mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-03-07 18:27:40 +08:00
Code Issues Packages Projects Releases Wiki Activity
acce4dd050
eigen/test
History
Rasmus Munk Larsen acce4dd050 Change Eigen's ColPivHouseholderQR to use the numerically stable norm downdate formula from http://www.netlib.org/lapack/lawnspdf/lawn176.pdf, which has been used in LAPACK's xGEQPF and xGEQP3 since 2006. With the old formula, the code chooses the wrong pivots and fails to correctly determine rank on graded matrices. 
This change also adds additional checks for non-increasing diagonal in R11 to existing unit tests, and adds a new unit test with the Kahan matrix, which consistently fails for the original code.

Benchmark timings on Intel(R) Xeon(R) CPU E5-1650 v3 @ 3.50GHz. Code compiled with AVX & FMA. I just ran on square matrices of 3 difference sizes.

Benchmark               Time(ns)     CPU(ns) Iterations
-------------------------------------------------------
Before:
BM_EigencolPivQR/64        53677       53627      12890
BM_EigencolPivQR/512    15265408    15250784         46
BM_EigencolPivQR/4k  15403556228 15388788368          2

After (non-vectorized version):
Benchmark               Time(ns)     CPU(ns) Iterations  Degradation
--------------------------------------------------------------------
BM_EigencolPivQR/64        63736       63669      10844         18.5%
BM_EigencolPivQR/512    16052546    16037381         43          5.1%
BM_EigencolPivQR/4k  15149263620 15132025316          2         -2.0%

Performance-wise there seems to be a ~18.5% degradation for small (64x64) matrices, probably due to the cost of more O(min(m,n)^2) sqrt operations that are not needed for the unstable formula.
2016-01-28 15:07:26 -08:00
..
adjoint.cpp bug #977: avoid division by 0 in normalize() and normalized(). 2016-01-21 20:43:42 +01:00
array_for_matrix.cpp bug #178: remove additional const on nested expression, and remove several const_cast. 2016-01-28 21:43:20 +01:00
array_replicate.cpp Add unit test for m.replicate(...)(index). 2015-06-08 15:42:15 +02:00
array_reverse.cpp Add row/column-wise reverseInPlace feature. 2015-03-31 21:35:53 +02:00
array.cpp Protect digamma tests behind a EIGEN_HAS_C99_MATH check. 2015-12-24 21:28:18 -08:00
bandmatrix.cpp bug #877, bug #572: Introduce a global Index typedef. Rename Sparse*::Index to StorageIndex, make Dense*::StorageIndex an alias to DenseIndex. Overall this commit gets rid of all Index conversion warnings. 2014-12-04 22:48:53 +01:00
basicstuff.cpp Add unit tests for bug #981: valid and invalid usage of ternary operator 2015-09-09 11:38:25 +02:00
bdcsvd.cpp split compiler intensive bdcsvd_1 unit test 2015-06-26 16:14:23 +02:00
bicgstab.cpp Do go to full accuracy when testing BiCGSTAB. 2015-06-05 14:32:26 +02:00
block.cpp Fix compilation of coeff(Index) on sub-inner-panels 2014-09-08 09:50:03 +02:00
cholesky.cpp Factorize VERIFY_EVALUATION_COUNT in unit tests 2015-06-19 16:38:26 +02:00
cholmod_support.cpp bug #1095: add Cholmod*::logDeterminant/determinant (from patch of Joshua Pritikin) 2016-01-22 16:05:29 +01:00
CMakeLists.txt Extend superlu cmake script to check version 2015-11-30 14:48:11 +01:00
commainitializer.cpp
conjugate_gradient.cpp Add unit tests for CG and sparse-LLT for long int as storage-index 2015-03-09 14:33:15 +01:00
conservative_resize.cpp
corners.cpp
ctorleak.cpp Constructing a Matrix/Array with implicit transpose could lead to memory leaks. 2015-04-16 13:25:20 +02:00
