### Table 10: Results for the symmetric version of the code.

2000

Cited by 67

### Table 2: Results for the symmetric version of the code on matrix CRANKSEG2.

"... In PAGE 18: ... Furthermore, comparison of symmetric and unsymmetric performance shows a good performance of the symmetric code, which is almost twice as fast as the unsymmetric version. Finally, Table2 presents results for the larger test problem CRANKSEG2 both on the SP2 and an Origin 2000 located at Parallab (Bergen). The speedups on the SP2 are due to both algorithmic and memory e ects.... ..."

### Table 2: Operational semantics (symmetric versions of (Sum), (Par) and (Com) omitted)

1999

"... In PAGE 16: ...?!, i.e.: if P Q and P ??! P 0 then there exists Q0 such that Q ??! Q0 and P 0 Q0 (the proof goes by inspection of the rules; see also [16]). The key to soundness is the following proposition, that relates equivalence on environ- ments ( ) to the (conventional) operational semantics of Table2 (its proof can be found in Appendix B). Proposition 4.... ..."

Cited by 54

### Table 4.8: Results for the symmetric version of the code.

1998

### Table 4.8: Results for the symmetric version of the code.

### Table 4.8: Results for the symmetric version of the code.

### Table 1. Deadlock and starvation freedom for symmetric (unfair) version.

2005

"... In PAGE 12: ... We also computed a lower bound e pZ on pZ by solving inequation (*) for with N0 set to the number of samples taken before for the rst counter-example was found, and set to 10 1 as previously mentioned. This yields: e pZ 1 eln( )=N The results for the symmetric unfair case are given in Table1 . The meaning of the column headings is the following: ph is the number of philosophers; time is the time to nd a counter-ex.... ..."

Cited by 17

### Table 4: Internal Relation (symmetrical versions of rules IR7-10 omitted)

"... In PAGE 7: ...n in/read operation. We shall use to range over Act (i.e. sequences of actions). In rules IR12 and IR13 in Table4 , we make use of a complementation notation for labels. It is de ned in the obvious way, namely ot! = ot? and ot? = ot!; as usual = .... In PAGE 8: ...Table 4: Internal Relation (symmetrical versions of rules IR7-10 omitted) before tuple t is actually accessed. Thus out(t):P is rendered as (out(t):nil)jP (rule IR5 in Table4 ), and tuples can be used independently of what the remainders of producer processes do. In Table 3, rule AR1 shows that process in(t):E consumes a tuple ot matching the tuple I[[t]] resulting from the evaluation of t; this causes the substitution, denoted by E[ot=I[[t]]], in E of the free occurrences of the variables in the formals of I[[ t ]] with the correspond- ing values in ot.... In PAGE 8: ... According to the terminology of [38], the PAL operational rules adopt an early instantiation scheme; value variables bound by in/read are instantiated when input transitions are inferred, not when communications take place (late instantiation). In Table4 , rule IR6 shows that eval causes dynamic process creation; eval(out(t).nil) can be used to express the original Linda eval(t), that allowed tuples and not terms as arguments of eval.... ..."

### Table 4: Operational semantics for asynchronous -calculus (symmetric versions of Par, Com and Close omitted)

2002

"... In PAGE 17: ... For any term (name, process, action) t, t denotes the result of replacing each free name a of t with (a), with renaming of bound names of t possibly involved to avoid captures. Operational semantics is given in terms of a labelled transition system, speci ed by the rules in Table4 . The label (action) on a transition P ??! P 0 can be of four forms: (interaction), ab (input at a of b), ab (output at a of b) or a(b) (bound output at a of b).... ..."

Cited by 22