Description of individual benchmarks

Version 1.9 Build 1556

News	FAQ
Search	Home

Next: benchmark - Tool Up: benchmark - Module Previous: Overview of benchmark tool functions:

Description of individual benchmarks

Each benchmark is identified by a unique benchmark code. These are hierarchical, and specify all relevant parameters applicable to the benchmark. This section includes any scripts from other packages which perform the same reduction. Package inter-comparison is subject to the caveats mentioned above however.

Clark Clean single-field imaging

This benchmark has the following code descriptor:

CC-SF-data -SP win - stokes - weight - npixel - nclean

where data is the input data description, further broken down as data= [instrument][size][compression][obsmode]. The parameters are specified as follows: i) instrument - telescope abbreviation (VL=VLA); ii) size - dataset size on an integer scale of 1 to 10 (4 $\sim$ 630k UVFITS visibilities); iii) compression - MS compression (U=uncompressed; C=compressed); iv) obsmode - observing mode (C=continuum, L=line); v) win - number of spectral windows imaged; vi) stokes - Stokes parameters imaged (e.g. IQUV or I); vii) weight - imaging weight (UN=uniform, NA=natural); viii) npixel - image size in pixels; vix) nclean - number of Clean components in deconvolution.

The individual Clark Clean single-field benchmarks implemented in the default run of the ccsf () method at present are enumerated in the tables below. The common portion of benchmark code mentioned in the table caption. The variable part of the code is listed in the ``Benchmark'' column of the tables. Full benchmark code can be constructed by concatenating the constant and the variable part of the code

**Table 1.862:** Table for continuum VLA 125K rows dataset benchmark. Common part of the benchmark code: 'CC-SF-VLAC-U125K-SP1-'
Benchmark code	Dataset	N_spw	Stokes	Wgt.	N_pixel	N_chan	N_clean
I-UN-512-C1-1000	vlac125K	1	I	UN	512	1	1000
IQUV-UN-C1-512-1000	vlac125K	1	IQUV	UN	512	1	1000
I-UN-1024-C1-1000	vlac125K	1	I	UN	1024	1	1000
IQUV-UN-1024-C1-1000	vlac125K	1	IQUV	UN	1024	1	1000
I-UN-2048-C1-1000	vlac125K	1	I	UN	2048	1	1000
IQUV-UN-2048-C1-1000	vlac125K	1	IQUV	UN	2048	1	1000
I-NA-512-C1-1000	vlac125K	1	I	NA	512	1	1000
IQUV-NA-512-C1-1000	vlac125K	1	IQUV	NA	512	1	1000
I-NA-1024-C1-1000	vlac125K	1	I	NA	1024	1	1000
IQUV-NA-1024-C1-1000	vlac125K	1	IQUV	NA	1024	1	1000
I-NA-2048-C1-1000	vlac125K	1	I	NA	2048	1	1000
IQUV-NA-2048-C1-1000	vlac125K	1	IQUV	NA	2048	1	1000

**Table 1.863:** Table for continuum VLA 1M rows dataset benchmark. Common part of the benchmark code: 'CC-SF-VLAC-U1M-SP1-'
Benchmark code	Dataset	N_spw	Stokes	Wgt.	N_pixel	N_chan	N_clean
I-UN-512-C1-1000	vlac1M	1	I	UN	512	1	1000
IQUV-UN-512-C1-1000	vlac1M	1	IQUV	UN	512	1	1000
I-UN-1024-C1-1000	vlac1M	1	I	UN	1024	1	1000
IQUV-UN-1024-C1-1000	vlac1M	1	IQUV	UN	1024	1	1000
I-UN-2048-C1-1000	vlac1M	1	I	UN	2048	1	1000
IQUV-UN-2048-C1-1000	vlac1M	1	IQUV	UN	2048	1	1000
I-NA-512-C1-1000	vlac1M	1	I	NA	512	1	1000
IQUV-NA-512-C1-1000	vlac1M	1	IQUV	NA	512	1	1000
I-NA-1024-C1-1000	vlac1M	1	I	NA	1024	1	1000
IQUV-NA-1024-C1-1000	vlac1M	1	IQUV	NA	1024	1	1000
I-NA-2048-C1-1000	vlac1M	1	I	NA	2048	1	1000
IQUV-NA-2048-C1-1000	vlac1M	1	IQUV	NA	2048	1	1000

