Very strange measurements

[Brecht Vermeulen]

> Maybe you could take these numbers for at least two loads, heavy (=low
> latency result) and light (=hi)? And any ToDevice statistics? If it's not a
> big deal?
> 

the linux scheduling is still on 4096.

light load (10 MBit, 64 byte packets)

latency: min	avg	max	# packets
10.0	14.2	4407.1	9331.3	148,809 

PollDevice :
148809 packets received
15153571353 cycles poll
14347536 cycles refill
46661811 cycles allocskb
6703385240 cycles pushing
0 perfctr1 poll
0 perfctr1 refill
0 perfctr1 allocskb
0 perfctr1 pushing
0 perfctr2 poll
0 perfctr2 refill
0 perfctr2 allocskb
0 perfctr2 pushing
149434896 empty polls
148805 activations

ToDevice
0 packets rejected
8002 hard start xmit
9590366 device busy returns
148809 packets sent
15153421 cycles pull
55584353 cycles clean
51685042 cycles freeskb
22410571 cycles queue
0 perfctr1 pull
0 perfctr1 clean
0 perfctr1 freeskb
0 perfctr1 queue
0 perfctr2 pull
0 perfctr2 clean
0 perfctr2 freeskb
0 perfctr2 queue
83639 transmit activations

High load (70 Mbit, 64 byte packets)

Latency:
	min	avg	max	# packets
70.0	16.4	40.7	10010.4	1,038,709 

(hmm the avg is quite good now, not 20us as in the other machine)

PollDevice
1040594 packets received
4690537360 cycles poll
76346199 cycles refill
141668725 cycles allocskb
3831047093 cycles pushing
0 perfctr1 poll
0 perfctr1 refill
0 perfctr1 allocskb
0 perfctr1 pushing
0 perfctr2 poll
0 perfctr2 refill
0 perfctr2 allocskb
0 perfctr2 pushing
76933218 empty polls
1038530 activations

Todevice
0 packets rejected
31 hard start xmit
58942 device busy returns
1038709 packets sent
102081505 cycles pull
345637295 cycles clean
80526459 cycles freeskb
135846534 cycles queue
0 perfctr1 pull
0 perfctr1 clean
0 perfctr1 freeskb
0 perfctr1 queue
0 perfctr2 pull
0 perfctr2 clean
0 perfctr2 freeskb
0 perfctr2 queue
1034375 transmit activations

regards,
Brecht