devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity
13,188 Commits 15 Branches 93 Tags
gpu-cg-interop
Commit Graph

13188 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Rasmus Munk Larsen
014f12f11a GPU: Add BLAS-1 ops, DeviceScalar, device-resident SpMV, and CG interop (5/5) 
Add the operator interface needed for GPU iterative solvers:

- BLAS Level-1 on DeviceMatrix: dot(), norm(), squaredNorm(), setZero(),
  noalias(), operator+=/-=/\*= dispatching to cuBLAS axpy/scal/dot/nrm2.
- DeviceScalar<Scalar>: device-resident scalar returned by reductions.
  Defers host sync until value is read (implicit conversion). Device-side
  division via NPP for real types.
- GpuContext: stream-borrowing constructor, setThreadLocal(), cublasLtHandle(),
  cusparseHandle().
- GEMM upgraded from cublasGemmEx to cublasLtMatmul with heuristic algorithm
  selection and plan caching.
- GpuSparseContext: GpuContext& constructor for same-stream execution,
  deviceView() returning DeviceSparseView with operator* for device-resident
  SpMV (d_y = d_A * d_x).
- geam expressions: d_C = d_A + alpha * d_B via cublasXgeam.
- GpuSVD::matrixV() convenience wrapper.

These additions make DeviceMatrix usable as a VectorType in Eigen algorithm
templates. Conjugate gradient is the motivating example and is tested against
CPU ConjugateGradient for correctness.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-09 20:19:59 -07:00
Rasmus Munk Larsen
43a95b62bb GPU: Add sparse solvers, FFT, and SpMV (cuDSS, cuFFT, cuSPARSE) 
Add GPU sparse direct solvers (Cholesky, LDL^T, LU) via cuDSS, 1D/2D FFT
via cuFFT with plan caching, and sparse matrix-vector/matrix multiply
(SpMV/SpMM) via cuSPARSE.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-09 19:11:49 -07:00
Rasmus Munk Larsen
8593c7f5a1 GPU: Add dense cuSOLVER solvers (QR, SVD, EigenSolver) 
Add QR (geqrf + ormqr + trsm), SVD (gesvd), and self-adjoint eigenvalue
decomposition (syevd) via cuSOLVER. All support host and DeviceMatrix input.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-09 19:11:34 -07:00
Rasmus Munk Larsen
58c44ef36d GPU: Add library dispatch module (DeviceMatrix, cuBLAS, cuSOLVER) 
Add Eigen/GPU module: A standalone GPU library dispatch layer where
DeviceMatrix<Scalar> operations map 1:1 to cuBLAS/cuSOLVER calls.
CPU and GPU solvers coexist in the same binary with compatible syntax.

Core infrastructure:
- DeviceMatrix<Scalar>: RAII dense column-major GPU memory wrapper with
  async host transfer (fromHost/toHost) and CUDA event-based cross-stream
  synchronization.
- GpuContext: Unified execution context owning a CUDA stream + cuBLAS
  handle + cuSOLVER handle. Thread-local default with explicit override
  via setThreadLocal(). Stream-borrowing constructor for integration.
- DeviceBuffer: Typed RAII device allocation with move semantics.

cuBLAS dispatch (expression syntax):
- GEMM: d_C = d_A.adjoint() * d_B (cublasXgemm)
- TRSM: d_X = d_A.triangularView<Lower>().solve(d_B) (cublasXtrsm)
- SYMM/HEMM: d_C = d_A.selfadjointView<Lower>() * d_B (cublasXsymm)
- SYRK/HERK: d_C = d_A * d_A.adjoint() (cublasXsyrk)

cuSOLVER dispatch:
- GpuLLT: Cached Cholesky factorization (cusolverDnXpotrf + Xpotrs)
- GpuLU: Cached LU factorization (cusolverDnXgetrf + Xgetrs)
- Solver chaining: auto x = d_A.llt().solve(d_B)
- Solver expressions with .device(ctx) for explicit stream control.

CI: Bump CUDA container to Ubuntu 22.04 (CMake 3.22), GCC 10->11,
Clang 12->14. Bump cmake_minimum_required to 3.17 for FindCUDAToolkit.

