This PR cleans many warning messages (more than 500) triggered by -Wextra. A detailed list of changes is given below:
* Fix maybe-uninitialized warnings
* Fix unused-parameter warnings
* Fix sign-compare warnings
* Fix implicit-fallthrough warnings
* Add HYPRE_UNUSED_VAR
* GenerateCoordinates -> hypre_GenerateCoordinates
This PR relaxes the check of CUDA compile flags, which checks the equality of the major CUDA version of the compiled code, and also the minor version should not be larger than the GPU running on.
* Add hypre_Solver struct
* Add hypre_PCGSetPreconditioner
* Add hypre_Solver info to AMG's struct
* Use public level function pointers
* Add HYPRE_PtrToDestroyFcn
* Remove redundant definitions and put external base objects in HYPRE_utilities.h
* Make use of new HYPRE_PCGSetPreconditioner in the IJ driver
---------
Co-authored-by: Rob Falgout <rfalgout@llnl.gov>
* Fix segfault in ILU solve when not using reordering and running on host (only affects block Jacobi case)
* Fix segfault in ILU solve when using iterative triangular solves with block Jacobi
---------
Co-authored-by: Daniel Osei-Kuffuor <oseikuffuor1@llnl.gov>
This enhances what print_level can achieve in MGR. Particularly, now we can dump linear system info to files according to the print_level code. We also have the ability now of printing a sequence of linear systems to file (useful when hypre is used in time-stepping application).
A detailed list of changes is given below:
* Add utilities for creating/checking directories
* Add print_level codes to MGR and new info_path member
* Add hypre_MGRDataPrint
* Add call to hypre_MGRDataPrint and logic to update the print_level variable
* Update MGRSolve with new print_level logic
* Remove hypre_MGRWriteSolverParams
* Update documentation for HYPRE_MGRSetPrintLevel
* Implement new logic for HYPRE_MGR_PRINT_MODE_ASCII
This PR improves statistics reporting for MGR.
A list with detailed changes is given below:
* Add MatrixStats and MatrixStatsArray
* Add hypre_squared utility
* Fix divisor line location for Rectangular matrices
* Minor fix on hypre_squared definition
* Move nonzero variable definitions up
* Initialize global number of nonzeros at matrix creation
* Add hypre_ParCSRMatrixStatsArray and helper functions
* Add par_csr_matstats_device
* Add par_mgr_stats
* Print F-relax data only once
* Fix clang-13 build
* IJ driver now passes in print_level option to MGR
* GPU runs always require A_FF in MGR
* Add HYPRE_PRINT_SHIFTED_PARAM macro
* Add hypre_IntArraySetInterleavedValues (host/device implementations)
* Fix F-relaxation reporting + refactoring
* Update global number of nonzeros of the matrix
* Move new BoomerAMG functions to par_stats
* Apply astyle
This PR adds CUDA support to dense direct solver options (98, 99, 198, and 199) of BoomerAMG and MGR:
- Options 98 and 99 compute the LU factorization with pivoting.
- Options 198 and 199 compute the dense inverse matrix explicitly.
Detailed list of changes below:
* Add hypre_ParCSRMatrixToCSRMatrixAll_v2
* Add hypre_SeqVectorMigrate
* Add hypre_ParVectorToVectorAll_v2
* Refactor implementation of BoomerAMG's Gaussian Elimination
* Add hypre_GaussElimAllSetup and hypre_GaussElimAllSolve
* Add device support via MAGMA and cuSOLVER to BoomerAMG's LU coarsest linear solver (options 98, 99)
* Add device support via MAGMA and cuSOLVER to BoomerAMG's exact inverse solver (options 198, 199)
* Add wrappers to MAGMA's getrf and getrs
* Add MAGMA info on AMG stats + code formatting
* Add wrappers to cuSOLVER and cuBLAS functions
* Add wrapper hypre_magma_getri_nb
* Add header file for collecting hypre functors
* Add memory location to Gaussian elimination data structure
* Improve description of coarsest level solver options
* Update GE data structure in MGR
* Change Ainv to Awork
* Updated documentation for clarity and to clean up a few typos.
