)]}'
{
  "log": [
    {
      "commit": "aca8f01014f2eb835a004d568e5591f88a5afadf",
      "tree": "24954225c6978d6c12cfdd02a015721d7e38db77",
      "parents": [
        "c536fd6cba19b925cb6a9d9856b9356b3b91a747"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Wed Oct 21 12:40:38 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Oct 21 12:41:47 2026 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 1004151930\nChange-Id: I6a0f64f1a5308c65c81502fb004cebbe70cda743\n"
    },
    {
      "commit": "c536fd6cba19b925cb6a9d9856b9356b3b91a747",
      "tree": "da55d58a5614999b1e0ca5f0fd9c2e043123cba6",
      "parents": [
        "3ac4a63b8f65bfc146e700ace498d9546ed46952"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Mon Oct 12 10:25:13 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Oct 12 10:26:06 2026 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 998003033\nChange-Id: I63251a7185b868b9c25ba36194bf33195b616541\n"
    },
    {
      "commit": "3ac4a63b8f65bfc146e700ace498d9546ed46952",
      "tree": "acd33f594b651734ff468173c020fbf5dfc25fbc",
      "parents": [
        "8e8874cddbf373521cea9a7be8422846c1bac2d6"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Wed Oct 07 17:16:46 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Oct 07 17:18:04 2026 -0700"
      },
      "message": "Update Eigen to commit:b1f8b0c3b8db9c3711349144acf951226700ddb4\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nb1f8b0c3b - Core: Support flexible indexers in `VectorwiseOp::replicate`\n8f1a18ff7 - Core: Gate product_packet_cascade on MightVectorize and mark assignPacket device-callable\n630f81dbf - RVV: Enable generic packet sinh, cosh, asinh, acosh, atan, atanh, log10, log1p, expm1, pow, cbrt, rsqrt and erfc\n5b5b44a71 - LU: Unroll PartialPivLU for fixed sizes up to 12\n32a4f127f - Eigenvalues: Defer far-column reflector updates in the Francis QR step\nab8a55665 - Householder: Apply two- and three-element reflectors from the right in one pass\n97898f371 - AltiVec: Fix out-of-bounds tail accesses in VSX and MMA GEMM kernels\n0013bcb18 - Geometry: Preserve single-rounding pivot and trace order in Quaternion(Matrix3)\n07d4528c8 - MSVC: Fix the ILP64 BLAS DLL and the stencil test\n2774a036e - StructuredMatrices: Identity and nested factors for KroneckerOperator\n869eb9850 - Core: Pin the TRSM padded tail\u0027s view of the right-hand side to column-major\nb6225b7dc - Core: Finish partial GEMV packets in one pass with masked segments\nc541f8f43 - Core: Support flexible indexers in `DenseBase::replicate`\n7f03d8469 - NEON: Classify floats with absolute comparisons and reduce allFinite() through masks\nbc99fddc6 - Cholesky: Unblocked LLT up to 48 columns and 16-column minimum blocks\ne3755c9c7 - Core: Scalar tails for assignments bounded to a few packets\nff16ae80a - Added specific value tests (x2) and random number generated tests for autodiff...\n03d21656a - StructuredMatrices: Reuse Cauchy reciprocal columns across right-hand sides\n477435767 - SME: Fold a real alpha into the complex tile-pair store\nc8725af72 - StructuredMatrices: Drop extreme-scale hardening from DPR1EigenSolver\n3346f1a2b - StructuredMatrices: Drop extreme-scale hardening from DiagonalPlusLowRank\na29e46e2e - StructuredMatrices: Solve Kronecker least squares by per-factor COD\n036a81d7c - NEON: Clean up complex arithmetic and fix GCC ARM32 float compares\n1c7d00c27 - Core: Fix out-of-bounds B-panel lookahead loads in AVX512 GEMM kernel\n4be755d45 - Core: Decouple the fixed-size product threshold and retune it for SME\ndee1af8bb - Householder: Restore AutoDiffScalar support in QR and JacobiSVD\nf7a2dd7a9 - FindCoeff: Require HasCmp for the packet path of minCoeff/maxCoeff\na82afb7c7 - Geometry: Fix AngleAxis::fromRotationMatrix accuracy and axis sign near pi\n7194105cc - StructuredMatrices: Drop extreme-scale hardening from the FFT operators\nd11bb9121 - NEON: Optimize stride-two float gathers\n7ac31bc91 - EulerAngles: Test the constructor from a data pointer\n038961d50 - NumericalDiff: Implement Fornberg finite difference stencil algorithm\nb4befd43f - SVD: add PreconditionSquareMatrix option.