Tests: gpu_cublas.cpp, gpu_cusolver_llt.cpp, gpu_cusolver_lu.cpp,
gpu_device_matrix.cpp, gpu_library_example.cu
Benchmarks: bench_gpu_solvers.cpp, bench_gpu_chaining.cpp,
bench_gpu_batching.cpp
 2026-04-09 19:05:25 -07:00
Rasmus Munk Larsen
6a9405bf7a GPU: Raise CUDA/HIP minimum and remove legacy guards 
- Raise CUDA minimum from 9.0 to 11.4 (sm_70/Volta).
- Raise HIP minimum to GFX906 (Vega 20/MI50) / ROCm 5.6.
- Remove EIGEN_HAS_{CUDA,HIP,GPU}_FP16 guards — FP16 is always available
  on sm_70+ and GFX906+.
- Remove obsolete __HIP_ARCH_HAS_* preprocessor branches.
- C++14 cleanup: remove pre-C++14 workarounds in GPU code.
- Fix NVCC warnings (deprecated register keyword, unreachable code,
  tautological comparisons).
- Fix HIP test execution on gfx1151.
- Update CI configuration for new minimum versions.
 2026-04-09 15:21:39 -07:00
Rasmus Munk Larsen
e055e4e415 Add plog_core_double with fallback for AVX without AVX2 
libeigen/eigen!2407

Co-authored-by: Rasmus Munk Larsen <rlarsen@nvidia.com>
 2026-04-08 19:41:07 -07:00
Rasmus Munk Larsen
b1d2ce4c85 Revert "Speed up plog_double ~1.7x with fast integer range reduction" 
This reverts merge request !2385
 2026-04-08 13:03:48 -07:00
Rasmus Munk Larsen
ab70739c9c Speed up plog_double ~1.7x with fast integer range reduction 
libeigen/eigen!2385

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-07 21:48:25 -07:00
Rasmus Munk Larsen
e778b5d22b Switch ASAN/UBSAN smoketest pipelines to large runners 
libeigen/eigen!2405

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-07 21:37:58 -07:00
Rasmus Munk Larsen
def45c5e1e Improve psincos_double: faster polynomials + accurate range reduction 
libeigen/eigen!2389

Closes #3052

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-07 21:24:24 -07:00
Rasmus Munk Larsen
110530a4d8 Fix bugs and improve robustness of SelfAdjointEigenSolver, improve test coverage 
libeigen/eigen!2396

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-07 21:08:29 -07:00
Rasmus Munk Larsen
bde3a68bae Improve dense linear solver docs with practical guidance 
libeigen/eigen!2395

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-05 21:40:42 -07:00
Rasmus Munk Larsen
8eabfb5342 Vectorize BLAS level 1/2 routines with Eigen expressions 
libeigen/eigen!2404

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-05 18:53:11 -07:00
Rasmus Munk Larsen
4ad90a60f1 Replace blas/f2c with clean C++ implementations 
libeigen/eigen!2402

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-05 16:04:41 -07:00
Rasmus Munk Larsen
fe6ada10be Prevent nightly CI pipelines from being auto-cancelled 
libeigen/eigen!2390

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-04 11:52:15 -07:00
Alexander Grund
8179474225 CI: Add AVX512-FP16 build tests with GCC 13 
libeigen/eigen!1652

Co-authored-by: Alexander Grund <alexander.grund@tu-dresden.de>
 2026-04-04 11:32:31 -07:00
Florian Maurin
b57d860f3e Fix GCC maybe-uninitialized warning in InnerProduct 
libeigen/eigen!2386

Closes #3015
 2026-04-03 19:41:09 -07:00
Rasmus Munk Larsen
a3074053a6 Speed up pexp_double by ~15-17% 
libeigen/eigen!2388

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-03 17:09:11 -07:00
Rasmus Munk Larsen
a91913e961 Speed up plog_float by 1.6x with improved accuracy 
libeigen/eigen!2382

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-03 13:45:01 -07:00
Rasmus Munk Larsen
ebae0c7c10 ulp_accuracy: use dynamic work queue for thread load balancing 
libeigen/eigen!2383

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-02 22:40:03 -07:00
Charles Schlosser
5977635d64 fix singed integer overflow UB in integer_types and other trivial compiler warnings 
libeigen/eigen!2380
 2026-04-03 03:36:28 +00:00
Rasmus Munk Larsen
60df12437e Fix ulp_accuracy crashes in Release builds 
libeigen/eigen!2381

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-02 20:12:13 -07:00
Pavel Guzenfeld
e315a8cdd0 Inline IndexedViewMethods.inc into DenseBase.h 
libeigen/eigen!2330

Closes #2766
 2026-04-02 15:26:56 -07:00
Rasmus Munk Larsen
8ec68856a6 Fix basicstuff_8 casting test failure on loongarch64 
libeigen/eigen!2379

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-02 14:14:54 -07:00
Rasmus Munk Larsen
61a8662876 Improve log1p accuracy and speed with direct range reduction 
libeigen/eigen!2378

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-02 11:29:25 -07:00
Rasmus Munk Larsen
d31a73437f Vectorize asinh and acosh for float and double 
libeigen/eigen!2376

