Fast tomographic reconstruction with CPU+GPU co-processing
J.I. Agulleiro (1), F. Vázquez (1), E.M. Garzón (1), J.J. Fernández (2)
(1) Associated unit CSIC+UAL. Univ. Almeria. 04120 Almeria. Spain.
(2) Centro Nacional de Biotecnologia - CSIC. 28049 Madrid. Spain.
Contact: jjfernandez.software @ gmail.com
Tomo3Dhybrid implements a hybrid computing approach to tomographic reconstruction. It takes full advantage of the power in modern computers equipped with multicore processors (CPUs) and graphics processing units (GPUs). This approach collaboratively combines the CPUs and the GPUs in the system by using an 'on-demand' strategy whereby these computing devices asynchronously request a new piece of work to do when idle. The tasks allotted to the CPUs/GPUs consists of reconstructing slabs of slices. Tomo3Dhybrid has the most common reconstruction methods implemented, namely WBP and SIRT. Full resolution tomograms are thus generated at high speed with this hybrid approach (few seconds for WBP; few minutes for SIRT). Tomo3Dhybrid reads input tilt-series and writes output tomograms in MRC format. The program is compatible with standard packages, which allows easy integration in the electron tomography workflow.
A detailed description of the procedure implemented in the package can be found in the following articles:
Hybrid computing: CPU+GPU co-processing and its application to tomographic reconstruction
Please, cite this article if you use Tomo3Dhybrid in your works.
Developers might also be interested in Oddylos, our on-demand dynamic load scheduler implemented within Tomo3D and Tomo3Dhybrid.
Current version: February 2012