Loading...
Loading...

Go to the content (press return)

A case for malleable thread-level linear algebra libraries: The LU factorization with partial pivoting

Author
Catalán, S.; Herrero, J.; Quintana, E.; Rodríguez, R.; Van De Geijn, R.
Type of activity
Journal article
Journal
IEEE access
Date of publication
2019-01-31
Volume
7
First page
17617
Last page
17633
DOI
10.1109/ACCESS.2019.2895541
Project funding
High performance computing VII
Models de programació i entorns d'execució paral·lels
Repository
http://hdl.handle.net/2117/129939 Open in new window
URL
https://ieeexplore.ieee.org/document/8630926 Open in new window
Abstract
We propose two novel techniques for overcoming load-imbalance encountered when implementing so-called look-ahead mechanisms in relevant dense matrix factorizations for the solution of linear systems. Both techniques target the scenario where two thread teams are created/activated during the factorization, with each team in charge of performing an independent task/branch of execution. The first technique promotes worker sharing (WS) between the two tasks, allowing the threads of the task that com...
Citation
Catalán, S. [et al.]. A case for malleable thread-level linear algebra libraries: The LU factorization with partial pivoting. "IEEE access", 31 Gener 2019, vol. 7, p. 17617-17633.
Keywords
BLAS, Basic linear algebra subprograms, LAPACK, Linear algebra package, Multi-threading, Solution of linear systems, Thread malleability, Workload balancing
Group of research
CAP - High Performace Computing Group
inLab FIB

Participants

  • Catalán Pallarés, Sandra  (author)
  • Herrero Zaragoza, José Ramón  (author)
  • Quintana Ortí, Enrique Salvador  (author)
  • Rodríguez Sánchez, Rafael  (author)
  • Van De Geijn, Robert  (author)

Attachments