\n982e195f9 - StructuredMatrices: Batch KroneckerOperator products and solves, drop extreme-scale hardening\n221aa496a - StructuredMatrices: Drop extreme-scale hardening from Vandermonde and Cauchy\n963fffeb0 - NVHPC: Fix the nvc++ build\n19090299a - Core: Use k-masked packet segments on AVX-512 and predicated segments on SVE\n062290ffb - Tests: Guard pabs and pmadd/pmsub/pnmadd with packet traits in packetmath\n210e0a609 - Core: Return the cached outer stride from direct-access Block\n3cf994918 - CI: Build the AVX512-FP16 tests with clang-19\nbbefaffa6 - Tests: Compile split test parts with only the functions they call\n8c653283e - SME: Keep pack_direct in streaming mode under GCC\nc3123dc95 - BDCSVD: Cover the bidiagonal entry point in the switch-size test\nca0565a17 - CI: Pass the SME -march through CMAKE_CXX_FLAGS\ne4505f831 - SME: Move the streaming primitives into arch/SME/PacketMath.h\n1bd7cb0d1 - CI: Add an nvhpc-tests label for the NVHPC jobs\n48b196e67 - Core: Drop constexpr from the noinline complex sqrt fallback\n541f05781 - BDCSVD: Reallocate after setSwitchSize\ndae92cd3a - BLAS: Add ILP64 (64-bit integer) build and test support\nb79499c6f - AltiVec: Add pcmp_lt for Packet2l, fix the POWER7 build, and name the POWER8 check\n6accef0ca - Core: Default-initialize storage in the expression constructors\n0011719b1 - CMake: Disable if-conversion in the AVX512-FP16 tests on affected GCC\n47bf0b657 - Tensor: Reject u \u003d\u003d 0 in the normal random generator\n859c035ba - Tests: Compile cheap split-test parts together\nd22c0df89 - Tests: Build redux_bounded_compile_vectorized with the job\u0027s flags\nd9ec2c9b9 - GPU: Copy views and adopt from a Context in the dense solvers\u0027 rvalue overloads\nae47f25fb - CI: Isolate Windows sccache daemon port by CI_JOB_ID and ensure script teardown\n3d836af2d - Tests: Stop special_functions from failing on cancellation\nff16afbb1 - GPU: Give cuBLAS aligned copies of complex host scalars\ncbda2af3e - Core: Fix vector shape validation when automatic resizing is disabled\n077d29983 - Core: Clamp the exponent before vscalef in the AVX-512 pldexp\n9a97f13f6 - CI: Stop stale sccache servers on Windows runners\ncd1a734f2 - Core: Preserve bfloat16 packet sign under FTZ and DAZ\n93933c899 - Core: Split pldexp\u0027s scale into three factors and round it once\n356cdfd6e - Core: Preserve float and double sign under DAZ/FTZ\na63497bea - CI: Run the GPU jobs and the cuBLAS/cuSOLVER tests for GPU-module changes\n493c3a3a8 - GPU: Check CUDA and library calls in release builds\n8e08dc1f5 - Jacobi: Select bounded ratio operands before division\n851458df8 - Core: Compute subnormal integer powers exactly on ARMv7 NEON\n38c71b0cf - GPU: Reduce into a caller\u0027s DeviceScalar, and compute norm() from the dot product\na55314c6b - GPU: Stop solver result getters from waiting on unrelated GPU work\nab89b1ab1 - GPU: Fix complex device-scalar axpy/scal and empty-operand GEMM on DeviceMatrix\n367032aa0 - Core: Add adjoint() to diagonal, permutation and transposition matrices\nd17bdf5e9 - Tests: Index the FFT reference DFT\u0027s twiddles instead of calling exp\na44124c5a - Core: Preserve reduction bounds through unary expressions\n86e6ae8e2 - SME: one disjoint part of the result per SME unit, minimum task size per scalar type\na1afbf4dd - Core: Vectorize isZero() and isApproxToConstant() checks\n04bfe2287 - Core: Simplify packet traversal for strided coefficient assignments\n1bd8e6f4d - SME: thread cap at the unit count, in-place LHS, ZA packers, kernels for narrow and tiny results\n4df51993a - StructuredMatrices: Make DPR1 deflation exact for equal poles and option-independent\nef5a9d5c4 - Core: Refactor approximate comparisons and fix false positives\nd142522b7 - Core: Fix the TRSM padded solve\u0027s link error in unoptimized C++14 builds\nd1b87b4cd - Core: Solve whole TRSM diagonal blocks with the shared packet kernel\n651678320 - Core: Respect resizing restrictions and streamline small product evaluation\nff1458835 - CI: Use \u0027-\u0027 prefix for MSVC compiler flags so sccache caches Windows builds\n6d293af5e - Core: Add lazy sums of dense, triangular and diagonal operands with diagonal and permutation matrices\n563e4f40a - Compute integer powers by double-word repeated squaring, real and complex\ned56f0c1d - Eigenvalues: Keep subnormal couplings under FTZ on MSVC