News

• added missing hpro_matrix_conjugate/transpose in C bindings and
• simplified eigen functions in BLAS to prevent instantiation problems
• re-added aca_dense_fallback parameter

• fixed bug in Cholesky factorization

• clustering based on space filling curves (using CGAL library) in TSFCCTBuilder
• lowrank truncation support based on Frobenius norm
• support for adaptive quadrature order in BEM kernels
• added restriction to real/imaginary part of matrices (restrict_re/im)
• added make_symmetric/hermitian to enforce symmetry status for given matrices

• using generic value types for all major types (matrices, vectors and linear operators)
• officially renamed namespace and header files to Hpro and hpro (old names still valid for compatibility)
• C bindings split into separate functions per value type; also with hpro prefix now (old function names and functionality is still supported)
• support for mixed precision computations for matrix vector multiplication and in solver classes

• added apply_add with BLAS::Matrix as argument to TLinearOperator classes
• added absolute_prec to define TTruncAcc
• added support for NEON instruction set (Apple M1)
• added support for Mongoose graph partitioning library (TMongooseAlgPartStrat)
• enhanced TAlgAdmCond to define maximal number of allowed connecting edges
• fixes:
  • issue in BLAS::qrp for matrices with nrows < ncols
  • coefficient tests in TPermCoeffFn (was missing before)
  • issue in TMatrixProduct with single factor
  • issue in TMaternCovCoeffFn with different row/column coordinates

• replaced HLIB::complex by std::complex
• removed old DAG interface
• modified pivot strategy of standard ACA (better, more robust convergence)
• using geqr2 instead of geqrf for QR factorization (slightly faster)
• fixes:
  • replaced deprecated features of TBB
  • TGeomGroupCTBuilder>: fixed handling of offsets
  • missing instantiation of BLAS::random for BLAS::Vector added
  • wrong solving flags in TLDUInvMatrix
  • in-efficient dependency handling in DAG construction for TLR/Tile-H
  • memory leak in recursive DAG construction fixed

• fixed weak admissibility (was actually standard admissibility)
• fixed MBLR clustering (ordering only in one dimension led to extremely rectangular clusters)
• fixed bug in lowrank approximation (wrong type conversion)
• fixed non-SIMD implementation of TExpBF (still had additional factor)
• fixed TDenseCoeffFn in case given dense matrix had non-zero row/column offsets
• replaced tbb::mutex and tbb::atomic by std versions since marked obsolete in recent TBB versions

• improved new DAG generation system (better speed and parallel scalability) and made it the default system (old version still available with CFG::dag_version=1)
• moved non-generic include files into hpro sub-directory for better separation with other libraries
• added various approximation routines for sums of matrices, operators, e.g., SVD, pairwise SVD, Rand-SVD, Rand-LR and ACA for operator sums
• expanded lazy accumulator arithmetic to move all updates to leaves only (evaluating all updates simultaneously; see CFG::Arith::lazy_eval and CFG::Arith::sum_approx)
• added TZeroMatBuilder and build_zero_mat to construct empty matrix for given block clusters tree, e.g., as pre-initialized result of other H-matrix operation
• parallelization of various routines during clustering, e.g., sorting, etc. (may result in slightly different clustering with different number of CPU cores)
• bug fixes: -permutation of dense matrix in TMatBuilder removed (inconsistent behaviour compared to if H-matrix is built) -fixed update of aux. data in H-matrix in copy_nearfield -fixed reading of old HLIB files (wrong processor sets) -fixed return value of Mem::usage -fixed various issue when using single precision

• fixed various compiler issues with MS Visual C++
• added build method for coefficient functions to return dense matrix for given index set
• C bindings:
  • added hlib_matrix_to_dense/rank
  • added hlib_matrix_approx_rank to compute low-rank approximation of given matrix with different methods
• fixed issue in TBSPNDCTBuilder when no interface is present
• fixed issue in HLIB::Mem::usage

