CSR (Compressed Sparse Row)

E839061

CSR (Compressed Sparse Row) is a memory-efficient sparse matrix storage format that stores only nonzero elements and their indices in row-major order to enable fast arithmetic and matrix–vector operations.

Jump to: Statements Referenced by

Statements (46)

Predicate Object
instanceOf data structure
sparse matrix storage format
advantageOverDense avoids computation on zero entries
reduces memory for highly sparse matrices
alsoKnownAs CRS
CSR format
Compressed Sparse Row NERFINISHED
compressed row storage
complexity O(nnz) storage complexity
O(nnz) time for matrix–vector multiplication
component column indices array
row pointer array
values array
contrastedWith COO (Coordinate format)
CSC (Compressed Sparse Column)
DIA (Diagonal format) NERFINISHED
domain numerical linear algebra
sparse matrix computation
limitation inefficient for column slicing
insertion of new nonzeros is expensive
property enables fast arithmetic operations
enables fast matrix–vector operations
good cache locality for row-wise access
memory-efficient for sparse matrices
stores only nonzero entries
supports efficient row slicing
rowPointerArrayMeaning prefix sum of nonzeros per row
storageOrder row-major
stores column indices of nonzero elements
nonzero matrix elements
row pointer array
supportedBy Eigen NERFINISHED
MKL NERFINISHED
PETSc NERFINISHED
SciPy NERFINISHED
cuSPARSE NERFINISHED
typicalIndexBase 0-based indices GENERATED
typicalUseCase finite element methods
sparse linear algebra libraries
usedFor graph algorithms
high‑performance computing
iterative linear solvers
scientific computing
sparse matrix–matrix multiplication
sparse matrix–vector multiplication
storing sparse matrices

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

cuSPARSE supportsMatrixFormat CSR (Compressed Sparse Row)