This release contains program fixes and enhancements as follows:

Rule file gmake.init-

• Made changes to support more platforms/compilers. Compiler level has been added to structure to allow multiple compilers on a single architecture/OS.




Library XmCreateVA-

• Modified way va_arg arguments are retrieved to correct potential bug on x86_64 architecture.




Programs obc_sr3d1 and obc_sr3d2-

• First release of these two programs providing for sorted binning of 3D OBC data




Program xsb-

• Eliminated calls to XMstr() convenience function.




Program xsd-

• Eliminated calls to XMstr() convenience function.
• Fixed widget code to expose user message info field.




Program phzrot-

• Added capability to restore all mute zones with the -M flag, not just top and bottom. Also reworked the code to use dynamic memory allocation.




Program rnum-

• Changed to allow for the possibility of zero increment for either trace or record variable header renumbering.




Program putval-

• Added two new options: -SP gives the signed maximum peak value and -SA gives the simple signed average. The defaults continue to be the absolute maximum peak value and the average of the absolute trace values. Note that -SA will alter the rms calculation slightly since the mean is removed from the trace values prior to rms calculation. Also the -I has been deprecated.




Program hdrfil-

• Corrected a major logic bug (would not work unless the user was using the trace editting capability). Modernized the code and wrote a XIKP pattern file.




Program wind-

• A change has been made to the tau-p ( -TP ) option. Now the prediction distance as well as the operator length is scaled by the cosine of the emergence angle. This will affect only the highest angles, i.e. those greater then 80deg, but should result in more effective decon in that high angle zone.
• Added -z option to zero out data within the specified window. This is the ones complement of -Z which zeros out data outside the specified window. In both cases the length of the input trace remains unchanged. Also added dynamic memory allocation and fixed a call list problem in the -R restore mode.




Program tvcross-

• New release of program tvcross, which does time varying crosscorrelation of seismic traces.




Program fitter-

• Corrected the -p[] input which should be the order of the function not the number of terms in the function.
• Made the same correction to the 2D surface fit option: now -p[] refers to the order of surface fit, e.g. -p1 is a plane.




Program mixrec-

• Added -Z option to strip out zeros across traces at each time point prior to filtering. This avoids mixing zeros with samples at the mute zone edges.
• Corrected an indexing problem with the -Z option that caused floating point exception messages on the altix architecture.




Program tdecon-

• New program to do targeted predictive decon in either single gate or time varying mode. The strategy is to optimize the prediction filter coefficients for each gate by minimizing the rms ampltitudes in the autocorrelation gap zone by scanning over a range of scale factors applied to the coefs.




Program flatter, flatterX-

• New programs which use a picked horizon to flatten traces, enhanced by crosscorrelation with a pilot built from surrounding traces.




Program pad_lidi-

• Corrected a problem with the pre-stack mode: if the dead traces at the end of a gather (remember the gathers have to be square for this mode) did not have valid LI/DI or record/trace numbers the logic would fail. This would not affect post stack runs.




Programs pol2, pol3-

• Rebuilt the mapping algorithm for computing the frequency domain gaussian filter comb. Unveiled previously hidden gaussian function parameters -aa[] -bb[] -A[] -B[] -str[]. Added optional output of an xgraph file of the gaussian filters used. Added sorting to ensure the eigenvalues/vectors used in the polarization filtering are in proper order.
• Corrected header word menemonic misspelling ToTmAU describing the source component. Also added component descriptions to the man pages.




Program slicetrp-

• Added optional interpolation algorithms for handling nonseismic data like velocities. These options both fill in holes in time slice data and one of them also extrapolates the last nonzero values out to the edges.




Program las2usp-

• Added dynamic memory allocation to trace array to handle very long logs.




Program stack-

• For very large numbers of input records a math error could cause rounding problems leading to the apparent addition of one extra record to the output.




Program backus-

• Added -w[] window length cmd line parameter to the scan mode. This is the length in time over which to test the autocorrelation function for optimum reflection coefficients for all multiple events that fit within the trace length. For example -w36 will cause the energy of the autocorrelation to be calculated over 36ms centered on each bounce of the water bottom multiple and summed for all bounces down the trace. The default is to test the entire autocorrelation.
• Extensive modifications and modernizations. Removed card file input of water depths - it must now come from the trace header entry WDepDP or as a global cmd line override -h[]. Two modes involve the water bottom reflection coeff: a global fixed value -r[] or a trace-by-trace scan mode between two cmd line input values with a search on autocorrelation energy for the optimum refl coeff. This release also corrects some glaring errors in the old (ancient) version.




