TOMOCUPY

RESOURCE

Abstract

ANL REFERENCE SF-22-102 DESCRIPTION: Tomocupy is a Python package and a command-line interface for GPU reconstruction of tomographic/laminographic data in 16-bit and 32-bit precision. It implements an efficient data processing conveyor allowing to overlap all data transfers with computations. First, independent Python threads are started for reading data chunks from the hard disk into a Python data queue and for writing reconstructed chunks from the Python queue to the hard disk. Second, CPU-GPU data transfers are overlapped with GPU computations by using CUDA streams.
Developers:
NIKITIN, VIKTOR [1] DE CARLO, FRANCESCO [1]
  1. Argonne National Lab. (ANL), Argonne, IL (United States)
Release Date:
2022-08-19
Project Type:
Open Source, Publicly Available Repository
Software Type:
Scientific
Licenses:
BSD 3-clause "New" or "Revised" License
Sponsoring Org.:
Code ID:
92779
Research Org.:
Argonne National Laboratory (ANL), Argonne, IL (United States)
Country of Origin:
United States

RESOURCE

Citation Formats

NIKITIN, VIKTOR, and DE CARLO, FRANCESCO. TOMOCUPY. Computer Software. https://github.com/tomography/tomocupy. USDOE OFFICE OF BASIC ENERGY SCIENCE. 19 Aug. 2022. Web. doi:10.11578/dc.20220919.1.
NIKITIN, VIKTOR, & DE CARLO, FRANCESCO. (2022, August 19). TOMOCUPY. [Computer software]. https://github.com/tomography/tomocupy. https://doi.org/10.11578/dc.20220919.1.
NIKITIN, VIKTOR, and DE CARLO, FRANCESCO. "TOMOCUPY." Computer software. August 19, 2022. https://github.com/tomography/tomocupy. https://doi.org/10.11578/dc.20220919.1.
@misc{ doecode_92779,
title = {TOMOCUPY},
author = {NIKITIN, VIKTOR and DE CARLO, FRANCESCO},
abstractNote = {ANL REFERENCE SF-22-102 DESCRIPTION: Tomocupy is a Python package and a command-line interface for GPU reconstruction of tomographic/laminographic data in 16-bit and 32-bit precision. It implements an efficient data processing conveyor allowing to overlap all data transfers with computations. First, independent Python threads are started for reading data chunks from the hard disk into a Python data queue and for writing reconstructed chunks from the Python queue to the hard disk. Second, CPU-GPU data transfers are overlapped with GPU computations by using CUDA streams.},
doi = {10.11578/dc.20220919.1},
url = {https://doi.org/10.11578/dc.20220919.1},
howpublished = {[Computer Software] \url{https://doi.org/10.11578/dc.20220919.1}},
year = {2022},
month = {aug}
}