* Add warning messages to FEI, ParaSails, PILUT, Euclid.
* Improved and updated GPU information
* Added CMake build information
This solves an out-of-bounds memory error during `hypre_BoomerAMGSetup` when called multiple times without a call to `hypre_BoomerAMGDestroy` interleaved. This pull request makes sure that `smooth_num_levels` is reset to `hypre_ParAMGDataSmoothNumLevels(amg_data)` before the smoothers variable is allocated.
This PR adds new Print and Read functions for matrices and vectors to be stored/read in binary format. A detailed list of changes is given below:
* Add IJMatrix/ParCSRMatrix routines for binary I/O
* Add IJVector/ParVector routines for binary I/O
* Add typedefs for unsigned integer types and single-precision floating-point
* Change char sizes to HYPRE_MAX_FILE_NAME_LEN
* Add options to IJ driver for reading binary matrices/vectors
* Add regression tests for IJ input/output
Allow the use of MAGMA as local linear solver for FSAI.
Add `HYPRE_FSAISetLocalSolveType` for choosing the local linear solve type used in FSAI and add `HYPRE_BoomerAMGSetFSAILocalSolveType` for the case when FSAI is used as a smoother to BoomerAMG.
This PR adds CUDA and HIP support to FSAI according to a static pattern generation algorithm. The resulting method can also be used as a preconditioner for BoomerAMG. A detailed list of changes is given below:
* Add par_fsai_device.c
* Add hypre_FSAIApply
* Add function to dump local linear systems in dense format
* Implement static FSAI pattern computation via powers of A
* Improve filtering of candidate pattern
* Improve local linear systems extraction
* Add option for a 125pt matrix (27pt squared)
* Add options to control sizes of the memory pools with umpire
* Add hypre_GpuProfiling calls
* Improve candidate pattern truncation times
* Add max_nnz_row member and its private and public functions to FSAI
* Use max_nnz_row in FSAISetupDevice
* Add num_levels member and its private and public functions to FSAI
* Add threshold member and its public/private functions to FSAI
* Expose FSAI algorithm type to BoomerAMG
* Expose options to control FSAI setup
* Add cuSOLVER variables and calls
* Add batched dense linear solver calls to FSAI
* Improve execution time for generating random numbers
* Show FSAI parameters when amg_print_level >= 1
* Improve output of FSAIPrintStats
* Implement warp calls
* Add hypre_mask type and hypre_ballot_sync wrapper function
* Add hypre_popc and hypre_ffs wrapper functions
* Implement warp_allreduce_max calls
* Change: hypreDevice -> hypre_*Device
* Add rocSOLVER calls
* Apply astyle
* Remove redundant line
Add warnings for Euclid and PILUT redirecting users to hypre-ILU.
Rewrite hypre-ILU overview section.
Add new sections to hypre-ILU documentation: "User-level functions", "ILU as smoother for BoomerAMG", and "GPU support".
Include info about new iterative ILU options.
Update BoomerAMG complex smoothers section.
Change name "hypre-ILU" to "ILU"
Modified from #718, this PR squashes out zero columns of the off-diagonal part of a `hypre_ParCSRMatrix`.
The issue was in offd there exist empty columns (columns with no nonzeros), which correspond to "useless" entries in col_map_offd. This caused issues in at coarser grids in the communications with large number of ranks. We added a routine to compress the zero columns out and shorten col_map_offd. This should reduce communication cost even at higher levels.
Two sources of the empty columns have been located and fixed:
- Truncation after building P
- P^T(AP): only the transpose multiplication part.
---------
Co-authored-by: Noel Chalmers <noel.chalmers@gmail.com>
Co-authored-by: Ruipeng Li <li50@llnl.gov>
Co-authored-by: Wayne Mitchell <mitchell82@llnl.gov>
This PR adds HIP support to hypre_ILU (setup and solve phases):
- Algorithm type 0 (BJ-ILU0)
- Algorithm type 10 (GMRES-ILU0)
- Iterative triangular solves for backward and forward substitutions.
---------
Co-authored-by: Paul Mullowney <Paul.Mullowney@nrel.gov>
