NVIDIA cuDSS (Preview): A high-performance CUDA Library for Direct Sparse Solvers¶
NVIDIA cuDSS (Preview) is a library of GPU-accelerated linear solvers with sparse matrices. It provides algorithms for solving linear systems of the following type:
A X = B
with a sparse matrix A, right-hand side B and unknown solution X (could be a matrix or a vector).
The cuDSS functionality allows flexibility in matrix properties and solver configuration, as well as execution parameters like CUDA streams.
Note: Since the library is released as a preview, API is subject to change in later releases.
Provide Feedback: cuDSS-EXTERNAL-Group@nvidia.com
Real/complex general/symmetric/positive-definite sparse matrices
Single and double precision datatype for values and
intdatatype for indices
Single and multiple right-hand sides
Multi-stage execution with three main phases: reordering & symbolic factorization, numerical factorization and solving
Different algorthms for reordering and factorization phases
Supported configurations: single GPU
Supported SM Architectures:
SM 7.0and newer
Supported CPU Architectures:
- Release Notes
- Getting Started
- cuDSS Data Types
- cuDSS Functions
- cuDSS Logging Features
- Software License Agreement
- Third Party License Agreements