Program storeit-

• Corrected a dimensioning problem left over from converting to dynamically allocated memory. Corrected a problem with formatted read using the amplitude option. Added verbiage to man pages and pattern files concerning formats of the input files.
• Removed the old putfp2 method of storing amplitude scalers in the -amp option and replaced it with modern header word type detection.




Program taupmig-

• Cleaned up the code for some variable declarations and memory management.




Program specal-

• Modification made to Makefile for various Intel compilers.




Program velint-

• Fixed an array overflow bug in veltrp subroutine where the array index walked outside allocated memory. For whatever reason this bug has not affected the code until x86_64 architectures came into being.




Program mvstak-

• Fixed bug in code in which output line header was written AFTER the internal buffer containing that data was reallocated.




Program decon3d-

• Fixed a coding problem that caused the input traces to be written to the output rather than the processed traces.




Program vttd-

• Added -seisworks option. If present on the command line then the output RecNum, TrcNum entries will be passed unchanged. The default action, install sequential indexing, has not been changed.




Program rmmult-

• Rebuilt the memory allocation structure making the program more maintainable and fixed several coding inconsistencies.




Program radonr-

• Rebuilt the memory allocation structure and fixed some coding inconsistencies.




Program radonf-

• Reworked the memory allocation structure and corrected some internal coding inconsistencies.




Program smedian-

• Extra space for large filenames was allotted.




Program tim2hed3d-

• Fixed a variable mis-alignment in a subroutine call that was causing the -FT and -SM options to segment on some architectures.




Program vomit-

• Added another new input only format [essov23d].
• Added a new input only format [western3d3] to deal with yet another variation on a theme coming out of western.
• Corrected the vomit pattern file. A character missing from the first line caused and error in parsing by XIKP.




Library ut-

• A memory leak was found in the median() routine and fixed.
• Fixed some problems in the library pertaining to variable arguments, failure to include the malloc header file and other misc bug fixes.




Program putsis-

• Fixed an array dimension problem.
• Fixed a bug where ascii fields in the constructed USP line header were non blanked out correctly.




Program filt-

• Fixed an array problem in the -MD minimum phase option.
• Minor correction to the man page.
• Modification made to allow frequency limits to be read in using trace header words (allows spatially variable filtering). On the cmd line instead of -f1[] -f2[] -f3[] -f4[] for ormsby or -fl[] -fh[] for butterworth you put -hw1[] -hw2[] -hw3[] -hw4[] for ormsby or -hw1[] -hw2[] for butterworth. The defaults are Horz01, Horz02, etc since they are floating point words but short integer words ought to work also. All other cmd line args will be globally the same as before. For each input trace filt will then extract the contents of the header words and use these as frequency limits for this trace. The header words can be in any order since the program will internally sort the freq limits to ascending order. Some checking is done for truly absurd values. Putting -V on the cmd line will tell you what freqs got used for each live trace. In addition, the old global cmd line args should work exactly as before.




Programs presort, presortc, & presort3d-

• Corrected potential bug discovered by Roxo.




Program scale-

• Fixed bug in scale where -s[] -e[] parms were set 1 sample off in application.




Program ratdecon-

• Reworked decon code for data complying with the ARMA linear model. The old version was defunct. Added QC output to show goodness of fit for a sample trace. Added verbose output to monitor performance. Also option to auto-choose number of iterations.




Program avepred-

• Fixed a bad logic problem which happened if all the traces of the first record happened to be zero.




Program xpanel-

• Fixed a bug in the pick list option which exhibited itself on the Linux architecture.
• Modified to up the limit on the number of entries in the the "select" box. This was previously 1023 and was exceeded in at least one instance. The order of the items was also changed to reflect the order in which they are submitted in the panel.




Program agca-

• Corrected a problem with hard-coded array dimension for line headers that was too short. Also added dynamic allocations for trace data space.




Program avoan-

• Modified to pass degenerate data rather than abend when it is found, to allow for processing padded 3D data volumes.




