Dimitre> Seems like some people didn't understand what I said. Is Dimitre> it too complicated to understand that "the results Dimitre> happens not earlier than the reason for it" ?

But the examples you gave relied on more than this.

Besides - it isn't necessarily relevant to the problem. In a speculative parallelism implementation, the results may well be calculated before the need for them arises.

In fact, this is something I intend to explore next year - evaluating global variables in parallel prior to any reference to them. -- Colin Adams Preston Lancashire