The arguments for perf are now mostly optional, except for the input file. 
Here is the list:
	- output file will contain stats for each transaction. First line 
in the file tells the meaning of each field
	- victim file is used for one-way measurement so we need the clocks
synchronized. This is only needed for IRC and VoIP.
	- offset is input in milliseconds and it serves if ntp synchronization 
is imperfect. If a clock drifts between source trace and victim trace and you 
know the drift then use this value for the offset and it will help adjust the 
measurements.
	- you can specify either destination and mask (in changeable bits) or
source and mask (again in changeable bits so /24 would have a mask 8)
	- attack start and end are measured in seconds from the start of the 
trace. They are needed because DoS usually has effect only during an attack so 
it makes sense to measure pft only while attack lasts. If you don't specify 
them they default to 0 and 5000
	- monitoring interval should be 1 if you want timed stats every second
	- batch processing is something particular to my tests on a specific 
trace so please ignore this option
	- c flag says if we should match req/rep for DNS and ping based on 
identification. It defaults to true and this should be OK for most traces but 
some traces may have this ID anonymized. In this case req/rep will be matched 
solely (and imperfectly) on timing.
	- human vs machine output is just for the final pft statistics that 
are printed on the screen. It defaults to human.

All tresholds are in tresholds.h, in case you need to change some.
File tg.txt will have throughput and goodput per second for all traffic
going to the same dest (if you use -d flag) or coming from the same source
(if you use -s flag).
File timed.txt will have timed stats of live transactions and failed 
transactions at a given time slot, and the last column is the pft per second, 
i.e. the failure ratio.