and fix the test build\n92b9ed6b0 - StructuredMatrices: Apply alpha to x in non-finite sparse Kronecker products\na80a9cd50 - CI: Fix the GPU README docs failure and add a docs-build MR label\ndd16a1d7f - CI: Run the lint checks in one job on a prebuilt lint image\n38295af54 - Core: Fix RealView coefficient references and IndexedView/RealView data() constness\n0ec724769 - CI: Run clang-tidy on a prebuilt, pruned Ubuntu image\naf8d01b89 - Tests: Use explicit expected values for complex rsqrt corner cases\nac940d63c - CI: Build the Windows CUDA tests through ccache, not sccache\n9c7cebf31 - Core: Keep pldexp\u0027s scale factors from being reassociated under fast math\nac63c7844 - CI: Pass failtest compiles through a missing launcher\nde177ddbb - CI: Provision sccache on Linux test runners for failtest jobs\na7a0e35ee - Tests: Bound the scaled_permutation sum check by the terms, not the result\n5fac38847 - Householder: Multiply a HouseholderSequence by a row vector\n8746be0f2 - Core: Silence GCC 10 loop warnings in slice-vectorized reductions\n91c8a4d8b - Householder: Apply a one-row reflector from the left\n2dce125ec - CI: Fix sccache GCS credential JSON key and mask Windows token dump\n53c33b64b - Core: Fix strides reported by vector IndexedView and strided RealView\n9d194e7ad - Householder: Multiply HouseholderSequence by diagonal, triangular and permutation matrices\n7681b6b7a - Core: Add permutation products with triangular and self-adjoint views\n8a070f3e1 - CI: Add sccache with GCS remote backend with ccache fallback.\n4b0b97830 - Core: Accelerate real vector operations with SME2\n96aad0baa - Core: Fix scalar handling in mixed and self-adjoint products\n59c28e04c - GPU: run Eigen\u0027s ConjugateGradient on the GPU types, with solveWithGuessInPlace\n19292c62a - Docs: Fix the unresolved BSR section link in the GPU README\na1a19e6b2 - Core, SVD, StructuredMatrices: Scale all-subnormal data exactly on flush-to-zero hardware\n4cba415f3 - CI: Pre-bake curl, ca-certificates, and python3 into smoke build Dockerfiles\n087757ad5 - Docs: Use Doxygen inline-math delimiters in ScaledPermutationMatrix.h\ne89236d8f - GPU: Add cuSPARSE BSR products for BlockSparseMatrix\nd30a9e6d3 - Core: Fix reshape traversal order and unsafe packet assignments\n6f4f522f5 - EventCount: Fix signal loss in CancelWait\nb346103bd - Docs: Treat proprietary software as a black box\neef236cb4 - SME: numext::round_down for the non-streaming panel peels\n776f395d2 - Core: Add ScaledPermutationMatrix, the product of a permutation and a diagonal matrix\nec8593a7d - Fix deprecated enum-enum conversion warning\ncba8148e3 - Core: Add products of two triangular or self-adjoint views\n45e88a56d - Core: Harden the structured-product dispatch for shapes without a kernel\n34e0e0793 - Eigenvalues: Reorthogonalize globally unresolved block shifts\nf26a444af - Core: Fix implicit Matrix conversions in std::variant\nd5a19d2ca - Fix GCC array-bounds warnings in Core, Sparse and Tensor\nab4f7316d - CI: Auto-retry infrastructure drops on runner loss or preemption\n14e4b5405 - Jacobi: Use power-of-two scaling and preserve subnormal real Givens rotations\nd32e9702b - SME: NEON path for small blocks, 64-byte packed panels, generic crossovers\n647bc59a9 - Core: Fix scalar rewrites of structured products\n8c6b5162b - Tests: Fix redux_12 and structured_kronecker_12 on 32-bit ARM\n82a7cb22a - Agents: Present benchmark measurements as tables in merge request descriptions\n47a78413d - Eigen: Standardize LAPACK helper routines into native C++ implementations\n399d29312 - StructuredMatrices: Accept sparse factors in KroneckerOperator\neddc25291 - SparseLU: Divide by subnormal pivots instead of forming their reciprocal\n1b5286a2c - MSVC: Fix complex-component evaluator and tensor block-view builds\nb5c3e3da6 - Core: Use four accumulators in the vectorized reductions\nc7c04d737 - Core: Stabilize scalar complex square roots\n7051b1f9a - Tests: Materialize the random vector before makeHouseholder in householder_essential_expressions\n46a2d0c56 - Eigenvalues: Share and harden direct self-adjoint solver scaling\n3ca5ab304 - CI: Allow runner-level CCACHE_* overrides via EIGEN_CI_CCACHE_* defaults\ne3c8cb150 - Core: Respect packet access in in-place