rocSPARSE 4.1.0 for ROCm 7.1.0

Added

• Added brain half float mixed precision to rocsparse_axpby where X and Y use bfloat16 and result and the compute type use float.
• Added brain half float mixed precision to rocsparse_spvv where X and Y use bfloat16 and result and the compute type use float.
• Added brain half float mixed precision to rocsparse_spmv where A and X use bfloat16 and Y and the compute type use float.
• Added brain half float mixed precision to rocsparse_spmm where A and B use bfloat16 and C and the compute type use float.
• Added brain half float mixed precision to rocsparse_sddmm where A and B use bfloat16 and C and the compute type use float.
• Added brain half float mixed precision to rocsparse_sddmm where A and B and C use bfloat16 and the compute type use float.
• Added half float mixed precision to rocsparse_sddmm where A and B and C use float16 and the compute type use float.
• Added brain half float uniform precision to rocsparse_scatter and rocsparse_gather routines.

Optimized

• Improved the user documentation.

Upcoming changes

• Deprecate trace, debug, and bench logging using environment variable ROCSPARSE_LAYER.

rocSPARSE 4.0.3 for ROCm 7.0.2

Resolved issues

• Resolved an issue causing premature deallocation of internal buffers still in use.

rocsparse 4.0.2 for ROCm 7.0.1

rocSPARSE code for ROCm 7.0.1 did not change. The library was rebuilt for the updated ROCm 7.0.1 stack.

rocSPARSE 4.0.2 for ROCm 7.0.0

Added

• Adds SpGEAM generic routine for computing sparse matrix addition in CSR format
• Adds v2_SpMV generic routine for computing sparse matrix vector multiplication. As opposed to the deprecated rocsparse_spmv routine, this routine does not use a fallback algorithm if a non-implemented configuration is encountered and will return an error in such a case. For the deprecated routine rocsparse_spmv, the user can enable warning messages in situations where a fallback algorithm is used by either calling upfront the routine rocsparse_enable_debug or exporting the variable ROCSPARSE_DEBUG (with the shell command export ROCSPARSE_DEBUG=1).
• Adds half float mixed precision to rocsparse_axpby where X and Y use float16 and result and the compute type use float
• Adds half float mixed precision to rocsparse_spvv where X and Y use float16 and result and the compute type use float
• Adds half float mixed precision to rocsparse_spmv where A and X use float16 and Y and the compute type use float
• Adds half float mixed precision to rocsparse_spmm where A and B use float16 and C and the compute type use float
• Adds half float mixed precision to rocsparse_sddmm where A and B use float16 and C and the compute type use float
• Adds half float uniform precision to rocsparse_scatter and rocsparse_gather routines
• Adds half float uniform precision to rocsparse_sddmm routine
• Added rocsparse_spmv_alg_csr_rowsplit algorithm.
• Added support for gfx950
• Add ROC-TX instrumentation support in rocSPARSE (not available on Windows or in the static library version on Linux).
• Added the almalinux OS name to correct the gfortran dependency

Changed

• Switch to defaulting to C++17 when building rocSPARSE from source. Previously rocSPARSE was using C++14 by default.

Optimized

• Reduced the number of template instantiations in the library to further reduce the shared library binary size and improve compile times
• Allow SpGEMM routines to use more shared memory when available. This can speed up performance for matrices with a large number of intermediate products.
• Use of the rocsparse_spmv_alg_csr_adaptive or rocsparse_spmv_alg_csr_default algorithms in rocsparse_spmv to perform transposed sparse matrix multiplication (C=alpha*A^T*x+beta*y) resulted in unnecessary analysis on A and needless slowdown during the analysis phase. This has been fixed by skipping the analysis when performing the transposed sparse matrix multiplication.
• Improved the user documentation

Resolved issues

• Fixed an issue in the public headers where extern "C" was not wrapped by #ifdef __cplusplus, which caused failures when building C programs with rocSPARSE.
• Fixed a memory access fault in the rocsparse_Xbsrilu0 routines.
• Fixed failures that could occur in rocsparse_Xbsrsm_solve or rocsparse_spsm with BSR format when using host pointer mode.
• Fixed ASAN compilation failures
• Fixed failure that occurred when using const descriptor rocsparse_create_const_csr_descr with the generic routine rocsparse_sparse_to_sparse. Issue was not observed when using non-const descriptor rocsparse_create_csr_descr with rocsparse_sparse_to_sparse.
• Fixed a memory leak in the rocsparse handle

Removed

• The deprecated rocsparse_spmv_ex routine
• The deprecated rocsparse_sbsrmv_ex, rocsparse_dbsrmv_ex, rocsparse_cbsrmv_ex, and rocsparse_zbsrmv_ex routines
• The deprecated rocsparse_sbsrmv_ex_analysis, rocsparse_dbsrmv_ex_analysis, rocsparse_cbsrmv_ex_analysis, and rocsparse_zbsrmv_ex_analysis routines