cuda_basic.cu Fix compilation of cuda unit test 2015-09-02 16:59:07 +02:00
cuda_common.h NVCC: add more debug info 2014-01-24 12:51:33 +01:00
dense_storage.cpp
denseLM.cpp
determinant.cpp
diagonal.cpp bug #178: remove additional const on nested expression, and remove several const_cast. 2016-01-28 21:43:20 +01:00
diagonalmatrices.cpp Fix bug #987: wrong alignement guess in diagonal product. 2015-03-31 23:35:12 +02:00
dontalign.cpp
dynalloc.cpp Do not check NeedsToAlign if no static alignment 2015-11-30 22:36:14 +01:00
eigen2support.cpp
eigensolver_complex.cpp Limit matrix size for other eigen and schur decompositions 2015-10-30 18:06:03 +01:00
eigensolver_generalized_real.cpp bug #1098: fix regression introduced when generalizing some compute() methods in changeset 7031a851d4 2015-10-26 16:00:25 +01:00
eigensolver_generic.cpp Limit matrix size for other eigen and schur decompositions 2015-10-30 18:06:03 +01:00
eigensolver_selfadjoint.cpp Limit matrix size when testing for NaN: they can become prohibitively expensive when running on x87 fp unit 2015-10-30 14:44:22 +01:00
evaluator_common.h
evaluators.cpp Add unit test for CoeffReadCost 2015-10-09 14:28:48 +02:00
exceptions.cpp
fastmath.cpp fix typo 2015-10-28 20:18:00 +01:00
first_aligned.cpp Generalize first_aligned to take the requested alignment as a template parameter, and add a first_default_aligned variante calling first_aligned with the requirement of the largest packet for the given scalar type. 2015-08-06 17:52:01 +02:00
geo_alignedbox.cpp Fix a warning 2015-10-08 16:27:54 +02:00
geo_eulerangles.cpp Introduce EIGEN_PI, get rid of M_PI and acos(-1.0) 2015-06-10 17:12:10 +02:00
geo_homogeneous.cpp bug #997: add missing evaluators for m.lazyProduct(v.homogeneous()) 2015-06-08 15:43:41 +02:00
geo_hyperplane.cpp Many files were missing in previous changeset. 2015-07-29 11:11:23 +02:00
geo_orthomethods.cpp Fix nested_eval<Product<> > which wrongly returned a Product<> expression 2014-09-24 09:39:09 +02:00
geo_parametrizedline.cpp Many files were missing in previous changeset. 2015-07-29 11:11:23 +02:00
geo_quaternion.cpp Fix matrix to quaternion (and angleaxis) conversion for matrix expression. 2015-12-01 09:45:56 +01:00
geo_transformations.cpp Add missing Rotation2D::operator=(Matrix2x2) 2015-12-03 22:25:26 +01:00
hessenberg.cpp
householder.cpp
incomplete_cholesky.cpp bug #1150: make IncompleteCholesky more robust by iteratively increase the shift until the factorization succeed (with at most 10 attempts). 2016-01-23 22:13:54 +01:00
integer_types.cpp
inverse.cpp Fix shadow warnings triggered by clang 2015-06-09 09:11:12 +02:00
is_same_dense.cpp Fix unit tests wrt EIGEN_DEFAULT_TO_ROW_MAJOR 2015-12-11 10:06:28 +01:00
jacobi.cpp
jacobisvd.cpp Reduce compiler memory consumption for SVD unit tests 2015-06-22 09:58:06 +02:00
linearstructure.cpp Add regression using test for array<complex>/real 2015-11-20 15:29:32 +01:00
lscg.cpp Rename LSCG to LeastSquaresConjugateGradient 2015-03-05 10:16:32 +01:00
lu.cpp Add LU::transpose().solve() and LU::adjoint().solve() API. 2015-12-01 14:38:47 +01:00
main.h Protect all calls to isnan, isinf and isfinite with parentheses. 2015-08-14 17:32:34 +02:00
mapped_matrix.cpp First part of a big refactoring of alignment control to enable the handling of arbitrarily aligned buffers. It includes: 2015-08-06 15:31:07 +02:00
mapstaticmethods.cpp Fix many long to int implicit conversions 2014-07-08 16:47:11 +02:00