Program mvstak-

• Fixed potential dimension problem.




Program xyzin-

• Fixed a bug in the program that resulted in an output value of unity when a single input value was found exactly on the radius of investigation. This resulted in a weight of 0.0 and a NaN in the calculation. For some magical reason it resulted in a 1.0 or -1.0 going out.




Program sis_xy-

• Fixed bug in -R option. Internal array which contains number of elements per function was hard wired at 10000. If more than 10000 traces came in this array would overflow with unpredictable results depending on which architecture you were on. I upped this to 100000 traces for the time being.




Program replace-

• Fixed the -min[] -max[] subroutine control to allow the use of both -min[] and -max[] on the command line. The man page indicated that this was possible but the original code did not support the option.
• Added option to hang a gradient on the first occurence of a value, either on the cmd line or in a trace header word, i.e. either -vo[] or -hw5[]




Program velin-

• Corrected problem with the -rnmo option: the check for zero velocity was still in place even though residual velocities can easily be zero.




Program spike-

• Changed order of parse for -v[] and -v2[] so that -v2[] is picked up first.
• Added contributions to FreeUSP site by Roxo to the distribution. Changes made to velocity definition, man page, pattern file and table.sort
• Corrected problem with initialization of line header fields. This should help eliminate problems with processing and also with print files that rmprint refuses to remove.




Script usp-generic-set-path-

• Modified the script to properly find Linux 2.4 binaries when running on a machine with a 2.6 kernel. Also added blocks for intelligent location of shareable libraries among products.




Misc programs and library routine fold()-

• Makefiles for these applications using a common fold() algorithm from the USP utility library have been modified to use a version which utilizes SGI performance library code when possible.




Program laip-

• Make printout file lower case so rmprint will not remove it. The user must now explicitly delete that file if so desired. The printout file contains too much information that is required later in processing to have rmprint nuke it by accident.




Program usp2segy-

• Fixed a bug which occurred when the USP line header had an HLH greater than 9000 bytes.
• Bug found which affected header value translation on 64 bit Linux architectures. The output segy data header values were potentially incorrect.
• Fixed a reported bug in both usp2segy and segy2usp which prevented setting of header values to a constant 0 on some architectures.
• Fixed a bug which could cause a segmentation fault if the input USP dataset line header exceeded 10000 bytes in length.




Program avgain-

• Corrected bug in code in handling of dead traces when -agc was used.
• Fixed another bug regarding dead traces in -agc option. Traces which were dead but had not been marked are now killed in the processing output.




Program saneusp-

• Corrected problem in the spectral output routine that handles leading/trailing zeros in the signal and noise spectra.




Program vred-

• Added -A option to force trace distances to be positive. 3D data most always has positive trace distances but sorted 2D can easily have both positive and negative distances in DstSgn causing the vred shift to reverse sign, which is not always what is desired.




Program attin-

• Found a program bug which caused program failure on occasion.




Program etch2d-

• Fixed a potential bug in the code.




Program bytknt-

• Modified the XIKP pattern file to make the default xcat window if run from within XIKP a transient window. Otherwise, the xcat window will hang net execution until dismissed by the user. In addition, a few code changes were made to avoid potential bugs.




Program bdnmo-

• Added dynamic memory allocation to the input trace and velocity data buffers to be able to handle very long traces.




Program rand-

• Completely reworked the random statics functions. There are 4 new header word entries: -hws[] -hwr[] to detect shot and receiver indexing (when a change in either indexing is detected a new random static is computed) and -sws[] -swr[] for storing the shot and receiver statics respectively.
• Added -n option to use a gaussian normal random number distribution. The default is a uniform distribution.
• Added automatic generation of the seed for the internal random number generator. This is now the default. To get exactly the old random number sequence use -C911 on the cmd line.




Library ieeeio-

• Fixed some problems in the I/O library pertaining to 64 bit and logic concerning proprietary compressed formats




Program filt-

• Added an option to output only the filter response and then stop. -I triggers this option and the length of the response in ms or us can be specified with -L[] (def = 201 samps). The response will be centered within the single trace output and will consist of the response of whatever filter the user specifies to a spike of unit amplitude. A -N[] input file is still required.
• Found a minor bug which could sometimes cause the program to falsely report that requested memory was not available.