transpose\n1eef8230d - Core: Fix reverse row and column iterator indexing\n04cc35554 - Core: Fix packet twoprod residuals and correct its error bound\na076632bf - Performance: Avoid temporaries in non-aliasing product assignments\nc1a957f96 - Core: Share inner product dispatch and retain a tiny-vector path\n59d9e069a - Tensor: Fix the GPU warnings that overflow the build log\n0039ee376 - CI: Run 128-bit SVE tests natively on dedicated ARM runner\n33fd96d09 - Docs: Re-attach the Transform::rotate Doxygen block and fix the GMRES \\param name\n93a193c58 - Core: Vectorize triangular solves across right-hand sides\ne1c311414 - Core: Enable linear counting and simplify visitor unrolling\ne7e5178d1 - Core: Exclude unreachable wide reductions for bounded expressions\n2396b9dc2 - Core: Keep complex and class operands of the PowerPC GCC barrier in memory\n02f77fd76 - CI: Skip the ccache round trip when an affected build compiles nothing\nff285117c - Tests: Check backward error in dontalign QR solves\n94f0b704f - Core: Reject Boolean subtraction and negation\n8f8d4ed4c - Geometry: Support mixed-scalar operations in Transform::translate, pretranslate, and toRotationMatrix\ndd755f75e - Householder: Suppress false-positive -Warray-bounds warnings in BlockHouseholder\nbbcff8c67 - Eigenvalues: Avoid double promotion in scaling test\n2800561ee - Core: Simplify and vectorize permutation operations\n2a57319e6 - CMake: Compile GPU tests through CUDA and HIP language support\n60c5525bb - Eigenvalues: Reuse QZ workspaces during compute\n3f21c28a5 - CI: Parallelize the failtest suite and run native x86-64 tests on the large runner\n154a40bd4 - Tensor: Avoid copying strided block operands\n2002cae43 - Core: Avoid -Wfloat-equal warnings in SafeScaling for long double\n2e7712e63 - CI: Use saas-linux-medium-amd64 for smoke build jobs\nda8cda1da - Core: Vectorize complex component reductions\na14735e7e - Core: Add move support for DiagonalMatrix\n4990b8359 - Eigenvalues: Break stalled ComplexQZ iterations with exceptional shifts\n20b41e40b - QR: Document transposed solve limits and condition wide tests\n99f1cccfd - Tests: Bound erfc comparisons in the subnormal range\ncd38c5ddc - Core/QR: Vectorize complex partial norms and simplify QR initialization\n6a0625b84 - Sparse: Share packet scatter updates in Cholesky and LU\n1373c7a36 - CI: Build only the affected selection in the SVE affected jobs\n2298fbd89 - CI: Raise the SME full build timeout to 3h\n20f5a26ba - Core: Remove misc directory and simplify LU expressions\n15f227178 - Eigenvalues: Avoid cache-conflicting strides in RealSchur workspaces\n74601a9e3 - CI: Log machine configuration before builds and tests\n4a6bf607a - Eigenvalues: Use safe scaling in tridiagonal eigensolvers\ncaac397f4 - Tensor: Enable double outer reductions with native sum atomics\n27b0c620e - Tests: Allow for MSVC expm1 reference error\n8facea8ba - SVD: Pack active secular terms for contiguous evaluation\n6e86f64a0 - Core: Consolidate structured matrix norm utilities\n5ad7d1788 - SVD: Reuse the saved row across blocked Jacobi rotations\nec5876775 - SVD: Reuse operands while constructing singular vectors\na4d943d1f - CI: Cross-compile the aarch64 builds on amd64\n219e6c5a3 - HIP: Build tests for the requested GPU architectures\n0f96e0f16 - Tests: Bound SVD normal-equation checks by backward error\n7ec23fec7 - Core: Fix diagonal strides and subnormal identity scaling\n68af15188 - Benchmarks: Add GPU kernel timing and validated baselines\n50c7c77f3 - Tensor: Query attributes for the stream\u0027s owning GPU\n4eaf42d05 - Docs: Ground performance changes in cost models and evidence\nddf4a5479 - Eigenvalues: Use safe scaling in the iterative self-adjoint solver\n3f4260bfe - Eigenvalues: Fix inconsistent nonfinite input status in in-place eigensolvers\n5df2459f1 - Polynomials: Refine PolynomialSolver roots with Ehrlich-Aberth iterations\n99baace37 - Eigenvalues: Use power-of-two scaling in Schur decompositions\n6a026803b - CI: Cap parallel compiles in the aarch64 gcc-10 affected build\ne5b838361 - Geometry: Work around an LLVM ARM load/store miscompile in vectorwiseop\n37e72f8df - Tests: Remove duplicate and near-duplicate test coverage\n3bda1a485 - Warings: Fix some compiler warnings\n133b7aa9d - StructuredMatrices: Vectorize scaling and reuse spectral work\nf0a977a96 - Tests: Work around an LLVM ARM load/store miscompile in vectorwiseop\n0b9759a73 - Eigenvalues: Support inplace decomposition through Ref\u003c\u003e\nc0d6e9944 - GPU: Fix packet capabilities and simplify half operations\n7a6afd466 - Geometry: Restore 4D vector assignment delegation in AngleAxis::operator\u003d\nb17fadee5 - Core: Avoid -Wfloat-equal warnings in ceil_power_of_two for long double\ncae72577d - SparseCore: Silence unused variable warning under NDEBUG in triangular_solve_over_reach_iter\n2d148dc46 - Tests: Add device packet-math coverage shared with the host suite\n688bc7a5c - Core: Fix scalar complex GEMM accumulation on RVV\n\nPiperOrigin-RevId: 995455097\nChange-Id: Idbbc059f9c9e5e999ec6c62d97ec2a5392c935d5\n"
    },
    {
      "commit": "8e8874cddbf373521cea9a7be8422846c1bac2d6",
      "tree": "40558f28f21a46c19302d3e2155bdd20d47c1fb0",
      "parents": [
        "c533110be8303e7887b82e18a5176a3b0d0b0b63"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Oct 01 09:16:02 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Oct 01 09:16:48 2026 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 991719321\nChange-Id: I93434ec095d2716a5ecaf520926f0394afc5a031\n"
    },
    {
      "commit": "c533110be8303e7887b82e18a5176a3b0d0b0b63",
      "tree": "5781cfe0836a675fe70ed51936a21efd985b08d6",
      "parents": [
        "aecd9b8ceaf389fa43c981de9990a391ce42bafb"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Wed Sep 30 13:28:55 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Sep 30 13:30:03 2026 -0700"
      },
      "message": "Update Eigen to commit:087757ad50159b5bd86ab73e4df34e6b4ea74258\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n087757ad5 - Docs: Use Doxygen inline-math delimiters in ScaledPermutationMatrix.h\ne89236d8f - GPU: Add cuSPARSE BSR products for BlockSparseMatrix\nd30a9e6d3 - Core: Fix reshape traversal order and unsafe packet assignments\n6f4f522f5 - EventCount: Fix signal loss in CancelWait\nb346103bd - Docs: Treat proprietary software as a black box\neef236cb4 - SME: numext::round_down for the non-streaming panel peels\n776f395d2 - Core: Add ScaledPermutationMatrix, the product of a permutation and a diagonal matrix\nec8593a7d - Fix deprecated enum-enum conversion warning\ncba8148e3 - Core: Add products of two triangular or self-adjoint views\n45e88a56d - Core: Harden the structured-product dispatch for shapes without a kernel\n34e0e0793 - Eigenvalues: Reorthogonalize globally unresolved block shifts\nf26a444af - Core: Fix implicit Matrix conversions in std::variant\nd5a19d2ca - Fix GCC array-bounds warnings in Core, Sparse and Tensor\nab4f7316d - CI: Auto-retry infrastructure drops on runner loss or preemption\n14e4b5405 - Jacobi: Use power-of-two scaling and preserve subnormal real Givens rotations\nd32e9702b - SME: NEON path for small blocks, 64-byte packed panels, generic crossovers\n647bc59a9 - Core: Fix scalar rewrites of structured products\n8c6b5162b - Tests: Fix redux_12 and structured_kronecker_12 on 32-bit ARM\n82a7cb22a - Agents: Present benchmark measurements as tables in merge request descriptions\n47a78413d - Eigen: Standardize LAPACK helper routines into native C++ implementations\n399d29312 - StructuredMatrices: Accept sparse factors in KroneckerOperator\neddc25291 - SparseLU: Divide by subnormal pivots instead of forming their reciprocal\n1b5286a2c - MSVC: Fix complex-component evaluator and tensor block-view builds\nb5c3e3da6 - Core: Use four accumulators in the vectorized reductions\nc7c04d737 - Core: Stabilize scalar complex square roots\n7051b1f9a - Tests: Materialize the random vector before makeHouseholder in householder_essential_expressions\n46a2d0c56 - Eigenvalues: Share and harden direct self-adjoint solver scaling\n3ca5ab304 - CI: Allow runner-level CCACHE_* overrides via EIGEN_CI_CCACHE_* defaults\ne3c8cb150 - Core: Respect packet access in in-place transpose\n1eef8230d - Core: Fix reverse row and column iterator indexing\n04cc35554 - Core: Fix packet twoprod residuals and correct its error bound\na076632bf - Performance: Avoid temporaries in non-aliasing product assignments\nc1a957f96 - Core: Share inner product dispatch and