mapstride.cpp Many files were missing in previous changeset. 2015-07-29 11:11:23 +02:00
meta.cpp bug #826: fix is_convertible for MSVC and add minimalistic unit test for is_convertible 2014-07-16 13:17:06 +02:00
metis_support.cpp Switch to MPL2 2015-10-23 10:36:33 +02:00
miscmatrices.cpp
mixingtypes.cpp Extend mixing type unit test with trmv, and the following not yet supported products: trmm, symv, symm 2016-01-27 17:29:53 +01:00
mpl2only.cpp Make the IterativeLinearSolvers module compatible with MPL2-only mode 2015-10-26 15:17:52 +01:00
nesting_ops.cpp Fix unit tests wrt EIGEN_DEFAULT_TO_ROW_MAJOR 2015-12-11 10:06:28 +01:00
nomalloc.cpp add nomalloc unit test for rank2 updates 2016-01-27 17:26:48 +01:00
nullary.cpp Re-enable vectorization of LinSpaced, plus some cleaning 2015-10-08 17:27:01 +02:00
packetmath.cpp Merged in ebrevdo/eigen (pull request PR-148) 2015-12-11 11:52:09 +01:00
pardiso_support.cpp
pastix_support.cpp bug #1149: fix Pastix*::*parm() 2016-01-20 19:01:24 +01:00
permutationmatrices.cpp Check number of temporaries when applying permutations 2015-06-19 16:39:24 +02:00
prec_inverse_4x4.cpp
product_extra.cpp Add regression test for bug #817 2015-09-02 13:16:03 +02:00
product_large.cpp bug #1144: fix regression in x=y+A*x (aliasing), and move evaluator_traits::AssumeAliasing to evaluator_assume_aliasing. 2016-01-09 08:30:38 +01:00
product_mmtr.cpp Add unit test for bug #879 2015-06-16 22:11:41 +02:00
product_notemporary.cpp bug #1144: fix regression in x=y+A*x (aliasing), and move evaluator_traits::AssumeAliasing to evaluator_assume_aliasing. 2016-01-09 08:30:38 +01:00
product_selfadjoint.cpp Workaround dead store warnings in unit tests. 2015-02-18 11:30:44 +01:00
product_small.cpp Extend unit test of coeff-based product to check many more combinations 2015-12-08 11:27:43 +01:00
product_symm.cpp
product_syrk.cpp Workaround dead store warnings in unit tests. 2015-02-18 11:30:44 +01:00
product_trmm.cpp Limit matrix sizes for trmm unit test and complexes. 2015-10-30 15:07:50 +01:00
product_trmv.cpp Workaround dead store warnings in unit tests. 2015-02-18 11:30:44 +01:00
product_trsolve.cpp Extend unit test of dense triangular solvers 2014-07-15 11:15:36 +02:00
product.h bug #1144: fix regression in x=y+A*x (aliasing), and move evaluator_traits::AssumeAliasing to evaluator_assume_aliasing. 2016-01-09 08:30:38 +01:00
qr_colpivoting.cpp Change Eigen's ColPivHouseholderQR to use the numerically stable norm downdate formula from http://www.netlib.org/lapack/lawnspdf/lawn176.pdf, which has been used in LAPACK's xGEQPF and xGEQP3 since 2006. With the old formula, the code chooses the wrong pivots and fails to correctly determine rank on graded matrices. 2016-01-28 15:07:26 -08:00
qr_fullpivoting.cpp Use VERIFY_IS_EQUAL instead of VERIFY(a==b) to get more feedback in case of failure 2015-06-26 16:22:49 +02:00
qr.cpp reduce false negative in the qr unit test 2014-02-14 09:58:30 +01:00
qtvector.cpp
rand.cpp Specified signedness of char type in test so that test passes 2015-09-28 14:26:10 -07:00
real_qz.cpp Addendum to last patch: k is Index and not int 2015-03-31 00:42:14 +02:00
redux.cpp Refactoring of the cost model: 2015-10-28 11:42:14 +01:00
ref.cpp Workaround i387 issue in unit test 2015-11-16 13:33:54 +01:00
resize.cpp
rvalue_types.cpp
schur_complex.cpp Limit matrix size when testing for NaN: they can become prohibitively expensive when running on x87 fp unit 2015-10-30 14:44:22 +01:00
