7/30/02 -- AAL 
changed it only to print once at the end of sorting things twice -- because the cycle iteration count was too low otherwise.



first  iteration done at 0x00000706b (28779 cycles)
second iteration done at 0x00000df3e (57150 cycles) (delta=28371)
third  iteration done at 0x000014e11 (85521 cycles) (delta=28371)


==> iter takes 28371/28371 cycles (average = 28371)


since we are producing a bitonic sort of 32 elements 2 times per printing "done", we are producing  
64 outputs every 28371 cycles, so normalized to 10^5 cycles, 64*(100000/28371) = 225.58246

No FLOPS because this is an integer app.

workCount = 459030 / 460464 = 0.99688575
workCount = 452953 / 453936 = 0.9978279
workCount = 452953 / 453936 = 0.9978279



Uniprocessor (7/30/2002):
(Xenon 2.2 GHz, 512MB cache)
10 million iterations (64 outputs/iteration)
99.9% utilization
runtime for 10^7 iterations = 121.61 seconds

We want cycles/iteration:
10^7 iterations/ 121.61 sec * 
64 outputs/1 iteration
1 second /2.2x10^9 cycles
* 10^5 cycles 
= 239.21478 outputs/10^5 cycles


----- Old calculations without -O3 ------

first  iteration done at 0x000010e51 (69201 cycles)
second iteration done at 0x000021b08 (137992 cycles) (delta=68791)
third  iteration done at 0x0000327bf (206783 cycles) (delta=68791)


==> iter takes 68791/68791 cycles (average = 68791)



since we are producing a bitonic sort of 32 elements 2 times per printing "done", we are producing  
64 outputs every 68791 cycles, so normalized to 10^5 cycles, 64*(100000/68791) = 93.035426


flops reported are 0 (not surprising given that this is an integer app) flops, which is 0 MFLOPS, by the way

workCount = 1091749 / 1107216 = 0.98603073
workCount = 1085641 / 1100656 = 0.98635814
workCount = 1085641 / 1100656 = 0.98635814



Uniprocessor (7/30/2002):
(Xenon 2.2 GHz, 512MB cache)
10 million iterations (64 outputs/iteration)
99.9% utilization
runtime for 10^7 iterations = 195.41 seconds

We want cycles/iteration:
10^7 iterations/ 195.41 sec * 
64 outputs/1 iteration
1 second /2.2x10^9 cycles
* 10^5 cycles 

= 148.87114 outputs/10^5 cycles