Panua-Pardiso
Panua-Pardiso 8.2 was released in February 2024. It contains full support of multi-threaded Schur-complement computations, parallel selected inversion, incremental updates, and many others accelerations. Some of the new features have been used, e.g., in the following papers:
- D. Pasadakis, M. Bollhöfer, and O. Schenk, Sparse quadratic approximation for graph learning, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 9, pp. 11256-11269, 2023.
- A. Eftekhari, D. Pasadakis, M. Bollhöfer, S. Scheidegger, and O. Schenk, Block-enhanced precision matrix estimation for large-scale datasets, Journal of Computational Science, vol. 53, 2021.
- L. Gaedke-Merzhäuser, J. van Niekerk, O. Schenk, and H. Rue, Parallelized integrated nested laplace approximations for fast bayesian inference, Statistics and Computing, vol. 33, no. 1, p. 25, 2023.
Additional publications related to Pardiso are available on the webpage of the research group of Prof. Olaf Schenk, Advanced Computing Laboratory, Institute of Computing, USI Lugano, Switzerland.
Important: Please note that the Intel MKL version of Pardiso is based on our version from 2006 and that a lot of new features and improvements of Pardiso are not available in the Intel MKL library.
Features
Solves unsymmetric, structurally symmetric or symmetric systems, real or complex, positive definite or indefinite, hermitian.
LU decomposition with complete pivoting and extremely fast selected inversion strategies.
Unlocks the rapid and and accurate reduction of computing time by taking advantage of SMP and MPP parallel processing technology.
Industrial strength - Licensed to various leading industrial software vendors such as Dassault, NXP, Siemens, or Formula One racing teams.
Supporting Intel and ARM architectures.
Enables easy integration as a dynamic or static library.
Benchmarks of Pardiso 8.2
MKL Pardiso vs Pardiso 8.2 performance comparison, NXP semiconductors
Pardiso 8.2 is benchmarked against Intel MKL Pardiso version 2020 on up to 16 cores on an Intel Xeon E7-4880. The matrices are from NXP's industrial circuit simulator Mica. The plots shows the performance acceleration for the factorization and solution phase against MKL Pardiso.
MKL Pardiso vs Pardiso 8.2 performance comparison, Silvaco circuit simulator
Pardiso 8.2 is benchmarked against Intel MKL Pardiso version 2020 on up to 32 cores on an Intel Xeon Intel E5-2650 with 2.30GHz architecture. The matrices are from Silvaco industrial circuit simulator. The plots shows the performance acceleration for the factorization and solution phase against MKL Pardiso.
Pardiso 8.2 k-rank update analysis, NXP semiconductor
Regression analysis on the k-rank update LU factorization in Pardiso. The scatter plot shows the number of k-rank updates and the corresponding factorization time in milliseconds. The regression analysis cleary demonstates a linear trend both for the single and the multiple core version. The dashed line shows the time for the full factorization.
Examples
C
- symmetric linear systems
- unsymmetric linear systems
- complex unsymmetric linear systems
- symmetric linear systems, example based on a discretized 2D Laplacian equation (laplace.c, laplace.h)