Building the project in this branch is as follows:

• Laptop/workstation using NVIDIA :

  • CPU : cd build cmake .. make make test sudo make install

  • GPU : cd build_gpu cmake -DUSE_HIP=ON -DCUDA_SM=86 .. # make use to change CUDA_SM to value relevant to your specific GPU (check on https://en.wikipedia.org/wiki/CUDA ) make make test sudo make install

• Setonix - at least temporarily there are separate build_setonix.sh and CMakeLists.txt_SETONIX scripts for Setonix/HIP architecture

  • CPU version ./build.sh cpu - "-DENABLE_PROFILER=ON"

  • GPU version salloc -p gpu --gres=gpu:1 --time 0:30:00 -n1 --account=??? ./build.sh gpu - "-DENABLE_PROFILER=ON -DUSE_HIP=ON"

pacer_dirty_imager

Pacer BLINK imaging software (including test and demonstration programs). Fast imaging program developed in C++ for the BLINK PaCER project.

required

sudo apt-get install libnova-dev fftw2 fftw-dev fftw3-dev libhdf5-dev libcfitsio-dev

msfitslib follow the installation instructions at https://github.com/marcinsokolowski/msfitslib

astroio follow the installation instructions at https://github.com/PaCER-BLINK-Project/astroio

libpal follow the installation instructions at https://github.com/Starlink/pal

installation:

On super computer :

git clone [email protected]:blink/imager.git

./build.sh

On any desktop computer:

mkdir build

cd build

cmake .. ( for debug version : cmake -DCMAKE_BUILD_TYPE=Debug .. )

make

sudo make install

current programs :

• The package includes script to dump correlation matrix from CASA or UVFITS files

  • Starting with demonstration software using CASA dumps of visibility and UVW data
  • Scripts for dumping CASA ms are in cotter_wsclean/scripts/casa/

• pacer_dirty_image - reads visibilities and UVW data from FITS files as produced by script cotter_wsclean/scripts/casa/get_corrmatrix_from_casa.py (may be moved here now) then gridds the data on UV grid, performs invFFT and gets dirty image. Has already been demonstrated to work well with EDA2 data (SKA-Low station data) and currently work is underway to make it working with MWA data.

  Example usage :

    2049x2049 image : pacer_dirty_image_test 20191104_033537_eda2_ch1_ant256_midday_avg1 -p _channel000_time000000 -n 2049
  

  generate visibilties and do inverse FFT :

         pacer_uvgrid_generator -A 100.00 -n 129
         pacer_dirty_image in out -g 0 -r uv_grid_re.fits -i uv_grid_im.fits
  

  or 8193x8193 : pacer_uvgrid_generator -A 100.00 -n 8193 pacer_dirty_image in out -g 0 -r uv_grid_re.fits -i uv_grid_im.fits -n 8193

Test data:

EDA2 test dataset for Topaz (taken from as in scp.sh script):

 cd data/topaz/test1_eda2

 ./doit_topaz!

 The program is expected to create :

    Image of the sky named : dirty_image_20220406T035747506.fits

    Imaginary image named : 20220406T035747506_imag.fits

 WARNING : the name conventions are being changed. So, they are likely to change soon, especially the name of imaginary part is just WRONG now !

 The reference images are included in the test dataset, hence 

    dirty_image_20220406T035747506.fits should be equal to dirty_image_20220406T035747506_real_TEMPLATE_512x512.fits

    20220406T035747506_imag.fits should be equal to dirty_image_20220406T035747506_imag_TEMPLATE_512x512.fits

For any desktop computer just execute ./doit! script which should give the same results as above.