**Table 1.864:** Table for spectral line VLA 125K rows, 64 channel dataset benchmark. Common part of the benchmark code: 'CC-SF-VLAL-U125K-SP1-'
Benchmark code	Dataset	N_spw	Stokes	Wgt.	N_pixel	N_chan	N_clean
I-UN-512-C64-1000	vlac125K	1	I	UN	512	64	1000
I-UN-1024-C64-1000	vlac125K	1	I	UN	1024	64	1000
I-UN-2048-C64-1000	vlac125K	1	I	UN	2048	64	1000
I-NA-512-C64-1000	vlac125K	1	I	NA	512	64	1000
I-NA-1024-C64-1000	vlac125K	1	I	NA	1024	64	1000
I-NA-2048-C64-1000	vlac125K	1	I	NA	2048	64	1000

**Table 1.865:** Table for continuum BIMA 139M rows dataset benchmark. Common part of the benchmark code: 'CC-SF-BIMAL-U139M-SP1-'
Benchmark code	Dataset	N_spw	Stokes	Wgt.	N_pixel	N_chan	N_clean
I-NA-256-C20-1000	BIMADATA	1	I	NA	256	20	1000

The corresponding AIPS script for these benchmarks is as follows. The environment variable DATA must be set to the point to the directory where the benchmark related data files are kept (/aips++/data/demo/benchmark by default). Note that since AIPS++ datafile names have lower case letters, you will have either make a symbolic link (speaking for UNIX only) or copy the relevant data files to a files with all capital letters. This is required since AIPS internally converts all letters to upper case.

<FITSFILE> in the following script is the name of the UV FITS file to be loaded (E.g. VLAC125K.FITS).

<N> is the AIPS catalog number corresponding to the UVDATA file loaded using FITLD.

<NPIX> is the size of the image in pixels.

<STOKES> is 'I', 'Q', 'U', or 'V'. IMAGR task in AIPS accepts a single stokes value per execution.

<WEIGHT> is 'U' for uniform weighting or 'N' for natural weighting.

<NITER> is the number of Clean iterations required.

 TASK 'FITLD'
 INFILE 'DATA:<FITSFILE>'

 GO FITLD

 TASK 'IMAGR'
 GETN         <N>
 INSEQ         1             
 INDISK        1             
 SOURCES      ' '         
 QUAL         -1             
 CALCODE      ' '           
 TIMERANG      0
 SELBAND      -1             
 SELFREQ      -1             
 FREQID        1             
 SUBARRAY      0             
 DOCALIB      -1             
 GAINUSE       0             
 DOPOL        -1             
 BLVER        -1             
 FLAGVER       0             
 DOBAND       -1             
 BPVER        -1             
 SMOOTH        0           
 STOKES     <STOKES>
 BCHAN         1             
 ECHAN         1             
 CHANNEL       0             
 NPOINTS       1             
 CHINC         1             
 BIF           1             
 EIF           1             
 OUTNAME      ' '   
 OUTDISK       1             
 OUTSEQ        0             
 OUTVER        0             
 IN2NAME      ' '   
 IN2CLASS     ' '         
 IN2SEQ        0             
 IN2DISK       0             
 CELLSIZE      5,5 
 IMSIZE     <NPIX>, <NPIX>
 NFIELD        1             
 DO3DIMAG     -1             
 FLDSIZE       0           
 RASHIFT       0           
 DECSHIFT      0           
 UVTAPER       0 
 UVRANGE       0 
 GUARD         0 
 ROTATE        0             
 ZEROSP        0           
 UVWTFN     <WEIGHT>
 UVSIZE        0
 ROBUST        0             
 UVBOX         0             
 UVBXFN        1             
 XTYPE         5             
 YTYPE         5             
 XPARM         0           
 YPARM         0           
 NITER      <NITER>
 BCOMP         0           
 ALLOKAY       0             
 NBOXES        0             
 CLBOX         0           
 BOXFILE      ' '         
 OBOXFILE     ' '         
 GAIN          0.1           
 FLUX          0             
 MINPATCH     51             
 BMAJ          0             
 BMIN          0             
 BPA           0             
 OVERLAP       0             
 PHAT          0             
 FACTOR        0             
 CMETHOD      ' '         
 IMAGRPRM      0           
 NGAUSS        0             
 WGAUSS        0           
 FGAUSS        0           
 MAXPIXEL      0             
 DOTV         -1             