retain a tiny-vector path\n59d9e069a - Tensor: Fix the GPU warnings that overflow the build log\n0039ee376 - CI: Run 128-bit SVE tests natively on dedicated ARM runner\n33fd96d09 - Docs: Re-attach the Transform::rotate Doxygen block and fix the GMRES \\param name\n93a193c58 - Core: Vectorize triangular solves across right-hand sides\ne1c311414 - Core: Enable linear counting and simplify visitor unrolling\ne7e5178d1 - Core: Exclude unreachable wide reductions for bounded expressions\n2396b9dc2 - Core: Keep complex and class operands of the PowerPC GCC barrier in memory\n02f77fd76 - CI: Skip the ccache round trip when an affected build compiles nothing\nff285117c - Tests: Check backward error in dontalign QR solves\n94f0b704f - Core: Reject Boolean subtraction and negation\n8f8d4ed4c - Geometry: Support mixed-scalar operations in Transform::translate, pretranslate, and toRotationMatrix\ndd755f75e - Householder: Suppress false-positive -Warray-bounds warnings in BlockHouseholder\nbbcff8c67 - Eigenvalues: Avoid double promotion in scaling test\n2800561ee - Core: Simplify and vectorize permutation operations\n2a57319e6 - CMake: Compile GPU tests through CUDA and HIP language support\n60c5525bb - Eigenvalues: Reuse QZ workspaces during compute\n3f21c28a5 - CI: Parallelize the failtest suite and run native x86-64 tests on the large runner\n154a40bd4 - Tensor: Avoid copying strided block operands\n2002cae43 - Core: Avoid -Wfloat-equal warnings in SafeScaling for long double\n2e7712e63 - CI: Use saas-linux-medium-amd64 for smoke build jobs\nda8cda1da - Core: Vectorize complex component reductions\na14735e7e - Core: Add move support for DiagonalMatrix\n4990b8359 - Eigenvalues: Break stalled ComplexQZ iterations with exceptional shifts\n20b41e40b - QR: Document transposed solve limits and condition wide tests\n99f1cccfd - Tests: Bound erfc comparisons in the subnormal range\ncd38c5ddc - Core/QR: Vectorize complex partial norms and simplify QR initialization\n6a0625b84 - Sparse: Share packet scatter updates in Cholesky and LU\n1373c7a36 - CI: Build only the affected selection in the SVE affected jobs\n2298fbd89 - CI: Raise the SME full build timeout to 3h\n20f5a26ba - Core: Remove misc directory and simplify LU expressions\n15f227178 - Eigenvalues: Avoid cache-conflicting strides in RealSchur workspaces\n74601a9e3 - CI: Log machine configuration before builds and tests\n4a6bf607a - Eigenvalues: Use safe scaling in tridiagonal eigensolvers\ncaac397f4 - Tensor: Enable double outer reductions with native sum atomics\n27b0c620e - Tests: Allow for MSVC expm1 reference error\n8facea8ba - SVD: Pack active secular terms for contiguous evaluation\n6e86f64a0 - Core: Consolidate structured matrix norm utilities\n5ad7d1788 - SVD: Reuse the saved row across blocked Jacobi rotations\nec5876775 - SVD: Reuse operands while constructing singular vectors\na4d943d1f - CI: Cross-compile the aarch64 builds on amd64\n219e6c5a3 - HIP: Build tests for the requested GPU architectures\n0f96e0f16 - Tests: Bound SVD normal-equation checks by backward error\n7ec23fec7 - Core: Fix diagonal strides and subnormal identity scaling\n68af15188 - Benchmarks: Add GPU kernel timing and validated baselines\n50c7c77f3 - Tensor: Query attributes for the stream\u0027s owning GPU\n4eaf42d05 - Docs: Ground performance changes in cost models and evidence\nddf4a5479 - Eigenvalues: Use safe scaling in the iterative self-adjoint solver\n3f4260bfe - Eigenvalues: Fix inconsistent nonfinite input status in in-place eigensolvers\n5df2459f1 - Polynomials: Refine PolynomialSolver roots with Ehrlich-Aberth iterations\n99baace37 - Eigenvalues: Use power-of-two scaling in Schur decompositions\n6a026803b - CI: Cap parallel compiles in the aarch64 gcc-10 affected build\ne5b838361 - Geometry: Work around an LLVM ARM load/store miscompile in vectorwiseop\n37e72f8df - Tests: Remove duplicate and near-duplicate test coverage\n3bda1a485 - Warings: Fix some compiler warnings\n133b7aa9d - StructuredMatrices: Vectorize scaling and reuse spectral work\nf0a977a96 - Tests: Work around an LLVM ARM load/store miscompile in vectorwiseop\n0b9759a73 - Eigenvalues: Support inplace decomposition through Ref\u003c\u003e\nc0d6e9944 - GPU: Fix packet capabilities and simplify half operations\n7a6afd466 - Geometry: Restore 4D vector assignment delegation in AngleAxis::operator\u003d\nb17fadee5 - Core: Avoid -Wfloat-equal warnings in ceil_power_of_two for long double\ncae72577d - SparseCore: Silence unused variable warning under NDEBUG in triangular_solve_over_reach_iter\n2d148dc46 - Tests: Add device packet-math coverage shared with the host suite\n688bc7a5c - Core: Fix scalar complex GEMM accumulation on RVV\n\nPiperOrigin-RevId: 991186382\nChange-Id: I53fe28eda1ccec3a60bd9c62acb6b54d72c363e6\n"