Upcoming changes

• Deprecated the rocsparse_spmv routine. Users should use the rocsparse_v2_spmv routine going forward.
• Deprecated rocsparse_spmv_alg_csr_stream algorithm. Users should use the rocsparse_spmv_alg_csr_rowsplit algorithm going forward.
• Deprecated the rocsparse_itilu0_alg_sync_split_fusion algorithm. Users should use one of rocsparse_itilu0_alg_async_inplace, rocsparse_itilu0_alg_async_split, or rocsparse_itilu0_alg_sync_split going forward.

rocSPARSE 3.4.0 for ROCm 6.4.4

rocSPARSE code for ROCm 6.4.4 did not change. The library was rebuilt for the updated ROCm 6.4.4 stack.

rocSPARSE 3.4.0 for ROCm 6.4.3

rocSPARSE code for ROCm 6.4.3 did not change. The library was rebuilt for the updated ROCm 6.4.3 stack.

rocSPARSE 3.4.0 for ROCm 6.4.2

rocSPARSE code for ROCm 6.4.2 did not change. The library was rebuilt for the updated ROCm 6.4.2 stack.

rocSPARSE 3.4.0 for ROCm 6.4.1

rocSPARSE code for ROCm 6.4.1 did not change. The library was rebuilt for the updated ROCm 6.4.1 stack.

rocSPARSE 3.4.0 for ROCm 6.4.0

Added

• Added support for rocsparse_matrix_type_triangular in rocsparse_spsv
• Added test filters smoke, regression, and extended for emulation tests.
• Added rocsparse_[s|d|c|z]csritilu0_compute_ex routines for iterative ILU
• Added rocsparse_[s|d|c|z]csritsv_solve_ex routines for iterative triangular solve
• Added GPU_TARGETS to replace the now deprecated AMDGPU_TARGETS in cmake files
• Added BSR format to the SpMM generic routine rocsparse_spmm

Changed

• By default, build rocsparse shared library using --offload-compress compiler option which compresses the fat binary. This significantly reduces the shared library binary size.

Optimized

• Improved the performance of rocsparse_spmm when used with row order for B and C dense matrices and the row split algorithm, rocsparse_spmm_alg_csr_row_split.
• Improved the adaptive CSR sparse matrix-vector multiplication algorithm when the sparse matrix has many empty rows at the beginning or at the end of the matrix. This improves the routines rocsparse_spmv and rocsparse_spmv_ex when the adaptive algorithm rocsparse_spmv_alg_csr_adaptive is used.
• Improved stream CSR sparse matrix-vector multiplication algorithm when the sparse matrix size (number of rows) decreases. This improves the routines rocsparse_spmv and rocsparse_spmv_ex when the stream algorithm rocsparse_spmv_alg_csr_stream is used.
• Compared to rocsparse_[s|d|c|z]csritilu0_compute, the routines rocsparse_[s|d|c|z]csritilu0_compute_ex introduce a number of free iterations. A free iteration is an iteration that does not compute the evaluation of the stopping criteria, if enabled. This allows the user to tune the algorithm for performance improvements.
• Compared to rocsparse_[s|d|c|z]csritsv_solve, the routines rocsparse_[s|d|c|z]csritsv_solve_ex introduce a number of free iterations. A free iteration is an iteration that does not compute the evaluation of the stopping criteria. This allows the user to tune the algorithm for performance improvements.
• Improved user documentation

Resolved issues

• Fixed an issue in rocsparse_spgemm, rocsparse_[s|d|c|z]csrgemm, and rocsparse_[s|d|c|z]bsrgemm where incorrect results could be produced when rocSPARSE was built with optimization level O0. This was caused by a bug in the hash tables that could allow keys to be inserted twice.
• Fixed an issue in the routine rocsparse_spgemm when using rocsparse_spgemm_stage_symbolic and rocsparse_spgemm_stage_numeric, where the routine would crash when alpha and beta were passed as host pointers and where beta != 0.
• Fixed an issue in rocsparse_bsrilu0 where the algorithm was running out of bounds of the bsr_val array.

Upcoming changes

• Deprecated rocsparse_[s|d|c|z]csritilu0_compute routines. Users should use the newly added rocsparse_[s|d|c|z]csritilu0_compute_ex routines going forward.
• Deprecated rocsparse_[s|d|c|z]csritsv_solve routines. Users should use the newly added rocsparse_[s|d|c|z]csritsv_solve_ex routines going forward.
• Deprecated AMDGPU_TARGETS using in cmake files. Users should use GPU_TARGETS going forward.

rocSPARSE 3.3.0 for ROCm 6.3.3

rocSPARSE code for ROCm 6.3.3 did not change. The library was rebuilt for the updated ROCm 6.3.3 stack.