the HcSPARSE project has been deprecated. Please use rocSPARSE instead.
This repository hosts the HCC implementation of linear algebraic routines for sparse matrices/vectors (HcSparse), on AMD devices. To know what HCC compiler features, refer here.
Following list the routines that are currently supported by HcSparse that are grouped into four major categories.
- Level 1 BLAS Routines:
- axpy : Product of Sparse Vector with constant and followed by addition of dense Vector.
- dot : Dot product of Sparse Vector with dense Vector
- Level 2 BLAS Routines:
- csrmv : Sparse Matrix - dense Vector multiply (SpM-dV)
- csrmm : Sparse Matrix - dense Matrix multiply (SpM-dM)
- Level 3 BLAS Routines:
- csrgemm : Sparse Matrix - Sparse Matrix multiply
- csrgeam : Sparse Matrix - Sparse Matrix addition
- Conversion Routines:
- dense2csr : Dense to CSR conversions
- csr2dense : CSR to Dense conversions
- coo2csr : COO to CSR conversions
- dense2csc : Dense to CSC conversions
- csc2dense : CSC to Dense conversions
- Input Reader functions:
- mm-reader : Functions to read matrix market files in COO or CSR format
- Supports COO, CSR, CSC, Dense matrix formats.
- Level 1 Routines for sparse vector with dense formats
- Level 2 routines for sparse matrix with dense vector/matrix formats
- Level 2 routines for sparse matrix with sparse matrix operations.
- Conversion routines that supports conversion between different matrix formats.
- conjugate-gradients : iterative conjugate gradient solver (CG)
- biconjugate-gradients-stabilized : Iterative biconjugate gradient stabilized solver (BiCGStab)
- Refer Prerequisites section here
- Refer Tested environments enumerated here
- Follow installation steps as described here
- Follow testing procedures as explained here