Can GUIDs be defined in a more precise way?
Specifically, in order for an execution trace E to be valid, it needs to
obey unithread semantics, but the obvious way to do that is to execute
the program in a sequentially consistent machine (where the reads return
the same values as specified in the execution trace), and see if the
resulting trace matches E, but to do that I need to be able to determine
which read in the execution trace corresponds to the step in my
sequentially consistent machine deterministically generate a GUID.
Would a pair (ThreadID, Counter) where ThreadID is some unique
identifier for a thread, and counter starts at 0 and increments each
time an instruction is executed be sufficent?
Hope this makes sense,
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:42 EDT