• fixed issue with HDF5 library but removed support from binary distribution due to linking problems with newer version of libHDF5
• added missing functions to sequential NET interface
• added functions to directly set LR matrices in TRkMatrix
• using internal grid generation also in example code (laplace/helmholtz)
• additional spherical grid (different start grid for “inbetween” steps)
• improved coordinate visualization in PostScript format (better minimal distance estimate)

• new DAG generation based on recursive algorithms with automatic deduction of dependencies between nodes (default: previous DAG; see CFG::dag_version)
• new coefficient function for Matern kernel (TMaternCovCoeffFn) and exponential bilinear form (TExpBF)
• functions for computing low-rank approximations for sum of matrices directly (using pair-wise SVD approx_sum_svd, randomized SVD approx_sum_randsvd or randomized low-rank approx approx_sum_randlr.)
• ACA: -modified stop criterion for ACA (user controllable maximal rank CFG::Build::aca_max_ratio) -added dense fallback for ACA if not converging with only computing those coefficients, not yet computed
• added MBLR cluster tree construction (TMBLRCTBuilder)
• modified handling of matrix coefficient functions, especially TPermCoeffFn
• (limited) grid generation and refinement in BEM library
• using libmvec for sin, cos and exp if available (glibc v2.22 and up) with significant speedups in complex valued computations
• new academic license without any user/host or date limitation

• Accumulator based H-arithmetic reducing number of truncations with support for lazy and eager evaluation
• added randomized SVD and implemented dense approximation and lowrank truncation for all types (SVD, RRQR and Rand-SVD)
• also added lowrank approximation algorithms for RRQR and Rand-SVD for H-construction (TRRQRLRApx, TRandSVDLRApx)
• support for special flat H-hierarchy with optimised arithmetic functions, e.g., in-place inversion.
• support for block refinement during matrix construction, e.g., if admissibility gives false positives
• added infinity matrix norm (TInfinityNorm and norm_inf())
• implemented TOffDiagAdmCond with all off-diagonal blocks being admissible
• massive code restructuring and cleanup
• initial support for HDF5 matrix IO (dense and lowrank)
• support for VSX instruction set (POWER CPUs)
• special handling for all BLAS functions in case of parallel Intel MKL
• added parameter configuration with config files
• added some functions to simplify solver stopping criterion
• changed behaviour/incompatibilities with previous versions:
  • removed all permutation handling from THMatrix and TNearfieldMulVec
  • no recompression in ACA/HCA (now only in matrix builders)
  • ptrcast() now consistent with cptrcast(), i.e., no * needed
  • some parameter reorganization
  • previous TMatrix::copy_struct is renamed to TMatrix::copy_struct_from (TMatrix::copy_struct will now return a matrix copy without data)

• added operator for matrix sum (TMatrixSum) in addition to matrix products
• missing bindings in C interface (matrix product/sum, apply, apply_add)
• fixed issue in clustering with predefined partition with many groups
• support for more matrix formats and more robust IO (if files do not follow standard) for Harwell-Boeing/Matrix-Market format
• re-enabled parallel block cluster tree construction on shared memory
• bug fixes:
  • in solve_diag_left_block
  • matrix solves in TLLInvMatrix

• implemented rank revealing QR based low-rank truncation
• Solvers:
  • added CGS and TFQMR solvers
  • added support for matrix solves in linear iteration (also \mcH-matrices!)
  • optional computation of exact residual during iteration
  • simplified handling of stop criterion parameters
  • using status field in TSolverInfo instead of exception if solver fails (e.g., breakdown)
  • some code restructuring
• support for block-wise Jacobi and Gauss-Seidel operators
• support for AVX512 instruction set
• many new user controllable parameters
• misc.:
  • handling of diagonal in factorisation (inverse or normal) now a runtime option (default: inverse)
  • added optional distance for TWeakAlgAdmCond to support distances other than one
  • support for fixed rank 0
  • correct progress bar support for WAZ factorisation and inversion
  • some reorganization of source/header files
• C bindings:
  • added hlib_admcond_geom_hilo for THiLoFreqGeomAdmCond
  • additional parameter for blockdiag functions (blocksize)
• fixed serious issue with Intel TBB and with current Intel TBB-based Intel MKL
• various bug fixes

• Added factorization of inverse matrix WAZ = I, enabling vector solves using matrix vector mult. instead of forward/backward solves with much better parallel speedup.
• Significant improvements in parallel performance of matrix inversion.
• Improved performance of LU, matrix-vector mult., forward/backward solves.
• Added function nearfield_sparse to extract H-matrix nearfield as sparse matrix.
• Switched to adaptive_split_axis as default for clustering.
• Minor Changes:
  • Additional options for matrix visualization (colormap, etc.)
  • Basic VTK output of block clusters.

