This software computes the tomographic reconstruction of spatial-spectral data from raw detector images of the Computed-Tomography Imaging Spectrometer (CTIS), which enables transient-level, multi-spectral imaging by capturing spatial and spectral information in a single snapshot. The CTIS can be used for surveying planetary landscapes through spectral imaging. It can also be used for battlefield surveillance and the spectral imaging of live tissues for disease detection.
A Message Passing Interface Library (MPI) is used to parallelize the original serial version of the code without modifying its initial structure. By parallelizing the code, a speedup of up to 20 is reached by using 32 processors. The software does not use any third-party libraries that require licenses. It is written in Fortran and MPI, and the storage of matrix elements is efficient, thus reducing memory requirements.
This work was done by Seungwon Lee of Caltech for NASA's Jet Propulsion Laboratory.
This software is available for commercial licensing. Please contact Karina Edmonds of the California Institute of Technology at (626) 395-2322. Refer to NPO-45831.
This Brief includes a Technical Support Package (TSP).
Parallel Computing for the Computed- Tomography Imaging Spectrometer
(reference NPO-45831) is currently available for download from the TSP library.
Don't have an account?
Overview
The document titled "Parallel Computing for the Computed-Tomography Imaging Spectrometer" from NASA's Jet Propulsion Laboratory (JPL) outlines advancements in parallel computing techniques aimed at enhancing the performance of the Computed-Tomography Imaging Spectrometer (CTIS). CTIS is a cutting-edge instrument designed for transient-event multi-spectral imaging, capturing both spatial and spectral information in a single snapshot.
The primary challenge addressed in the document is the computational intensity of tomographic reconstruction, which is essential for generating a spectrum for every pixel in the captured scene. This process involves complex matrix-vector multiplications between scene voxels (or detector image pixels) and a system matrix, which collectively yield the detector's response. To improve the efficiency of these computations, the document discusses the implementation of parallel computing strategies.
The approach taken involves parallelizing the matrix-vector multiplication process, which is crucial for speeding up the tomographic reconstruction. The tests were conducted using the Cosmos cluster, which consists of 1024 processors, and the Intel 7.1 Fortran compiler was utilized for coding. Notably, the original code encountered issues when compiled with a newer version of the Intel compiler, leading to segmentation faults. To facilitate communication between processors, the Message Passing Interface (MPI) was employed, allowing for effective interprocessor communication during the parallel computations.
The document also provides specific details about the sizes of the matrices and vectors involved in the computations, indicating a system matrix size of 4,194,304 x 2,889,900, a detector vector size of 2048 x 2048, and a scene vector size of 195 x 195 x 76. The performance improvement is quantified through speedup metrics, which compare the computational time of the parallelized approach to that of a single processor.
Overall, this technical support package not only highlights the innovative use of parallel computing in enhancing the capabilities of CTIS but also emphasizes the broader implications of these advancements for aerospace-related developments. The document serves as a resource for further inquiries and underscores the importance of compliance with U.S. export regulations regarding proprietary information.