Program mute-

• Corrected a bug which occured with the flat file format -off mute: a zero mute time would leave the trace untouched. A zero off mute time should mute the entire trace




Program smute-

• New option -threshold to allow thresholding rather than just a straight mute in the surgical mute zone.




Program slvr-

• Fixed a segmentation problem.
• Corrected a problem reading vds3d format velocity file: standard vds3d fmt has a header line starting with 1MC3D. slvr would misinterpret this and then blow out with a segmentation error. slvr now checks to see if the MC3D header line is there and bypass it.




Program rhgain-

• Added mute replacement feature to prevent unintended brightening due to zeros in the top and/or bottom mute zone.




Program windstat-

• Reworked the man page, pattern file and program help for the -P12 and -P13 options to indicate that they output peak amplitude at the time of largest positive or negative value. To get those times use -TLP or -TLN. Also fixed reference to -MTA which is really -TMA in the code.
• Added functionality to -RMS option to allow the calculation without utilizing floating point horizons.
• Added two new attributes; -ABF which gives average value between horizons defined by floating point samples as opposed to AVA which does the same by defining the window to the nearest integer sample. Also added -RMS again to calculate the RMS value between floating point sample positions.




Program gentrp-

• Fixed a bug in the code when running under XIKP. Command line arguments passed by execution under XIKP caused gentAdded off mute detect/restore (det/res mute from the bottom of the trace, up)rp to falsely complain about invalid arguments.
• Removed policeman in code that prevented out number of samples from exceeding an internal limit. Now can handle any number of samples per trace on input.
• Added off mute detect/restore (det/res mute from the bottom of the trace, up)
• Previously did not allow the input sample interval -i[] to default to the input line header value (SmpInt). Also if the user wishes to use microsecs he must explicitly flag -M and use microsecs for -i[] and -s[] (input and output s.i.). The old automatic trigger for us if either or both -i[] or -s[] were less than 1.0 is disabled. If the user uses any value less than 1.0 the program exits with an error message. Also if either of these entries contains a fractional part the program issues a warning that perhaps using -M and microsec values would be desirable.




Script usp-set-env-

• Modification made to the USP environmental variables so that the scheme used in naming the USP print files is now used for all processing.
• Corrected problem which occurred when program auditting was turned on.




Script rmprint-

• Modification to rmprint script to catch printout files created on platforms where machine names are of form compute-N-N.local.




Program rest-

• Fixed a nasty feature that had to do with the interaction of -nullvalue[] with the line header entry HrzNul. The intent of the -nullvalue[] command line entry was to be able to override the value present in the line header of the input dataset. These values are used with SSAM and other attribute calculation scripts in the USP world. The override was not working. Even worse, the program and the man page made reference to a -null flag that was not used in the program. The user was left with the mistaken impression that no null based activity would take place in the code unless -null was flagged. Such activity was and still is the default, there is no way to shut it off, there is only a way to override the horizon null value on the command line if desired. Any trace with a horizon value equal to the null value will be passed unchanged as far as the time series is concerned. The StaCor entry in that trace header will be set to 30000 so that future calculations etc. will see it as a dead trace.
• Basic functionality has not changed. However, upon suggestion, the phase option now simply applies whatever is found in the trace headers and uses the exact phase rotation code as phzrot (before there was a 1 part in 10^5 difference due to a difference in the FFT calculation). Also the -F floating point flag has been deprecated since rest knows which header word is a float or integer or short integer.
• Added extra explanation for the -u[] time scaler cmd line entry in the online help and the man pages and the xikp files.
• Addition of "-L", linear mode option; for use with non-seismic data, e.g. velocities. A linear interpolator is used to super sample the data and then an integer shift is applied.
• Fixed a dynamic memory allocation bug which affected some datasets with a memory overflow internally.




Program ufh-

• Modified for swapped byte conditions. Fixed some missing lex variables.




Program zombie-

• Added policeman to prevent execution on datasets whose input NumTrc is too small to support the zombie algorithm, for instance single trace records.




