next up previous contents
Next: Sequence non-empty Up: Sequence parallelization Previous: Sequence parallelization   Contents

In case we have no parallel tasks currently

We introduce yet another marker (or pointer) into our code. We want to find the instruction I, that succeeds T if T was actually found in the code we know about, but is either before or equal to the start of our main sequence. In the case where we faked T to point to the beginning of the main sequence, we simply set I to point to the beginning of the main sequence.

So summarize: If T is found in code we know, then T $<$ I $\leq$ beginning of main. Otherwise, T $=$ I $=$ beginning of main.

If the sequence of instructions $[\mathsf{I};\mathsf{C}[$ is non-empty, it means that between the current opcode and the opcode that touches what we depend upon, there is something else which we don't care about. This is where we start two parallel tasks !