Well,
There are of course several ways to show threads in sequence diagrams.
Usually I start modelling threads (or thread of control) with one composite (structured) class that controls and maintain all thread specific attributes (such as thread prio, sched policy, affinity, name etc). This "Thread Boundary class" defines also the refined software architecture in its inner structure where all contained parts executes within the thread context of the Thread Boundary class. If you then add a sterotype ,i.e <THREAD_A>, to the Thread Container class this will show up nicely in the sequence diagram.
One example of how to model the actual parallel execution is shown in Figure 17 in
http://www.ibm.com/developerworks/ratio ... /3101.html.
In short:
1. Model one execution block, within the object of the same thread, with one interaction operator
2. Set the type of of this interaction operator to "parllell" (you can also define a "PARALLEL" sterotype for these types interaction operators")
3. Add execution occurances
4. Add operand seperators for the different parallel activities
Best Regards//
LGJ