    },
    {
      "commit": "aecd9b8ceaf389fa43c981de9990a391ce42bafb",
      "tree": "c0da1ef1f01bfac2de9e34e6d5dffcb67332b1c3",
      "parents": [
        "746fe26e4c097835f874a7f43ccd94d7ba6d5bd9"
      ],
      "author": {
        "name": "Junwhan Ahn",
        "email": "junwhan@google.com",
        "time": "Thu Sep 24 10:07:45 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Sep 24 10:08:25 2026 -0700"
      },
      "message": "Fix a signal loss bug in `Eigen::EventCount::CancelWait` that leads to task stranding and deadlocks in thread pools under concurrent scheduling.\n\n### Deadlock Scenario\n`EventCount` maintains a packed 64-bit atomic state containing:\n- `waiters`: count of threads preparing to wait (between `Prewait` and `CommitWait`/`CancelWait`).\n- `signals`: count of pending wakeup notifications.\n- `stack`: stack of parked threads sleeping on condition variables in `CommitWait`.\n\nThe deadlock occurs via the following race:\n1. **Worker A parks**: Calls `Prewait()` then `CommitWait(\u0026w_A)`. Worker A is added to `stack` and blocks (`waiters \u003d 0, signals \u003d 0, stack \u003d [A]`).\n2. **Worker B prepares to park**: Calls `Prewait()`. State becomes `waiters \u003d 1, signals \u003d 0, stack \u003d [A]`.\n3. **An external thread schedules a task**: Calls `Notify(false)` to wake a worker.\n4. **`Notify(false)` sees `signals \u003c waiters` (0 \u003c 1)**: Instead of popping Worker A from `stack`, it increments `signals` from 0 to 1, intending for Worker B to consume the signal in `CommitWait()`.\n5. **Worker B finds work and aborts parking**: Worker B calls `CancelWait()`.\n6. **The Bug**: `CancelWait()` observes `waiters \u003d\u003d signals` (1 \u003d\u003d 1). To preserve the invariant `signals \u003c\u003d waiters`, it decrements both `waiters` and `signals` (`newstate -\u003d kSignalInc`). This completely discards the notification! Worker A remains parked on `stack` indefinitely, even though work is queued in the pool. If another worker blocks waiting for work to complete, the entire pool deadlocks.\n\n### Fix\nWhen `CancelWait()` decrements `signals` because `waiters \u003d\u003d signals`, set `notify \u003d true` and call `Notify(false)` after the CAS succeeds. This forwards the consumed signal to unpark a waiting thread from `stack`.\n\n### Testing\nAdded `test_cancel_wait_signal_forwarding()` in `threads_eventcount.cpp`, which deterministically reproduces this interleaving. Without the fix, the parked thread hangs and the test fails 100% of the time; with the fix, the test passes immediately.\n\nPiperOrigin-RevId: 987583984\nChange-Id: I6dc28ad1474a9ad2f9dea18e268838f0b1f76144\n"
    },
    {
      "commit": "746fe26e4c097835f874a7f43ccd94d7ba6d5bd9",
      "tree": "cebe6a4dcbcc9796c56015e2b82b386e25081199",
      "parents": [
        "04a6c55da3977f8f9e096f6306f4e0d68aeb284a"
      ],
      "author": {
        "name": "Junwhan Ahn",
        "email": "junwhan@google.com",
        "time": "Thu Sep 24 10:06:20 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Sep 24 10:06:56 2026 -0700"
      },
