ATNF VLBI Wiki

slalib is not needed if you replace libate by a dummy routine:

      SUBROUTINE EPHMRS (RP_UTCMTAI, RP_C, RP_DJMREFP, RP_DJMREFT,
     *        XYUT, NXYUT, CPOS, NCPOS, SU_RA, SU_DEC,
     *        EPOCH, PR, PD, APLX, RVEL, DPLX0, NANT, PV, GOAD, PLATE,
     *        EPOCRD, RP_PRESSURE, RP_TEMP, RP_HUMID, TLAPSE, TRPAUS,
     *        AXIS_OFFSET, PHREF, TAI, INTGRN, RP_DEFEAT, RAP, DAP,
     *        TE, DELAY, DEL1D,  DEL2D, VEL, B, PARLCT, SW, SW1D)
        write(*,*) 'CALL to libate!!!'
        stop 1
      return
      end

make libate.a from the relocateble. This should never be called for DiFX data.

In order for your checked in source files to come out with SVN repository version and other useful information, you can do the two following steps:

1. Add stub comment code to yor files, making sure to use legal comment characters. Example for c++:

//===========================================================================
// SVN properties (DO NOT CHANGE)
//
// $Id: $
// $HeadURL: $
// $LastChangedRevision: $
// $Author: $
// $LastChangedDate: $
//
//============================================================================

2. Set the properties within svn to cause replacement using the following command line

svn propset svn:keywords 'Id Revision LastChangedDate LastChangedBy LastChangedRevision' example.cpp

Three packages within DiFX depend on IPP: mpifxcorr, vdif_server and difxfilterbank. These instructions apply to all of those.

1. Make a file called ipp.pc and install in ${DIFX_PREFIX}/lib/pkgconfig . The following template should be modified to reflect the path

base = /home/swc/difx/intel/ipp/7.0.1.084/ia32
 
Name: ipp
Description: Intel Performance Primitives
Requires: 
Version: 7.0.1.084
Libs: -L${base}/lib/ia32 -L${base}/ipp/lib/ia32  -lipps -lippvm -lippcore -liomp5
Cflags: -I${base}/ipp/include

2. Edit the configure.ac (or configure.in) file for each package, replacing all of the IPP configuration section (ending in fi) with the single line

PKG_CHECK_MODULES(IPP, ipp)

3. Make sure paths to both library directories within the IPP install are in the LD_LIBRARY_PATH. In the case of the Intel 32 bit version of IPP version 7, this means, for example, replacing:

PrependPath LD_LIBRARY_PATH  ${IPPROOT}/sharedlib

with

PrependPath LD_LIBRARY_PATH  ${IPPROOT}/lib/ia32
PrependPath LD_LIBRARY_PATH  ${IPPROOT}/ipp/lib/ia32

in the setup.bash file.

4. Note that for 64-bit installations, replace ia32 in the above instructions with intel64 and you might want to change the version number in the ipp.pc file to match more accurately the actual version installed.

Often it is useful for benchmarking purposes (of either the network or the CPUs) and for debugging certain problems to be able to feed DiFX with a source of fake data. This has been possible for some time with the difxfake program (by Chris Phillips). An equivalent feature has been added straight into DiFX version 2.1 and later. This is enabled by changing the DATA SOURCE entry for each datastream to FAKE. If set, DiFX will ignore the data table and simply flow a perfect stream of senseless data into the correlator. Everything else about correlation will remaing the same: you will still need a .machines file, a .threads file, and all of the other files. The output can be assembled into a FITS file if desired. At the moment no attempt is made to produce test vector data with any particular properties. If there is eventual demand for using this for correctness testing that won't be a difficult task. One could even add to the simulated data bandpass effects, switched power, pulse cal tones, òÀæ, but at the moment these are only posibilities.