 Chip Kerchner
|
0753ebcbf0
|
Merge remote-tracking branch 'origin/master' into fixSignedZerosWithFastMath
|
2026-07-08 13:10:31 +00:00 |
|
|
Chip Kerchner
|
30f58a8411
|
Add signed-zero flags to fast-math - without it results can be wrong.
|
2026-07-08 12:50:30 +00:00 |
|
|
Florian Maurin
|
fbc0c78a6d
|
Core: Improve fixed-size intermediate matrix product performances
libeigen/eigen!2695
Closes #1791
|
2026-07-08 00:20:38 -07:00 |
|
|
Charles Schlosser
|
70034bc005
|
Reach-based sparse triangular solveInPlace (Gilbert–Peierls) for col-major L / Lᵀ
libeigen/eigen!2678
|
2026-07-07 16:52:33 -07:00 |
|
|
Kasper Nielsen
|
bc5f5f685a
|
SME: fix gebp packers for non-direct-GEMM consumers
libeigen/eigen!2675
|
2026-07-06 22:16:17 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
bb18f8ad25
|
Core: Fix dangling reference in non-const imag() of real-valued expressions
libeigen/eigen!2682
Closes #3096
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-07-06 11:47:50 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
b33822a2b6
|
Test: Fix block sparse identity guard for C++14
libeigen/eigen!2683
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-07-06 11:02:46 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
f7001a24dc
|
Cholesky: Fuse the 2x2-pivot trailing update in unblocked Bunch-Kaufman
libeigen/eigen!2679
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-07-05 20:18:24 -07:00 |
|
|
Florian Maurin
|
3923421454
|
Geometry: restrict transform matrix product overloads
libeigen/eigen!2645
Closes #1322
|
2026-07-04 17:07:14 -07:00 |
|
|
Florian Maurin
|
23815f4f62
|
Test: Fix platform-sensitive test expectations
libeigen/eigen!2674
|
2026-07-04 16:51:28 -07:00 |
|
|
Charles Schlosser
|
cd70e0c4a9
|
BlockSparseMatrix: add block-sparse data structure and operations
libeigen/eigen!2644
|
2026-07-02 22:20:12 +00:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
b4e88b90d9
|
packetmath: fix signed-integer-overflow UB in test reference helpers
libeigen/eigen!2671
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-30 16:32:22 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
3fccdcd589
|
IterativeLinearSolvers: fix flaky LSMR tolerances test
libeigen/eigen!2670
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-28 22:38:07 -07:00 |
|
|
Charles Schlosser
|
e943ebae60
|
Fix compiler warnings in cross-compile CI jobs
libeigen/eigen!2667
|
2026-06-28 00:49:44 +00:00 |
|
|
Florian Maurin
|
82ab9b71c9
|
GMRES: normalize convergence by preconditioned RHS
libeigen/eigen!2660
Closes #926
|
2026-06-27 11:29:56 -07:00 |
|
|
Charles Schlosser
|
e478052efd
|
Product: vectorize small fixed-size compound products via packet cascade
libeigen/eigen!2656
|
2026-06-27 10:26:30 +00:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
2e20f3533b
|
Add LSMR iterative least-squares solver
libeigen/eigen!2663
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-26 18:54:20 -07:00 |
|
|
Kasper Nielsen
|
65bf174fc2
|
SME: adapt the fp32 GEMM kernel to the runtime streaming vector length
libeigen/eigen!2639
|
2026-06-25 18:21:35 -07:00 |
|
|
Florian Maurin
|
d4940f8a2f
|
Core: preserve pointer result types in coefficient functors
libeigen/eigen!2646
Closes #1700
|
2026-06-25 18:11:51 -07:00 |
|
|
Florian Maurin
|
6b77b5b65b
|
Perf: x2 speedup compilation time for BDCSVD and JacobiSVD
libeigen/eigen!2655
|
2026-06-25 18:02:14 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
fdfece7c16
|
Vectorize reductions and norms on runtime-contiguous strided expressions
libeigen/eigen!2651
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-25 17:34:57 -07:00 |
|
|
Florian Maurin
|
94152044e8
|
Fix empty product blocking in triangular updates
libeigen/eigen!2659
Closes #3093
|
2026-06-25 01:29:13 +00:00 |
|
|
Florian Maurin
|
bf7d62172f
|
BDCSVD: fix accuracy issue by protecting secular corrections from reassociation
libeigen/eigen!2653
Closes #1588
|
2026-06-22 09:36:11 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
7bd98698c5
|
Eigenvalues: add inverse-iteration eigenvectors to TridiagonalEigenSolver
libeigen/eigen!2617
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-18 14:40:43 -07:00 |
|
|
Florian Maurin
|
dd3b764968
|
Test: fix some CI test failures
libeigen/eigen!2648
|
2026-06-18 13:33:35 -07:00 |
|
|
Florian Maurin
|
69ec633d01
|
BiCGSTABL: Improve convergence accuracy
libeigen/eigen!2649
|
2026-06-18 12:50:51 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
fc56f71c7e
|
