SCIENTIFIC COMPUTING

BACK

CONTACT

Leader: Eduardo Xavier Silva Miqueles

Number: +55 19 3510 1043

E-mail: eduardo.miqueles@lnls.br

The Scientific Computing Group (GCC) is a support group in high performance processing for the Brazilian Synchrotron Light Laboratory (LNLS). Currently the focus of the group is on the manipulation, treatment, restoration and reconstruction of images generated in the beamlines of LNLS’ current synchrotron light source, UVX.

Among the objectives of the group are (i) the development of innovative algorithms that can compete with or surpass existing ones; (ii) the implementation of these algorithms in high performance platforms, essentially based on GPU; and (iii) the integration of these algorithms with real data from the beamlines.

STAFF

Name Position
Alan Zanoni PeixinhoSpecialist
Eduardo Xavier Silva MiquelesResearcher, Leader
Fernando Seiti FurusatoSpecialist
Gilberto Martinez JuniorScientific Associate
Giovanni Lenzi BaraldiScientific Associate
Matheus Fernandes SarmentoScientific Associate
Patricio Guerrero PradoResearcher
Thiago Vallin SpinaResearcher

IMAGE PROCESSING

Image processing refers specifically to:

1. Reconstruction of images from projections;
2. Segmentation of images from reconstructions;
3. Extraction of information from the segmentations and / or reconstructions.

Classes (1) and (2) contain a finite set of specific sub-operations in image processing, which contribute to the better performance in the analysis of data generated in LNLS. Usually, (2) is a consequence of (1), but not exclusively; being (3) one of the main objectives of an experimental imaging process. It is understood that the extraction of information from data generated in the beamlines is crucial to the success of the facilities.

The cycle of fundamental operations of a beamline, integrated to the scientific computation and within the network topology specified to the Laboratory is presented below:

Currently, the operations follow $\rm A \rightarrow C$ with the subsequent processing being performed by $\rm C \rightarrow D$, where (A) is a fixed, but arbitrary, beamline. For the new synchrotron light source, Sirius, the focus on high performance is idealized in operations $\rm A \rightarrow B \rightarrow C$, with high local processing performance at (B), within the beamline. Further processing will be done within the $\rm C \leftrightarrow D$ pipeline. Network topology connections are designed at 100 Gb/s to accelerate access and flow of measured data.