 GO IMAGR

The equivalent script for spectral line benchmarks would set the keywords BCHAN=C0, ECHAN=C1 where C0 and C1 are the start and end frequency channel numbers. This will generate a spectral cube. For multi-frequency synthesis, NPOINTS must be set to a value >=C1-C0+1 (this will grid all the specified channels on a single uv-grid).

VLA G and D term calibration

This benchmark has the following code descriptor:

CALVLA COMPRESSION-JONES-NAnt-SNR-NSolInt

where COMPRESSION is the compression state of the input data (compressed or uncompressed), JONES is the name of the Jones matrix being solved for ('G', `D', or 'B'), NAnt is the total number of antennas in the data, SNR the signal-to-noise ratio of the input data, and NSolInt is the number of solution intervals involved.

The individual calibrator benchmarks implemented in the default run of the calvla() method at present are enumerated in the Table 1.866.

**Table 1.866:** Table for VLA calibrator benchmark (for 'G' and 'D' Jones).
Benchmark code	Data	Compressed?	Jones	NAnt	SNR	NSolInt
CALVLAU-G-27-10-100	calvlac27s10.fits	Nope	G	27	10	100
CALVLAU-D-27-10-100	calvlac27s10.fits	Nope	D	27	10	100

<FITSFILE> in the following script is the name of the UV FITS file to be loaded (E.g. CALVLAU27S10.FITS).

<N> is the AIPS catalog number corresponding to the UVDATA file loaded using FITLD.

Following is the script for solve for the antenna based complex gain terms (GJones) as a function of time.

 TASK 'FITLD'
 INFILE 'DATA:<FITSFILE>'

 GO FITLD

This must be followed by a few, simple operations to get the UV database inside AIPS in proper order. The AIPS++ generated FITS files, though are in Time-baseline order naturally, don't seem to indicate so via whatever keyword AIPS looks for. Hence, one needs to run the AIPS task UVSRT to write out a Time-baseline (TB) sorted UV database as follows:

 TASK 'UVSRT'
 GETN        <N>
 SORT        'TB'
 GO

Next, since the simulated database is a single source database, it does not have the equivalent of the AIPS default CL table. This is required by PCAL later. To generate the blank CL table, one needs to convert the database into a multi-source database (even though it has only one source in it). This is done by running the AIPS task MULTI.

 TASK 'MULTI'

 GETN        <N>
 GO

Next, run the AIPS task INDXR as follows:

 TASK 'INDXR'

 GETN        <N>
 INFILE      ' '
 PRTLEV       0
 CPARM        0
 BPARM        0
 IN2FILE    ' '
 GO

Finally, run the task SETJY to put the source full stokes flux density in the AIPS SU tables.