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-01 21:46:36 -07:00
Rasmus Munk Larsen
9513d3878e Vectorize sinh, cosh, and log10 
libeigen/eigen!2368

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-01 20:41:18 -07:00
Rasmus Munk Larsen
30e669cfe1 Tensor module: const-correctness and constexpr improvements 
libeigen/eigen!2239

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-01 17:49:56 -07:00
Rasmus Munk Larsen
64885cc6a3 Fix remaining MSVC warnings in Windows CI (C4804, C4244, C4146, C4305) 
libeigen/eigen!2374

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-01 17:20:31 -07:00
Rasmus Munk Larsen
6a07970d7d CI: split NVHPC build and make fallback parallelism configurable 
libeigen/eigen!2372

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-01 16:43:33 -07:00
Rasmus Munk Larsen
4be66f2830 CI: fail test jobs when no tests are found (--no-tests=error) 
libeigen/eigen!2373

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-04-01 12:50:54 -07:00
Rasmus Munk Larsen
1df89cbc21 Right-size CI runners to reduce waste and shuffle build order to avoid OOM 
libeigen/eigen!2367

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-31 19:10:34 -07:00
Rasmus Munk Larsen
b54640df19 Fix NVHPC warnings in Visitor.h and Memory.h 
libeigen/eigen!2370

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-31 15:09:37 -07:00
Rasmus Munk Larsen
7fcbed7acb Fill packet math coverage gaps across multiple architectures 
libeigen/eigen!2237

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-31 14:52:32 -07:00
Rasmus Munk Larsen
80ab2898e2 CI: install libclang-rt-14-dev for sanitizer smoketest 
libeigen/eigen!2369

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-31 00:16:18 -07:00
Rasmus Munk Larsen
798d7f2bec CI: drop Clang-6, bump base image to Ubuntu 24.04 and Clang 12 to 14 
libeigen/eigen!2366

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-30 22:00:17 -07:00
Rasmus Munk Larsen
1ade3636b9 Fix BDCSVD bidiagonal hard-case failures on ARM with GCC 
libeigen/eigen!2365

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-30 20:17:37 -07:00
Rasmus Munk Larsen
801a9ee690 Fix ~1,460 MSVC warnings from generic code instantiated with bool 
libeigen/eigen!2364

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 21:05:49 -07:00
Rasmus Munk Larsen
806c7b6590 CI: fix Windows build cache key containing invalid path characters 
libeigen/eigen!2362

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 19:57:45 -07:00
Rasmus Munk Larsen
2776ba55eb Update slicing tutorial docs to reflect Eigen::placeholders namespace 
libeigen/eigen!2360

Closes #3064

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 18:51:02 -07:00
Rasmus Munk Larsen
09581fda38 Modernize tensor contraction code: bug fixes, dead code removal, and cleanup 
libeigen/eigen!2248

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 18:03:06 -07:00
Rasmus Munk Larsen
732ebc8cc2 Modernize evaluator files 
libeigen/eigen!2245

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 17:40:39 -07:00
Rasmus Munk Larsen
255f522e2e Fix bugs, docs, and structure in unsupported/ public headers 
libeigen/eigen!2254

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 17:06:40 -07:00
Pavel Guzenfeld
bd276fbb28 Map .inc files to C++ in Doxygen extension mapping 
libeigen/eigen!2338
 2026-03-29 16:48:13 -07:00
Rasmus Munk Larsen
c8633ceeea Clean up top-level Eigen headers 
libeigen/eigen!2252

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 16:28:09 -07:00
Rasmus Munk Larsen
409296d91d Add nightly benchmark regression detection pipeline 
libeigen/eigen!2349

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 16:03:56 -07:00
Pavel Guzenfeld
753a6ac5b3 Fix private shadowing of protected base members in iterative solvers 
libeigen/eigen!2357

Closes #1859
 2026-03-29 15:40:48 -07:00
Rasmus Munk Larsen
9fe2f03fa4 Revert "Lower BDCSVD crossover threshold from 16 to 8" 
This reverts merge request !2358
 2026-03-29 15:25:09 -07:00
Rasmus Munk Larsen
12fe90db8b Lower BDCSVD crossover threshold from 16 to 8 
libeigen/eigen!2358

Co-authored-by: Rasmus Munk Larsen <rmlarsen@gmail.com>
 2026-03-29 14:33:22 -07:00
Pavel Guzenfeld
b7f6aed1b9 Fix dangling reference in IndexedView with expression indices 
libeigen/eigen!2335

Closes #1943
 2026-03-29 09:39:13 -07:00