ok, may be a better way to write it (with the same result) is: property p1; logic [7:0] lv; logic [7:0] lv1; logic [7:0] lv2; (1, lv = 1) ##1 (1, lv1 = lv) ##0 (1, lv2 = lv) ##0 // lv1 = 1 (1, lv = lv1 + 1) ##1 // lv1 = 2 (1, lv = (lv > 3) ? lv2+1 : lv2-1) // lv1 = 0 ##1 (lv == 0) ; // p pass endproperty The idea is that when passed, (when the evaluation of the sequence starts) the value of the actual arguments is being sampled and used as a constant for the rest of the evaluation. I agree that we need to work harder to describe the semantics this way. My question is, how people though about it until now, and whether we loose something if we stick to substitution. Doron Korchemny, Dmitry wrote: >Hi Doron, > >You are considering passing by value as lv/* new definition */ = >1/*usage, passed by value*/ + 1 > >Following your approach the following should be legal: > >sequence s(int lv); > (1, lv = 1); >endsequence > >property p; > s(1); >endproperty > >which doesn't look natural. > >Dmitry > >-----Original Message----- >From: owner-sv-ac@server.eda.org [mailto:owner-sv-ac@server.eda.org] On >Behalf Of Doron Bustan >Sent: Tuesday, October 17, 2006 4:25 PM >Cc: sv-ac@server.eda.org >Subject: Re: [sv-ac] local vars as actual arguments > >Dmitry, > >I think that my example > >property p1; >logic [7:0] lv; > >(1, lv = 1) ##1 // lv = 1 > (1, lv = 1 + 1) ##1 // lv = 2 > (1, lv = (lv > 3) ? 1+1 : 1-1); // lv = 0 > ##1 (lv == 0) ; // p pass >endproperty > >works like functions in c and other programing languages and has a >meaning. > >Doron > > > > > > >Korchemny, Dmitry wrote: > > > >>Hi Doron, >> >>I think that passing local variable by value is problematic. I would >>interpret it in your example as: >> >>property p1; >>logic [7:0] lv; >> >>(1, lv = 1) ##1 >> (1, 1 = 1 + 1) ##1 >> (1, 1 = (1 > 3) ? 1+1 : 1-1); >>##1 (lv == 0) ; >>endproperty >> >>which is meaningless. >> >>Thanks, >>Dmitry >> >>-----Original Message----- >>From: owner-sv-ac@server.eda.org [mailto:owner-sv-ac@server.eda.org] On >>Behalf Of Doron Bustan >>Sent: Tuesday, October 17, 2006 2:21 PM >>To: sv-ac@server.eda.org >>Subject: [sv-ac] local vars as actual arguments >> >>Hi, >> >> >>last meeting I took an action item to review passing local variables >>as actual arguments to sequences, who in their turn are passed as >>actual arguments to other sequences. >> >>It looks that as long as all local variables in the different >>sequences have different names, an approach based on substitution >>will work. >> >>However, it is not clear that substitution is consistent with >>the way local variables are being considered as actual arguments. The >>LRM does not make it clear whether local variable are passed by value >>or by substitution. To understand the differences, look at the >>following example: >> >>sequence r1(lv1, lv2); >> (1, lv1 = lv1 + 1) ##1 >> (1, lv1 = (lv1 > 3) ? lv2+1 : lv2-1); >>endsequence >> >>property p1; >>logic [7:0] lv; >> >>(1, lv = 1) ##1 r1(lv, lv) ##1 (lv == 0) ; >>endproperty >> >>When we take the substitution approach, we have >>that p1 is equivalent to: >> >>property p1; >>logic [7:0] lv; >> >>(1, lv = 1) ##1 // lv = 1 >> (1, lv = lv + 1) ##1 // lv = 2 >> (1, lv = (lv > 3) ? lv+1 : lv-1); // lv = 1 >>##1 (lv == 0) ; // p fails >>endproperty >> >> >>When we take the pass by value approach, we have >>that p1 is equivalent to: >> >>property p1; >>logic [7:0] lv; >> >>(1, lv = 1) ##1 // lv = 1 >> (1, lv = 1 + 1) ##1 // lv = 2 >> (1, lv = (lv > 3) ? 1+1 : 1-1); // lv = 0 >>##1 (lv == 0) ; // p pass >>endproperty >> >> >>At least in sv-ac discussion, we did refer to passing >>local variables as actual argument as "pass by value. >> >>I think that we need to discuss the differences, >>and see if it is ok to continue with the substitution. >> >>Doron >> >> >> >>Received on Tue Oct 17 08:02:17 2006
This archive was generated by hypermail 2.1.8 : Tue Oct 17 2006 - 08:02:21 PDT