog-AMS verification component or PSL-AMS
>assertion - or a connect to a SystemC component that
>implements a Matlab module.. 
>that's my two cents as a "Design Verifier" using these
>tools now! (but probably not a "voting" member of this
>committee!)
>
>Jonathan David
>Scintera Inc.
>  
>
That approach may work OK for digital stuff (although I would say not 
myself),
but the way AMS works requires that the code is analyzed together in 
order to
work out which modules are required to be added for analog-to-digital 
conversion
at the analog/digital boundaries, and the runtime signal connectivity 
can be quite
different (for the digital side) than it would be without the analog 
components being
there.

The reason I keep pushing for SV integration is that almost every time 
someone
comes along and asks for new functionality in AMS a chunk of the work has
already been done for SV, so we spend time unnecessarily discussing the 
reinvention
of wheels. Also, a single language was the goal of the committee at the 
outset (set by
OVI).

Kev.

>--- Kevin Cameron <kevin@sonicsinc.com> wrote:
>  
>
>>Muranyi, Arpad wrote:
>>
>>    
>>
>>>Thank you all for answering this question for me.
>>>Unfortunately this was the answer I suspected.
>>>
>>>I would like to make a request that this capability
>>>be added.  A function that is written to process
>>>arrays is not very useful if the arrays it can take
>>>as input (or return as output) must be a
>>>      
>>>
>>predetermined
>>    
>>
>>>fixed size.  (This is assuming that Marq's previous
>>>reply was correct in saying that arrays can be
>>>      
>>>
>>passed
>>    
>>
>>>at all).
>>>
>>>What is the procedure for submitting requests?  Can
>>>I (an observer) enter items in Mantis, or does it
>>>have to be done by one of the "officials"?
>>> 
>>>
>>>      
>>>
>>I would suggest that you vote for the earliest
>>possible integration with 
>>SystemVerilog since it has those capabilities
>>already. In particular: 
>>old style argument and return passing for routines
>>is by value which  is 
>>particularly inefficient with arrays, and
>>SystemVerilog allows 
>>pass-by-reference
>>(http://www.eda.org/sv/SystemVerilog_3.1a.pdf - Sec 
>>10.4.2).
>>
>>Kev.
>>
>>
>>    
>>
>>>Thanks,
>>>
>>>Arpad
>>>      
>>>
>>======================================================
>>    
>>
>>>________________________________
>>>
>>>From: owner-verilog-ams@eda.org
>>>      
>>>
>>[mailto:owner-verilog-ams@eda.org] On Behalf Of Marq
>>Kole
>>    
>>
>>>Sent: Friday, September 16, 2005 12:28 AM
>>>To: verilog-ams
>>>Subject: Re: Analog function question
>>>
>>>
>>>
>>>Oops, need to correct a small mistake below: 
>>>...
>>>...
>>> 
>>>
>>>      
>>>
>>>>>analog function real myFunctionName;
>>>>>input in1;
>>>>>real  in1[1:5];    <--------- How would I know
>>>>>          
>>>>>
>>the size?
>>    
>>
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>...
>>>...
>>>I mean an argument-controlled size given the fact
>>>      
>>>
>>that you cannot set 
>>    
>>
>>>a parameter value from the outside. So a call: 
>>>
>>> myArrayFunction(myArraySize, myArray); 
>>>
>>>Where myArraySize argument carries the length of
>>>      
>>>
>>the myArray argument 
>>    
>>
>>>cannot be handled in current Verilog-AMS. 
>>>...
>>>...
>>> 
>>>
>>>      
>>>
>>>>Marq Kole
>>>>Competence Leader Analog Simulation, Philips ED&T
>>>>   
>>>>
>>>>        
>>>>
>>> 
>>>
>>>      
>>>
>>    
>>
>
>
>  
>
Received on Fri Sep 16 14:19:00 2005

This archive was generated by hypermail 2.1.8 : Fri Sep 16 2005 - 14:19:58 PDT