$11.(5)$ A multiprocessor system contains $32$ nodes. Each node has a processor and a

local memory. All processors have a $0\mathrm{.}5$ ns cycle time. When executing a certain

application, the multiprocessor system achieves an effective or average CPI of one

instruction per cycle when there is no communication required between nodes.

Each instruction that requires communication with a different node incurs a penalty

of $400$ extra nanoseconds.

Suppose $0\mathrm{.}4\%$ of the instructions that are executed communicate with other nodes.

None of the remaining instructions communicate with other nodes. What speedup is

provided by somehow eliminating all communication between nodes?