Renamed SOCKET transport type to TCP.

first version (not yet working) of DFT imager code.

Added entries that have the phase centers in the middle of the *stations*, not
the middle of the *microstations*.

do not cause compile errors on machines that do not have MPI

remove untested transpose optimisation

Add support for UDP sockets to the inputsection. Also rename the SOCKET input method to TCP and introduce the UDP input method.

Avoid type-punning pointer warning.

Set precision.

CS1.parset
 Added option 'Observation.SubbandIDs'

CS1_Parset.py
 Added lines for calculate the RefFreqs from array of 'Observation.SubbandIDs'

LOAFR_Parset.py
 Corrected function getInt32Vector()

CS1_Stations.py
 Added centre positions for Beamforming

CS1_Sections.py
 Changed BGLProc bin name in CS1_BGL_Processing

CS1_Run
 Corrected NoRuns

Cleaned up much code, including replacement of RectMatrix by boost::multi_array

Bug fix: count the number of subbands correctly.

Increased buffer size

In function getObservationEpoch() the variable sz must be equal to noRuns.
These variables must be a multiple of 16, plus 16.

Changed source MACs of stations and dest MAC of lii001

Fix bug in administration of flagged data.

Better diagnostics

In process() first call calculateDelays() and the increment the loop counter.
Otherwise an incorrect assertion happens.

Fix byte order bug for big-endian machines.

Major cleanup

Removed time variable that was always zero

* Removed useless operators
* templated increment/decrement operators
* added cast operator to int64

Fix postfix -- operator.

Use MPI_Alltoallv instead of CEPframe's point-to-point communication.

The problem with writing 2 subbands in 1 MS is solved. The sample times of the
subbands are correct. Results, see table below. 1 group is a set of all
baselines from 1 subband (t0=1 means start time plus 1 minute):

Group       Band   Sample time
  0          0      t0
  1          1      t0
  2          0      t0+1
  3          1      t0+1
  4          0      t0+2
  5          1      t0+2
etc.

The uvw-coordinates will be calculated by using Aips++ libs
instead of doing it by your self.

Do logging in separate thread; printing from signal handler causes deadlocks.

1011
1011
1011
1011
1011
1011
1011
1011
1011
Fixed wrong argument order when creating MSWriter object

Quick fix. Not yet fully tested.

Added transpose.cc

Test to see if the BG/L torus is fast enough for moving the transpose from the
InputSection to BGL_Processing.

It is not necessary to print-out invalid samples of every outputs from
the same station.

Making changes to support MFS

* shift subband by half a channel due to the PPF in BGL_Processing
* put #if defined HAVE_MPI at appropriate places

Take care of the half-channel shift introduced by the PPF.  Also flag channel 0,
which does not contain valid data.

Changed:
 The name of the MS is subband-dependent.

Changed:
 1 - Filled in the name field of table SPECTRAL_WINDOW with value "SB-x".
     'x' is the subband number started by '0' (index array of Observation.
     RefFreqs in the parset file.
 2 - The name of the MS is subband-dependent.

Major changes to the correlator, which now does not depend on compile-time
constant number of stations and integration time anymore.  Number of PPF
taps, polarizations, and subband channels is still constant, but not expected
to change in the near future.

Not necessary anymore

Added counter in fuunction proces().

Enabled logging data flags.
Changed: subbandsToReadFromFrame <= Input.NSubbandsPerFrame in stead of subbandsToReadFromFrame < Input.NSubbandsPerFrame