• fixed various bugs and race conditions
• extended ctors of various matrix classes to accept optional value type field
• added example on how to assemble block matrices

• Fixed two bugs in point-wise LU.
• Solver changes:
  • Refactored solver classes (no interface changes); added TRichardson to replace TSolver in the future.
  • Fixed inconsistent computation of residual norm in solvers. Now Richardson, CG and BiCG will compute standard residual norm, while MINRES and GMRES compute preconditioned residual norm.
  • Made initialisation of start vector in solver classes optional (function initialise_start_value)
• Added function diagonal to extract diagonal of a matrix.
• Added example spectrum to compute spectrum of graph Laplacian (see also documentation).

• Fixed issues when solving dense matrices (used in new example for many RHSs).
• Modified THiLoFreqGeomAdmCond: now maximal number of wavelengths per cluster is tested.
• Refactored geometrical clustering classes and partitioning strategies, thereby fixing several issues.
• C++11 changes:
  • most object creating functions now return std::unique_ptr,
  • replaced typedef by using,
  • added iterators for TIndexSet, TNodeSet, TGraph, TProcSet (for range based for).
• Added parameter to algebraic clustering in C bindings to define partitioning algorithm (BFS, multi level, METIS or Scotch).
• Fixed issues with progress bar during factorisation (wrong block count).
• Removed BSP style communication functions (MPI only now).
• Finished conversion to new packed_t SIMD type. Using SSE3 instead of SSE2.
• Added lock to TScotchAlgPartStrat because Scotch is not multi thread safe.

• Removing implicit reordering of unknowns during matrix-vector multiplication to fix inconsistent behaviour. Please use permutations from cluster trees or ℋ-matrices to reorder vectors or TPermMatrix to represent permuted matrices instead.
• Speedup improvements for matrix inversion. Triangular inversion and matrix multiplication available in standard user interface.
• Import/export from/to CCS/CRS matrices simplified.
• Simplified (and faster) mutex wrapper.
• Several C++11 changes.

• Removing reference counters in BLAS interface due to major performance issue on multi-core (-socket) systems. See documentation on how to use the modified interface (and avoid errors).
• New, scalable matrix-vector multiplication implemented.
• Using generic datatype for SIMD instructions, thereby enabling generic SIMD algorithms, e.g. for BEM kernels, and fast adoptation of new SIMD instructions, e.g. AVX2.
• Removed TVirtualVector (replaced by TScalarVector).
• and, as usual: several bugs fixed

• fixed race condition in C bindings
• fixed issue with initialisation of static variables

• fixed some bugs

• Major Changes
  • Switched from OpenMP to Threading Building Blocks as interface to shared memory parallelism, thereby also changing most algorithms to task-based parallelism.
  • Reducing dependency on external libraries by using C++11 features. Also replacing some classes by default C++ versions (finally removing old code).
  • Alternative, non-recursive, level-wise ℋ-LU factorisation based on explicit block dependencies, which provides far better speedup on many-core systems, e.g. Intel MIC architecture.
  • New H-LU factorisation algorithm also applicable in distributed environments, yielding better load-balancing (albeit with limited speedup).
  • Added support for multiple CPUs to many algorithms, e.g. in clustering, norm computations, matrix-vector multiplication and solves, H²-conversion.
  • Minor Changes
    • Optimised BEM kernels for Intel MIC architecture.
    • Introduced TLinearOperator for operators not supporting TMatrix functionality, e.g. factorised matrices.
    • HLIBpro file format changed due to internal changes and due to some bugs in the format. However, backward read compatibility for most files written with earlier versions is kept.
    • Added Support for Cairo library, thereby providing PDF output.
• And of course: many smaller feature upgrades and bug fixes.

