Usage
Step 1
 Convert the data-trace (outgoing) to ASCII after filtering on TCP. Use absolute sequence numbers first, then sort the trace on connections, then use the sack_rel_seqno program to convert that to relative sequence numbers.
Step 2
Convert the ack-trace (incoming) to ASCII after filtering on TCP. Again, use absolute sequence numbers first, then use the 2wayprep program, just to revert src and dst. Then sort this, and convert to relative sequence numbers using sack_rel seqno.

The steps for ACK stream is similar to the DATA stream except that we use  the 2wayprep program before sorting in the above step. The 2wayprep program simple reverse the source and destination. This allows us to compare the same column later when we try to combine these two streams.
 
Step 3
Now use extract_dat_seq_v3 to convert the sorted data-trace to output one line per SYN, DATA and TRM (FIN). Use sack_extract_ack_seq_v3 to do the same for the ack-trace. Then merge the two to obtain a sorted file, sorted on connections, and time stamp within the connection.

e.g. for step 1 and 2 and 3 for data pkts

gunzip -c $1 | tcpdump -S -n -tt -r - tcp | egrep '^[0-9]+\.[0-9]+ [0-9]{1,3}\.' | sort -s +1 -2 +3 -4 +0 -1 -T $3 | sack_rel_seqno  | extract_dat_seq_v3 |gzip -c > $4.dat
where
$1 is the input dump file to use for data sequence numbers
$3 is the path of a directory to use for sort temp files
$4 data file

e.g. for step 1 and 2 and 3 for Ack pkts

gunzip -c $1 | tcpdump -S -n -tt -r - tcp | egrep '^[0-9]+\.[0-9]+ [0-9]{1,3}\.' | sort -s +1 -2 +3 -4 +0 -1 -T $3 | 2way_prep |  sack_rel_seqno  | sack_extract_ack_seq_v3 |gzip -c > $4.ack

sort -s -o $4.seq_sort +1 -2 +3 -4 +0 -1 +5 -6 $4.ack $4.dat
Step 4
At the end of this step you would have a single file with both the ack and the data packets sorted in this.

Now this is ready to be used by the analyzer. We have to run the above file through each of the 4 different OSes.

e.g.
zcat $4.seq-sort  | ./{OS}_machine  -c t.a.log -w t.a.rtt -v t.a.viol -mi t.a.manual -t {OS_GRANULARITY} -ret t.a.retran -u t.a.unneeded -tcpret t.a.tcp_retran -rec t.a.recover

t.a.log file is a log_file with one line for each connection in the trace.
t.a.rtt file is a detailed log file with one line for each packet in the trace.
t.a.viol file is a file which records the cwnd and advertised window violation. (the cwnd window is still under testing)
t.a.manual file is similar to above (it was meant for debugging purpose for earlier versions of the code)
OS_GANULARITY is the granularity of the clock for each OS (its 0.01 for everything but windows, windows had granularity of 0.1)
t.a.retran file has once line each for a OOS segment which is a real retranmission.
t.a.unneeded file has one line each for each of the unneeded packet
t.a.tcp_retran has one line each for a OOS segment that TCP thinks is a retranmission (it may be different then the t.a.retran file as TCP may mis-judge whether a packets is required retranmission or not)
t.a.recover has once line each for each of the loss detection, recovery and normal phase of TCP. (detail available in  Empirical Analysis of TCP Losses and Its Detection/Recovery Mechanisms

All the Files are available here