Home

What is GPUVMEM?

Radio interferometers are an array of antennas capable of sampling the sky collecting radio signals from specific sources. Each antenna's signal is correlated with every other signal to collect a large number of samples of the sky but on the uv-plane in order to cover the Fourier plane as much as possible. As it can be seen these noisy, irregular and sparse samples are a set of complex numbers in the uv plane called visibilities. Additionally, the process of solving the problem of reconstruct an image from these data is called Image Synthesis.

Supposing that the uv-plane is completely sampled a simple relationship between image and data (van Cittert-Zernike theorem) can be expressed mathematically:

where A(x,y) is called primary beam and is the solid angle reception pattern of the individual antennas, V(u,v) are the visibilities in a certain (u,v) position and I(x,y) is the image in the pixel (x,y). In this case, a Fourier inversion can be applied to V(u,v) to recover I(x,y).

Since in a real scenario, radio-interferometers collect noisy and irregular samples of data, the problem is then not well defined. A naive approach to solve this problem would be to sample the Fourier domain at discrete points on a regular grid (see next image on the right) using the sampling function

a weighting function W(u,v) and a convolution kernel C(u,v).

If a an inverse Fourier transform of the gridded data is done then we would obtain a poor quality image due to the incomplete spatial sampling of the interferometric array. This image is called dirty image and an example of the protoplanetary disk HL-Tauri is shown below.

It is clear that this image is not useful to study the rings of the disk and to make conclusions about the planet formation in it.

GPUVMEM is a framework that allows radio astronomers to do image synthesis using a Bayesian approach to solve the inverse problem. This process takes Fourier space data taken by a radio-interferometer like ALMA, ATCA, VLA, or a set of radio-interferometers as EHT, and returns an image from a certain a radio source.

---

What are GPUVMEM's current features?

1. Object oriented Framework: Object orientation makes easier the creation and building of a feature that you are interested in and plug that to the program.
2. Multi-GPU: Radio datasets and memory are distributed between different GPU using the P2P approach.
3. Multi-Field/Mosaic support: Mosaic images in interferometry allow the study of large scale objects in the sky. In this Bayesian approach we fit a single model image to the ensemble of all pointings.
4. Parameterized reconstruction: Although GPUVMEM requires a few parameters, it does not require user assistance during algorithm iteration. In this sense, we say GPUVMEM is an unsupervised image synthesis algorithm.
5. Regular grid-to-irregular grid approach: An interpolation step is done to compute model visibilities from the 2D Fast Fourier transform of the image estimate to compare them directly with observed visibilities.
6. Gridding to speed up the reconstructions: Gridding is the process of resampling visibility data into a regular grid. Usually this is done convolving the data with a kernel to avoid aliasing. Although convolutional kernels are not yet implemented in GPUVMEM, the gridding process reduce the amount of data and speeds up the algorithm reducing computation time.
7. Multi-frequency support: Spectral dependency can introduce strong effects into image synthesis. In this implementation we use the following image representation for multi-frequency synthesis:

where Inu0 is the image at reference frequency and alpha is the spectral index image.

How do I install it?

Requirements

Installation