Program holdif-

• Added cmd line flag -S to switch stdin from the default dataset 2 to dataset 1 to allow alternate pipe-in capability.
• Outputs N2 in place of N1 only if N2 StaCor not equal to 30000 (ie. live trace)




Program edit3d-

• Fixed bug in record length assignment of direct access vector database file. Also syncd this routine with dbvec for the number of entries read from the database header record.




Program dbvec-

• Fixed a major bug in the allocation of the direct access file used to build the vector database. Old code would crater if the number of bin in the grid system was greater than the number of traces coming in. Also syncd this routine with edit3d which was reading more from the database than dbvec was storing.




Program rwspec-

• Upgraded this code to utilize dynamic memory allocation for all input and output arrays. This allows analysis of very long traces.
• Fixed bug in Decompose routine that walked off the top and bottom of the input trace array by half the window length.




Program segy2usp-

• Found a byte count that was wrong, which could result in the first two bytes of the "spare" area of the SEGY line header getting swapped. This could affect values stored there if bits were set in either of these two bytes.
• Fixed a reported bug in both usp2segy and segy2usp which prevented setting of header values to a constant 0 on some architectures.
• Fixed a bug that caused some SEGY header words to be improperly read in on 64 bit little-endian machines.




Program sr3d2-

• Fixed a bug that showed up in the -off option when the corner coord system included the origin and negative values of X. In this case the last live trace of the first output record was erroneously repeated until the end of the first offset record and for all bins of the second offset record for which the X coordinate was negative.




Program angst-

• Added another optional output data set, -F[]. This will contain the stacking fold at each sample position. There will be one fold trace for each -O[] output stacked trace.
• Found a bug which caused memory alignment problems. There should have been no effect in processing as it only happened on traces that were to be killed.
• The program now assumes that the sample interval and UnitSc parameters are set correctly. The program will ensure that the UnitSc and sample intervals are the same between the seismic and velocity files.
• Added check for any zero sample in the input usp format velocity data set. The program exits gracefully with a message to the terminal and the printout file. Zero samples in the velocity traces have sometimes been a problem in the past.




Program rept-

• Added option -T to take single trace records and pad them out by duplicating the trace nr times. Total number of records will remain the same as input.




Program seltrc-

• Fixed a bug in seltrc which resulted in the -L option being turned on all the time by default. To pass only live traces you must now actually use the -L option.




Program killtr-

• Added option to the kill file to read in a record number followed by a range of traces to kill, e.g. 11 5 9 will kill TrcNums 5 to 9 in RecNum 11. The new option requires -hw1[] and -hw2[] entries.




Program merge-

• Modification made to gather (mergeNto1 in XIKP parlance) to allow input piped data to be allowed when "-N stdin" is flagged on the command line. This allows for more versatility in usage within scripts which may then be run as part of an XIKP net or connected with other scripts and/or individual processes.
• Added an option to specify the input dataset from which the output line header will be selected. The default is the last input dataset, but by using the new argument -lh n, where n is between 1 and the number of inputs, the line header from the n'th input is used.




Program splitr-

• Fixed a minor bug where running under XIKP would yield a SPLITR* printout file which could have non-printable characters in it, which causes rmprint to not remove it.
• Added new feature to allow connection to stdout even when using multiple -O output connections. This is done via specification of "-O stdout" as the first output and then either redirecting stdout or connecting descriptor 1 within XIKP.




Program ttds3d-

• Found a bug in the ut library code that puts float values in USP header words normally containing short integers. The ttds3d code was using this routine to put values starting at line header words 'ReSpFm' and 'RATTrc'. This bug only affected little-endian architectures.
• Updated the start script for ttds3d so that the temporary files on LINUX will be opened in the same directory where the command is executed. Previously, on that arch only, the scratch files were opened on /tmp which was a disaster when working with very large files.




Program scan-

• Added option -vh to scan which, trace by trace, prints the ordered samples within each histogram bin.




Program shape2-

• Corrected a 1-sample shift in the ouput.




Program XIKP-

• Fixed a bug in the code which prints out the informational message when a remote process has died during a batch processing run. A variable was being used before being set in all cases and the message could sometimes not get displayed.
• Modification of both client (ikp) and daemon (ikpd) modules to run under the Sun Grid Engine. Also enabled a more portable method of determining load on systems.




