Presentation
(PP20) ESSEX-II: Equipping Sparse Solvers for Exascale
SessionProject Posters Presentation
Event Type
Project Poster
Extreme-Scale Computing
Heterogeneous Systems
Math Library Design
Parallel Algorithms
Parallel Applications
TimeTuesday, June 26th3:15pm - 3:45pm
LocationBooth N-230
DescriptionThe ESSEX project is funded by the German DFG priority programme 1648 “Software for Exascale Computing” (SPPEXA). In 2016 it has entered is second funding phase, ESSEX-II.
ESSEX investigates programming concepts and numerical algorithms for scalable, efficient and robust iterative sparse matrix applications on exascale systems. Starting with successful blueprints and prototype solutions identified in ESSEX-I, the second phase project ESSEX-II aims at delivering a collection of broadly usable and scalable sparse eigenvalue solvers with high hardware efficiency for the computer architectures to come. ESSEX-II is now entering its third year, and we are working towards providing efficient and easily usable software components:
- GHOST, the General Hybrid Optimized Sparse Toolkit, a collection of highly parallel sparse linear algebra building blocks. Heterogeneous parallelism among CPUs, GPUs, and Xeon Phi ist fully supported. (Released)
- PHIST, the Pipelined Hybrid-Parallel Iterative Solver Toolkit. It provides general-purpose block Jacobi-Davidson Eigensolvers, Krylov methods, and a preconditioning interface. C, C++, Fortran 2003 and Python bindings are available, as well as mutiple backends (including GHOST). (Released)
- CRAFT, the Checkpoint/Restart and Automatic Fault Tolerance Library. It allows to efficiently add C/R and AFT functionality to C++ programs and supports shrinking and non-shrinking recovery as well as fully asynchronous checkpointing. (Released)
- RACE, the Recursive Adaptive Coloring Engine, provides efficient block multicoloring with performance superior to existing libraries. (Release in 2018 planned)
- ScaMaC, the Scalable Matrix Collection, allows the parallel generation of large sparse matrices from diverse applications fields in Quantum Physics. (Release in 2018 planned)
ESSEX investigates programming concepts and numerical algorithms for scalable, efficient and robust iterative sparse matrix applications on exascale systems. Starting with successful blueprints and prototype solutions identified in ESSEX-I, the second phase project ESSEX-II aims at delivering a collection of broadly usable and scalable sparse eigenvalue solvers with high hardware efficiency for the computer architectures to come. ESSEX-II is now entering its third year, and we are working towards providing efficient and easily usable software components:
- GHOST, the General Hybrid Optimized Sparse Toolkit, a collection of highly parallel sparse linear algebra building blocks. Heterogeneous parallelism among CPUs, GPUs, and Xeon Phi ist fully supported. (Released)
- PHIST, the Pipelined Hybrid-Parallel Iterative Solver Toolkit. It provides general-purpose block Jacobi-Davidson Eigensolvers, Krylov methods, and a preconditioning interface. C, C++, Fortran 2003 and Python bindings are available, as well as mutiple backends (including GHOST). (Released)
- CRAFT, the Checkpoint/Restart and Automatic Fault Tolerance Library. It allows to efficiently add C/R and AFT functionality to C++ programs and supports shrinking and non-shrinking recovery as well as fully asynchronous checkpointing. (Released)
- RACE, the Recursive Adaptive Coloring Engine, provides efficient block multicoloring with performance superior to existing libraries. (Release in 2018 planned)
- ScaMaC, the Scalable Matrix Collection, allows the parallel generation of large sparse matrices from diverse applications fields in Quantum Physics. (Release in 2018 planned)
Poster PDF