Above example indicates 9.400 microseconds (us) is used for 10 iterations execution time. Each iteration execution time is 0.94 us or 940 nanoseconds (ns) in average.
Per output file ulbf_out0.txt, 38400 lines for 100 iterations. Each iteration processes and outputs 384 lines. Each line has 2 cint16 samples.
Performance calculation:
1,000,000,000(AI engine frequency in HZ) / 940(clock cycles each iteration) x 384(lines each iteration) x 2(samples per line) = 817,021,276.59(samples/second). This number is close to profiling API reported, 818,527,715.90 samples/s.