• Matrix Construction:
  • Switched to template based coefficient functions (TCoeffFn and derived) and all depended classes, e.g. TDenseMBuilder, SVD and ACA low rank approximation.
  • Rewrote HCA:
    • Simpler interface containing all neccessary functionality in single class.
    • Using template for value type.
    • Added base classes for permuted indices and for BEM applications using quadrature.
    • Added implementation for Laplace and Helmholtz also for linear ansatz spaces and with support for SSE2 and AVX.
  • Cleaned up ACA implementation.
  • Changed handling of recompression: should now be handled by default for low rank approximation algorithm and not by matrix construction class (to avoid recompression of optimal results).
• Clustering Changes:
• Added TNDBSPPartStrat to be used in connection with nested dissection (trying various clusterings and choosing best for ND).
• Modified TNDBSPCTBuilder to more resemble algebraic version, e.g. average depth for interface clusters instead of maximal.
• Fixed bug in PCA based clustering and added version for cardinality based clustering.
• Added various flags to modify clustering, e.g. synchronisation of interface depth, enforcing block clusters with same depth of corresponding clusters, using symmetrised weights in algebraic clustering.
• Input/Output and visualisation:
  • Fixed bug in reading dense matrices.
  • Changed order of dimension for coordinate IO using Matlab format: now ncoord × dimension (e.g. as also used by Sparse Matrix Collection).
  • Added VTK visualisation for coordinates (with various options, e.g. marking clusters or index connectivity) and BEM grids.
  • Added Output of Grids in HLIB format.
  • Added coordinate IO in MatrixMarket format.
• Changes in LAPACK wrapper:
  • added LAPACK workspace queries for optimal workspace size instead of using predefined block size
  • using xGESDD for large matrices
• various bug fixes.

• Deactivated default coarsening during matrix construction.
• Added special H² matrix builder with predefined cluster bases.

• changes in BEM code:
  • Added support for AVX.
  • Performance speedups in SSE2 implementation of Helmholtz and Maxwell kernels.
  • Runtime detection of SSE2/AVX availability and automatic choice of optimal kernel.
• Added matrix_format function to matrix coefficient functions to define whether unsymmetric, symmetric or hermitian (default: unsymmetric).
  Default build function in matrix builders now without format argument.
• Added support for ILP64 BLAS/LAPACK implementations (64bit integers).
• Added support for AMD-LibM (integrated in binary Linux distributions).
• Added vector IO in MatrixMarket format.
• Cleaned up C++ examples (thereby also removing Boost link dependency).
• several bug fixes

• OpenMP exception handling changed: now all threads will stop as soon as possible in case of an error
• fixed several, previously undetected, non-critical compiler warnings (MS Visual C++)
• bug fixes

HLIBpro v1.0 is a major rewrite/reorganisation of many of the H-matrix algorithms. The following list of changes only covers the main topics and is by far not complete.

• added distributed computing via MPI for matrix construction and factorisation
• added H²-matrices
• added internal multi-level graph partitioning for blackbox clustering
• added support for piecewise linear basis functions and Maxwell EFIE/MFIE
• rewrote interface to BLAS/LAPACK
• rewrote C interface with better mapping of internal C++ and C types
• increased robustness of matrix factorisation in case of bad-conditioned matrices
• increased speedup of matrix factorisation in multi-threaded computations
• many performance improvements and bug fixes

• added optional diagonal scaling of H-matrices during LU factorisation
• added blockwise accuracy, e.g. accuracy depending on current matrix block
• rewrote accuracy handling in C bindings
• simplified BSP partitioning methods and added regular cardinality based and principle component based clustering
• added optional balancing of tree depth in cluster tree construction with predefined partitioning
• implemented optional double precision computation of matrix inversion and low-rank truncation in single precision mode
• fixed bug in calling single precision norm functions of LAPACK
• fixed bug in PostScript output and modified H-matrix output in PostScript format
• added support for Jacobi based SVD (sgejsv and dgejsv) in LAPACK v3.2

