原廠網址:http://support.tsunamidevelopment.com/home
鴻鵠國際為台灣地區官方合作廠商,並且可提供技術支援與收費的教育訓練
歡迎來電或來信諮詢。
PSTM & PSDM
B2012 version pstm_6.2 and psdm_6.2
Better Port Connectivity
Previous versions of Tsunami would choose a starting communication port number at random, and then increment the port number for each compute node. If the port was in use, the job would fail. While this rarely happened, it could fail sometimes.
The current version checks for available port numbers before deciding which ones to use. This almost completely reduces the chances for a failure, due to ports being already in use.
CDP Numbers in the output file
Tsunami will retrieve the CDP numbers from byte 21 in the input data and place it output file. For every input trace with an input trace, it will copy the CDP number into a CDP map; if an input trace does not have a CDP number it will determine the inline and xline increments and extrapolate the CDP number. The only exception is when the xline numbers are also in byte 21, which will ignore the CDP numbering.
CDP numbering is usually based on some combination of the inline number, and number of xlines in the survey. If the number of xlines output from Tsunami is larger than the number of xlines in the survey, then the CDP numbers on the edge of the inlines may be duplicated, or meaningless.
Better communication between server and compute nodes
There is a recent change in the system command to expedite transfer of data from the server to the computer nodes. This is possible due to Tsunami no longer supporting SGI IRIX operating system, which allowed for clusters to contain nodes in both Big Endian and Little Endian format, slowing the transfer rate.
We now see network data transfers on GigE networks as high as 80 mbytes per second. This is approximately a 2x improvement.
Trace.db weights
The weights for the trace.db file were being recalculated for every job, sometimes unnecessarily. This has been changed so that weights are recalculated only when needed.
Fixed scaling weights problem
A bug has been fixed that scaled far offsets incorrectly, resulting in high amplitudes.
Fixed 1/t scaling bug
Previous versions of Tsunami required that the 1/t2 scaling power parameter (sphdiv) be an integer. Now any real number can be used.
Skip Initializing Output
Initializing of an output file ensures there is adequate space on disk to handle the output. However, with large jobs and large number of computer nodes, the initialization phase could add to the job’s run time. A checkbox has been added to skip initializing of an output file for both PSDM and PSTM.
In addition, this option will shorten run time by writing output traces directly to file. This implies there will be no check pointing of an offset group, as the output traces are not being written to file based on trace counts.