Explain what phantom arms are and why they happen. Why is it possible that this could lead to bad concurrent processing even though the two-phase synchronisation method is being used?

For instance, a systolic array is a kind of array that may be thought of as a MISD device. Systolic arrays are networks of data processing nodes organized in a wavelike fashion, thus the name. None of these parts need the employment of a program counter since execution is started by the arrival of new data. In a timed systolic array, all of the nodes calculate and communicate synchronously, or "in lock-step," with one another.

A MISD machine is an example of a systolic array, a kind of array. Data processing units are organized into a "wavefront" in a systolic array. All of these components do not need the use of a program counter since execution is initiated when new data is received. Lock-step computation is achieved in a clocked-systolic array, where each unit completes its alternating calculation and communication phases in the same clock cycle.

Has to do with amdahls law.   Lets say I want to parallelize a program on 10 nodes. Assuming 40% of the program cannot be parallelized what is the speedup that can be achieved by performing a perfect speedup on the 10 nodes vs the 1 node.