schur_real.cpp typo 2015-10-30 22:16:23 +01:00
selfadjoint.cpp
simplicial_cholesky.cpp Skip too large real-world problems for solvers that do not scale (e.g., SimplicialLLT without reordering) 2015-06-09 09:29:53 +02:00
sizeof.cpp bug #973: update macro-level control of alignement by introducing user-controllable EIGEN_MAX_ALIGN_BYTES and EIGEN_MAX_STATIC_ALIGN_BYTES macros. This changeset also removes EIGEN_ALIGN (replaced by EIGEN_MAX_ALIGN_BYTES>0), EIGEN_ALIGN_STATICALLY (replaced by EIGEN_MAX_STATIC_ALIGN_BYTES>0), EIGEN_USER_ALIGN*, EIGEN_ALIGN_DEFAULT (replaced by EIGEN_ALIGN_MAX). 2015-07-29 10:22:25 +02:00
sizeoverflow.cpp Removed redundant typedef which confused old gcc versions. 2015-02-18 01:03:32 +01:00
smallvectors.cpp
sparse_basic.cpp bug #1105: fix default preallocation when moving from compressed to uncompressed mode 2015-11-06 15:05:37 +01:00
sparse_block.cpp Split sparse_basic unit test 2015-03-19 15:11:05 +01:00
sparse_permutations.cpp Fix gcc 4.4 compilation issue 2015-10-30 08:44:37 +01:00
sparse_product.cpp Fix ambiguous instantiation 2015-10-27 11:01:37 +01:00
sparse_ref.cpp bug #1064: add support for Ref<SparseVector> 2015-10-21 09:47:43 +02:00
sparse_solver.h bug #1092: fix iterative solver ctors for expressions as input 2015-10-26 16:16:24 +01:00
sparse_solvers.cpp
sparse_vector.cpp Add SparseVector::conservativeResize() method. 2016-01-25 11:55:39 +01:00
sparse.h Fix ForceNonZeroDiag for complexes 2015-08-04 16:12:16 +02:00
sparseLM.cpp
sparselu.cpp Switch to MPL2 2015-10-23 10:36:33 +02:00
sparseqr.cpp Add unit test for dense = SparseQR::matrixQ 2015-06-24 17:55:41 +02:00
special_numbers.cpp
spqr_support.cpp Fix many long to int conversion warnings: 2015-02-16 13:19:05 +01:00
stable_norm.cpp bug #977: add stableNormalize[d] methods: they are analogues to normalize[d] but with carefull handling of under/over-flow 2016-01-23 22:40:11 +01:00
stddeque.cpp
stdlist.cpp
stdvector_overload.cpp
stdvector.cpp
superlu_support.cpp Avoid blocking-size mismatch in unit tests calling Eigen's blas interface. 2014-04-25 11:04:02 +02:00
svd_common.h Fix overflow when checking SVD accuracy 2015-06-23 15:05:20 +02:00
svd_fill.h Make test matrices for eigensolver/selfadjoint even more tricky 2015-05-12 18:44:46 +02:00
swap.cpp Fix vectorization of swap for non trivial expressions 2015-03-31 20:16:02 +02:00
triangular.cpp Add unit test for bug #839. 2014-07-09 13:06:06 +02:00
umeyama.cpp Relaxed umeyama test. Problem was ill-posed if linear part was scaled with very small number. This should fix bug #744. 2014-02-17 13:48:00 +01:00
umfpack_support.cpp Avoid blocking-size mismatch in unit tests calling Eigen's blas interface. 2014-04-25 11:04:02 +02:00
unalignedassert.cpp Disable complex scalar types because the compiler might aggressively vectorize 2015-12-09 20:46:09 +01:00
unalignedcount.cpp Added support for AVX to Eigen. 2014-01-29 11:43:05 -08:00
upperbidiagonalization.cpp Make UpperBidiagonalization accept row-major matrices (bug #769) 2014-09-12 14:52:35 +01:00
vectorization_logic.cpp Fix unit tests wrt EIGEN_DEFAULT_TO_ROW_MAJOR 2015-12-11 10:06:28 +01:00
vectorwiseop.cpp bug #1158: PartialReduxExpr is a vector expression, and it thus must expose the LinearAccessBit flag 2016-01-28 13:16:30 +01:00
visitor.cpp Add regression unit test for prod.maxCoeff(i) 2015-11-18 23:29:07 +01:00
zerosized.cpp Implement boolean reductions for zero-sized objects 2013-11-13 16:47:02 +01:00