Program cross-

• Corrected error on help screen
• Added option -lim for convolution that automatically outputs only the number of samples per trace on data set N2 (the default is to output N1+n2-1 samples). In most cases the N1 data set is the wavelet or operator and the N2 data set is the data being operated on.




Program spacetrp-

• Fixed memory leak in spctrp subroutine; also changed cubic splice routine to reinitialize for headers after every record. This was required to deal with the case when a dataset starts with many dead records.
• Added optional restore early mute. Caution should be exercised when using it. If the mute function is well behaved spatially then it should get interpolated properly; if the function is erratic however the interpolator might run off the rails.




Program stolt -

• The IKP pattern file for stolt had a mis-typing of the -aper argument that forced you to enter it as an integer. This has been changed to float to match the actual code




Program timfreq/freqtim-

• Identified a confusing point about these two codes: they use the complex packing scheme which uses N/2 complex values, where N is the next power of 2 input samples. The assumption here is that nyquist is zero. While selfconsistent (full circle transforms give you back the input) the usual packing scheme is to use N/2+1 elements which allows for the nyquist to be stored as a real value. timfreq and freqtim have been modified to use the more standard packing. Beware: old timfreq data will now no longer be transformable by the new freqtim.




Program timfreq -

• The trend option always subtracted zero from the input phase due to an indexing error in the code.




Program rmprint -

• Modification to rmprint script to catch printout files created on cluster nodes where machine names are of form compute-N-N.local.
• Modified to correct a problem which occurred when a large number of print files were encountered in conjunction with the "-i" option.
• Fixed a bug which affected execution on the SGI systems. Also modified the way the script processes the files and added a new option "-s" to skip the printing of messages about bypassed files.




Program hdrswap -

• Increased the amount of error info sent to stderr.




Program clean_green -

• Added byte rotation logic necessary for running on Linux systems. Also modified printout in verbose mode to identify the sequential record/trace numbers of traces with bad green words to cover cases where the rec/trc numbers in the header may also be corrupt.




Program fxdecon -

• Changed the -pc[] option to add back a percentage of the noise data rather than the input data
• There was a bug which occurred when the frequency cutoff was less than nyquist. An array was not being cleared which resulted in a nasty notch or step in the output amplitude spectrum and which no doubt would lead to phase problems in the output.
• Fixed access error and memory leak, which only seemed to affect large datasets with large numbers of traces per record.




Program dslice -

• Modified to handle byte order correctly on little-endian machines such as Linux. This code does not use the regular I/O mode because access is done on a word-by-word basis during the rotation.




Program tfdnoise-

• Bug fixes in the median threshold calculation that cropped up when the number of samples in the calculation was less than 3. Added -biasend[] option for use with -hwt[] and -bias[] to control a variable replacement window.




Program vomit-

• Updated man page and XIKP pattern file.
• Fixed bug in USP -> TDQ conversion that would prevent -force from activating if chosen on the command line.
• Added a new output format [flat3d2] - [LI, DI, X, Y, unit, vel]. Also fixed a nasty memory leak that caused a segmentation fault on LINUX boxes when using USP data format as input or output.




Program recselect-

• New routine to do user defined splitting of an input data stream into two outputs. Splitting is controlled by the contents of a user defined trace header and user defined command line control.




Program qest-

• Corrected major bug which prevented the slopes being calculated from the spectral peak (were being calculated starting from zero frequency). Also corrected a bug in the time-weighted Qeff calculation: time now starts from the water bottom not the surface since the Q calculations also start from the water bottom. The water column time is no longer put int VPick1 to avoid clobbering prior WB time picks.




Program asig-

• Fixed severe memory bug in pikval1 subroutine which crashed app on IRIX platform when using -nop 7.
  




Program qint -

• Initial FreeUSP release of program qint, which is used to compute interval Q given effective Q.
  




Utility which -

• Modifications to the USP "which" script used for locating an actual execution binary that allows optional specification of product and version.
  An example execution would be "usp -product xaprtools -version test xbtnpress" (This script is only executed via the use of the "usp" utility script to avoid conflict with the OS which utility.




Program pad_lidi -

• Changed the padded traces so that TrcNum gets updated as well as RecNum.
  
• Corrected bug that could cause the output DI's to get out of sync




Program sctvf -

• Corrected several coding errors; changed the oversampling engine to be a sinc interpolator (for the t-v option).