• removed ID based cluster tree computations in matrices
• always computing SCC in algebraic clustering, also in nested dissection clustering
• reordering clusters depending on size ratio (large first)
• fixed bug with filenames without directories
• fixed non-exception safe OpenMP usage
• added matrix reduction to nearfield part
• added dense low-rank multiplication if result is large dense matrix

• fixed solve functions in TLU, TLDL (checking for NULL blocks)
• fixed OpenMP call with zero threads in TLU TLDL and TMatrixInv
• fixed operator = in autoptr (wrong const)
• removed unnecessary checks in TArray::copy
• fixed recursive call in restrict_blockdiag
• replaced fixed constants by type dependent constants in lapack.cc
• fixed TMatrixInv::multiply_diag when only D is dense

• fixed several warnings from Visual C++ and Intel C++ compilers
• moved all global variables and functions into HLIB namespace (except xerbla override)
• enabled user defined prefix for functions and types in C interface and added override for namespace name
• reactivated cardinality check when using HLIB_BSP_AUTO

• replaced threads and mutices by OpenMP (thread start only, no scheduling)
• included log file support in addition to stdout
• added parallel LDL^T factorisation (DD and blockdiag only)
• added parallel blockdiag LU factorisation
• added zero approximation during matrix construction (for nearfield only)
• fixed bug in algebraic nested dissection clustering (wrong path length in interface)

• reduced memory consumption/fragmentation in ACA generated matrices with large rank
• added Fiduccia/Mattheyses bisection optimisation for BFS clustering
• added FFT for vectors by implementing support for FFTW3 (optional)
• fixed bug in TBSPPartCTBuilder when using more than two partitions
• fixed potential issues in sorting algorithms
• fixed type issues with *_bytesize functions in C interface
• fixed bug in PostScript visualisation of matrices if matrix norm is zero
• fixed issues with GCC-4.3
• fixed bug in command line parsing of configuration system
• minor modifications to SCons system to increase userfriendliness

• general Algorithmic Changes
  • support for single precision arithmetic; has to be decided before compiling HLIBpro
  • made complete C++ functions and classes visible from outside instead of just C interface functions
  • rewrote complex arithmetic to distinguish between symmetric and hermitian matrices; added LDL^H and LL^H factorisations
  • inversion now based on LU, thereby reducing memory consumption (roughly halved)
  • added computation of the diagonal of the inverse without computing the inverse
  • added evaluation of LU, LDL^T factorisations (instead of just solving)
  • removed point-wise LU and LDL^T factorisation (only blocked) to improve robustness with zeroes on diagonal
  • added (optional) check and fix for singular sub matrices during inversion and factorisation
  • added complex valued HCA
  • new version of ACA+
  • multiplication C = ADB with diagonal D implemented
  • implemented bilinear forms for Helmholtz single and double layer potential
  • implemented bilinear form for acoustic scattering
  • rewrote algebraic clustering for sparse matrices; added support for Scotch and CHACO
  • added support for periodic coordinates in clustering
  • added clustering with user defined index partition on first level in cluster tree
  • added standard admissibility for algebraic clustering
  • added maximal level in clustering to prevent infinite recursion
  • modified solvers to handle complex valued data
  • added permutation of dense matrices without temporary storage (needed in IO)
• parallel Arithmetic
  • added thread parallel algorithms for matrix construction, matrix multiplication, inversion and LU factorisation
  • redesigned thread pool, thereby fixing race conditions
  • added support for Windows threads
  • fixed several issues with thread safety
• Input and Output
  • added general I/O functions with autodetection of file format
  • added output of matrices in Harwell/Boeing format
  • added MatrixMarket format
  • added support for Ply and surface mesh format (NetGen) for Grid I/O
  • fixed format errors in SAMG output
  • conversion of arbitrary matrices to sparse format when writing in SAMG or Harwell/Boeing format
  • fixed support for symmetric matrices in Harwell/Boeing format
• C interface
  • prefixed all functions, types and constants with hlib_ (or HLIB_) to prevent collisions with other definitions (OS or libraries)
  • added support for C99 complex types (if available)
  • added hlib_set_coarsening to activate/deactivate coarsening during matrix construction (default: on) and matrix arithmetic (default: off)
  • added hlib_matrix_inv_diag to return diagonal of inverse
  • added hlib_matrix_is_complex to test for real or complex valued matrices
  • added hlib_set_nthreads to set number of threads
  • added hlib_coord_t as special type for coordinates
  • separated stop criterion and solver in solver interface
• Miscellaneous
  • updated CPUflags and Rmalloc
  • fixed optimisation issues (leading to infinite loops) in enclosed CLAPACK

