We carry out a similar procedure for the remaining two test cases and summarize the results of all three in Figure 11.14:
Interesting; though the maximum latency of the RTL kernel is much below the other standard kernels, both the minimum and, more importantly, average latencies are superior for the standard kernels. This ultimately results in superior overall throughput for the standard kernels (this very same point was stressed upon earlier).
The latency_test.sh bash script invokes the gnuplot(1) utility to generate graphs, in such a manner that the title line shows the minimum/average/maximum latency values (in microseconds) and the kernel the test was run upon. Recollect that test case #1 and #2 ran on the Raspberry Pi 3B+ device, whereas test case #3 ran on a generic...