Eigenvalues: add TridiagonalEigenSolver, SIMD + OpenMP spectral bisection for symmetric tridiagonal eigenvalues
libeigen/eigen!2615
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-12 15:33:48 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
f7a1c99acb
|
IterativeLinearSolvers: simplify and harden the iterative solvers (follow-up to !2606)
libeigen/eigen!2609
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-12 14:41:46 -07:00 |
|
|
Florian Maurin
|
c1a4cd5864
|
Default trivial special members where semantics are unchanged
libeigen/eigen!2618
|
2026-06-10 10:53:13 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
2c95e5508e
|
Test: relax vectorization_logic_half unrolling expectation broken on AVX-512
libeigen/eigen!2633
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-09 20:15:46 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
2a2b25c064
|
IterativeLinearSolvers: promote solvers from unsupported/IterativeSolvers
libeigen/eigen!2606
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-08 10:52:03 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
2259c86dc5
|
Cholesky: Add Bunch-Kaufman factorization for symmetric/Hermitian indefinite matrices
libeigen/eigen!2612
Closes #1268 and #2479
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-06-07 12:06:23 -07:00 |
|
|
Florian Maurin
|
35d56e4ede
|
AVX512: Use integer_sequence for TRSM
libeigen/eigen!2582
|
2026-06-06 23:16:53 -07:00 |
|
|
|
d3cfde9db1
|
Core: fix small fixed-size kernel regressions (#3083)
libeigen/eigen!2581
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
Co-authored-by: Rasmus Munk Larsen <rlarsen@nvidia.com>
|
2026-06-06 22:59:46 -07:00 |
|
|
Florian Maurin
|
eb82caec1c
|
CUDA: Fix self-adjoint rank update on CUDA devices
libeigen/eigen!2600
Closes #1470
|
2026-06-06 21:26:31 -07:00 |
|
|
Kasper Nielsen
|
6dae401d6a
|
ARM SME2 GEMM backend
libeigen/eigen!2128
|
2026-06-06 16:08:37 -07:00 |
|
|
|
76beb092c5
|
SparseCore: threaded SpMV — cached operator + extend stock OMP to ColMajor
libeigen/eigen!2557
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
Co-authored-by: Rasmus Munk Larsen <rlarsen@nvidia.com>
|
2026-06-06 14:14:42 -07:00 |
|
|
Florian Maurin
|
dbf8cb6151
|
CUDA: Fix numext min/max for CUDA custom scalars
libeigen/eigen!2599
Closes #1936
|
2026-06-05 10:57:53 -07:00 |
|
|
Florian Maurin
|
525976782c
|
SVD: Fix TensorRef with fixed-size Eigen scalars
libeigen/eigen!2601
Closes #1082
|
2026-06-04 20:16:51 -07:00 |
|
|
Florian Maurin
|
95fd0c8f3e
|
Fix: preserve NaN and infinity behavior in LU, QR, and norms
libeigen/eigen!2597
Closes #2076 and #1018
|
2026-06-03 14:56:07 -07:00 |
|
|
Florian Maurin
|
e043dafb62
|
Fix GCC fast-math complex packet arithmetic
libeigen/eigen!2590
Closes #1840 and #1839
|
2026-06-01 22:11:59 +00:00 |
|
|
Florian Maurin
|
ed45bac320
|
Perf: Avoid heap allocation for fixed-max decomposition temporaries
libeigen/eigen!2593
Closes #1477
|
2026-06-01 01:40:21 +00:00 |
|
|
Florian Maurin
|
fc0f148ab4
|
Fix: GCC 9.2 fast-math complex products
libeigen/eigen!2579
Closes #1839
|
2026-05-27 04:37:37 +00:00 |
|
|
Florian Maurin
|
039764e20a
|
Constexpr compile time branches
libeigen/eigen!2575
|
2026-05-22 13:18:49 -07:00 |
|
|
Florian Maurin
|
381cae56f8
|
SparseQR: look ahead before skipping weak pivots
libeigen/eigen!2537
Closes #1563
|
2026-05-21 14:02:33 -07:00 |
|
|
Florian Maurin
|
32a2be5f15
|
Fix: row-major least-squares diagonal preconditioner
libeigen/eigen!2576
|
2026-05-21 12:31:18 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
af2fd98144
|
SparseCore: honor Eigen::NoChange in SparseMatrix/SparseVector resize (#656)
libeigen/eigen!2569
Closes #656
Co-authored-by: Rasmus Munk Larsen <rlarsen@nvidia.com>
|
2026-05-18 21:36:29 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
8a079e1a9e
|
IterativeSolverBase: default-construct the dummy matrix
libeigen/eigen!2567
Closes #1704
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-05-18 21:01:03 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
226490124a
|
VectorwiseOp: fix aliasing in compound op with self-reducing rhs (#1731)
libeigen/eigen!2560
Closes #1731
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-05-18 19:10:27 -07:00 |
|
|
Rasmus Munk LarsenandRasmus Munk Larsen
|
edc053f1b7
|
SparseCore: add operator==/!= to SparseCompressedBase InnerIterator (#1192)
libeigen/eigen!2571
Closes #1192
Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
|
2026-05-18 22:16:43 +00:00 |
|