• Algorithmic Changes
  • added (blocked) LDL^T factorisation (now default for symmetric matrices)
  • no longer need extra matrix in matrix inversion
  • using ACAFull in HCA (instead of SVD)
  • adaptively choosing quadrature and interpolation order in ACA and HCA
  • rewrote matrix addition to support general cases, e.g. low-rank to blocked
  • rewrote low-rank truncation handling
  • support for METIS in algebraic clustering routines
  • added basic support for “dense” sparse matrices, e.g. with highly coupled indices
  • added SSE2 based HCA algorithm
  • added infinity norm for vectors
  • using norm of preconditioned residual for all solvers if preconditioner is present
  • added MINRES iteration
  • using ADM_AUTO as default admissibility
  • finally removed all asserts and replaced by internal error checking
• Input and Output
  • VRML97 support
  • added Matlab compression (Matlab v7) and structs support
  • support for Harwell-Boeing matrix format (read-only)
  • modified PostScript output of block-wise SVD; now scaled w.r.t. 2-norm of matrix
• OS and Library support
  • MS Windows support
  • shared libraries for Linux and Windows
  • changed configure system to better handle MS Windows environment
  • added internal xerbla to handle LAPACK errors directly
• C interface
  • automatic choice of matrix building in hlib_matrix_build_bem_grid
  • introduced vector_t as type to vectors (no more C arrays)
  • added Gauss and Sauter triangle quadrature rules
  • added functions to access matrix and vector entries
  • added copyto and copyto_eps functions
  • added hlib_matrix_build_dense to build H-matrix from dense matrix
  • changed solver management
• Miscellaneous
  • several improvements and bug fixes
  • cleaned up error codes
  • updated CPUflags and Rmalloc

• Arithmetic
  • added ACA-Full
  • added HCA (hybrid cross approximation)
  • complex valued ACA and SVD
  • added copy with coarsening for H-matrices
  • added computation of spectral norm for the inverse of a matrix
  • support for permutations in matrix-vector multiplication of sparse matrices
  • added support for Laplace SLP/DLP and 3D triangle surface grids
  • fixed issues with degenerated bounding boxes in geometrical clustering
• Input/Output
  • support for PLTMG matrix format
• Miscellaneous
  • replaced error handling with exceptions
  • added modified CLAPACK as default implementation of LAPACK to HLIBpro
  • integrated CPUFlags into configure system
  • added function for fast reciprocal square root

• Arithmetic
  • initial support for complex arithmetic
  • support for symmetric matrices in arithmetic
  • implemented block LU factorisation
  • implemented LDL^T factorisation
  • added Frobenius norm for sparse matrices
  • support for CRS format in sparse matrices
  • added Jacobi and SOR matrix types (for matrix-vector multiplication)
  • implemented hierarchical domain decomposition with parallel arithmetics
• Parallel Algorithms
  • thread-parallel Cholesky factorisation
  • thread-parallel coarsening of H-matrices
  • fixed thread-parallel LU and inversion
  • fixed dead-locks in thread-pool
  • added direct communication in BSP mode
  • parallel addition of matrices and vectors via streams
• Input/Output
  • support for Matlab and SAMG format
• Miscellaneous
  • introduced C interface functions and types
  • added configure system for Makefiles
  • added progress meter support for arithmetic
  • added internal RTTI system
  • support for memory consumption query on HP-UX
  • rewrote error handling

• first public version as PHI (Parallel H-matrix Implementation)
• merged BSP-parallel and thread-parallel versions of H-matrix library