 TASK 'INDXR'

 GETN        <N>
 SOURCES     ' '
 QUAL         -1
 BIF           0
 EIF           0
 ZEROSP       14.76,0.66,1.48,0
 OPTYPE      ' '
 CALCODE     ' '
 SYSVEL        0
 RESTFREQ      0 0
 VELTYP      ' '
 VELDEF      ' '
 FREQID       -1
 APARM         0
 GO

The data is now ready to run the tasks CALIB (for the equivalent of GJones calibration) and PCAL (for the equivalent of DJones calibration).

 TASK 'CALIB'
 GETN         <N>
 CALSOUR      ' '
 QUAL         -1
 CALCODE      '    '
 SELBAND       0  
 SELFREQ       0  
 FREQID        0 
 TIMERANG      0  
 BCHAN         64
 ECHAN         64
 ANTENNAS      0
 DOFIT         0
 ANTUSE        0
 SUBARRAY      0  
 UVRANGE       0  0
 WTUV          0
 DOCALIB      -1
 GAINUSE       0 
 FLAGVER       0  
 DOBAND       -1 
 BPVER        -1 
 SMOOTH        0
 IN2NAME      '     '
 IN2CLASS     '     '
 IN2SEQ        0 
 IN2DISK       0
 INVERS        0 
 NCOMP         0
 FLUX          0 
 NMAPS         0
 CMETHOD      '    '
 CMODEL       '    ' 
 SMODEL       0
 OUTNAME      '    '
 OUTCLASS     '    '
 OUTSEQ        0   
 OUTDISK       1
 REFANT        0
 SOLINT        1.16 #to get 100 solution intervals for the checked-in benchmark dataset
 APARM         0
 SOLTYPE      '    ' 
 SOLMODE      'A&P'
 SOLCON        0
 MINAMPER      0
 MINPHSER      0
 CPARM         0
 SNVER         0
 ANTWT         0
 GAINERR       0
 BADDISK       0

 GO CALIB

Following is the script for solve for the antenna based polarization leakage term.

 TASK 'PCAL'
 GETN         <N>
 CALSOUR      ' '
 TIMERANG      0
 SELBAND      -1
 SELFREQ      -1
 FREQID       -1
 BIF           0
 EIF           0
 ANTENNAS      0
 UVRANGE       0  0
 SUBARRAY      0
 FLAGVER       0
 DOCALIB       1
 GAINUSE       0
 IN2NAME      ' '
 IN2CLASS     ' '
 IN2SEQ        0
 IN2DISK       0
 INVERS        0
 NCOMP         0
 FLUX          0
 NMAPS         0
 PMODEL        0
 SOLINT        0
 SOLTYPE      'APPR'
 PRTLEV        0
 REFANT        0
 BPARM         0
 CPARM         0
 BADDISK       0

 GO PCAL

Importing visibility data from FITS files

This benchmark has the following code descriptor:

FUV-RD-data

where data is the input data description, further broken down as data= [instrument][size][compression][obsmode][nchan]. The parameters are specified as follows: i) instrument - telescope abbreviation ii); size - dataset size on an integer scale of 1 to 10 (4 $\sim$ 630k UVFITS visibilities); iii) compression - MS compression (U=uncompressed; C=compressed); iv) obsmode - observing mode (C=continuum, L=line); v) nchan - number of frequency channels.

The individual FITS Filler benchmarks implemented in the default run of the fuvrd () method at present are enumerated in Table 1.867.

**Table 1.867:** Table for FITS UV read benchmark (import visibility data into AIPS++ from FITS files). Common part of the code is ' `FUV-RD-VLA`'.
Benchmark code	Dataset	Mode	N_chans	Compressed?	Data size
C-U125K-C1	vlac125k.fits	Continuum	1	False	125K
C-U1M-C1	vlac1m.fits	Continuum	1	False	1M
L-U125K-C64	vlal125K64Chan.fits	Line	64	False	125K

Tools

atmosphere	Tool for calculating atmospheric effects in the mm regime
benchmark	tool for standard AIPS++performance benchmarks

Next: benchmark - Tool Up: benchmark - Module Previous: Overview of benchmark tool functions: Contents Index