      "message": "Forward a consumed suppressed notification when a spinner has already stolen a task and other work remains queued. Otherwise the spinner can enter a blocking task while another newly scheduled task remains stranded in a queue with all other workers parked.\n\n### Deadlock Scenario\nWhen allow_spinning \u003d true, idle worker threads spin for up to kSpinCount (5,000) iterations in MaybeGetTask() calling GlobalSteal():\n1. While a worker is spinning (num_spinning \u003e 0), any concurrent Schedule() or ScheduleWithHint() calls elide ec_.Notify(false) to minimize wake-up overhead, incrementing num_no_notification instead. The thread pool assumes the spinning worker will pick up the newly submitted task.\n2. When the spinning worker finishes its spin loop, it calls StopSpinning(), which returns has_no_notify_task \u003d true if any notification was elided.\n3. If the spinning worker stole a different task t-\u003ef during its spin loop, unpatched code skipped the compensating steal and never notified a parked worker.\n4. If all other worker threads in the pool are parked, the elided task sits stranded in its run queue. If the stolen task currently executing blocks waiting for the stranded task (e.g. at an XLA collective rendezvous / barrier), the thread pool permanently deadlocks.\n\n### Fix\nIn MaybeGetTask(), when has_no_notify_task is true:\n```cpp\nif (has_no_notify_task) {\n  if (t-\u003ef) {\n    if (NonEmptyQueueIndex() !\u003d -1) ec_.Notify(false);\n  } else {\n    *t \u003d GlobalSteal();\n  }\n}\n```\nAlso acquire the queue publication in StopSpinning() and release in IsNotifyParkedThreadRequired() to pair queue publication with suppressed notifications.\n\n### Testing\nAdded DelayingThreadEnvironment in threads_non_blocking_thread_pool.cpp to widen the interval between taking a task and retiring the worker\u0027s spinner, reproducing the lost notification under test_parallelism.\n\nBackport of https://gitlab.com/libeigen/eigen/-/merge_requests/3038.\n\nPiperOrigin-RevId: 987583048\nChange-Id: I3336ed6ee95914579a54718e72f178865a3f27b5\n"
    },
    {
      "commit": "04a6c55da3977f8f9e096f6306f4e0d68aeb284a",
      "tree": "9439dcad874e2d9fe5ed12cbb38a4c883a9bc2b8",
      "parents": [
        "34de9d370db7d3d3e25d8f02330eaecf3875f54e"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Apr 23 13:17:47 2026 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Apr 23 13:18:09 2026 -0700"
      },
      "message": "Update Eigen to commit:ea13a98decd497a8c5588fb5de71b57bcf10d864\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nea13a98de - Fix imag_ref for real scalar types and clean up svd_fill.h\n929785924 - Fix more cache size queries.\nb2f95d373 - Fix more cache size queries.\n9ae0e0f19 - Remove include from within Eigen namespace.\nc1faa7473 - Add boundary test coverage: stableNorm, LinSpaced, complex GEMV, triangular solve\n6b9275d1a - Add test coverage for transpose, reverse, bool redux, select, diagonal-of-product at boundaries\n356a9ba1d - Add test coverage for matrix lpNorm, RowMajor partial reductions, selfadjoint boundaries\n15cae8348 - Add test coverage for strided maps, triangular blocking, and mixed storage orders\n93aa959b8 - Add vectorization boundary tests for redux and visitor\nc93116b43 - Improve test coverage for inner product, fill, reductions, and IO\n5e478d328 - Improve product test coverage at critical code-path boundaries\n3a2ba7c43 - Optimize `predux_any\u003cPacket4f\u003e`\n8190c82cb - Add missing SIMD math function benchmarks\n8368a12f0 - Add runtime cache size detection for ARM and improve GEMM blocking\n42c1dbd2c - Add aarch64 smoke test pipeline for MRs\n875fb48f0 - fix various irksome compiler warnings\n2a2456c87 - restore Eigen/src/Core/arch/Altivec/MatrixProduct.h to b1e74b1cc\nc4eb3c4f4 - fix custom visitors\n4387e3248 - Fix row-skipping bug in general_matrix_vector_product::run_small_cols\n81550faea - Use Web Archive for dead link for the PDF referenced in Geometry/EulerAngles.h\n42b6c43cf - Revert \"Remove random retry loops in tests (batch 2: indices and integer types)\"\n54458cb39 - Remove random retry loops in tests (batch 3: geometry, sparse, umeyama)\na3cb1c659 - cxx11_tensor_random: use retry loop for low-precision RNG collisions\nf80d7b825 - Fix three more flaky tests: igamma, tensor_random, matrix_power\n8eaa7552f - Fix three flaky tests: packetmath, array_cwise, polynomialsolver\ndd81698ae - Fix vectorization_logic test for wide SIMD widths\nab5878426 - Remove random retry loops in tests (batch 5: geometry, mixing types, triangular)\n411422f2d - Remove random retry loop in SVD min-norm test\n7c3a34476 - Remove random retry loops in tests (batch 2: indices and integer types)\nbe7538ed6 - Remove random retry loops in tests (batch 1: simple scalar cases)\n5790d716c - Simplify and optimize pow/cbrt special case handling\n3041ab44a - Fix GEBP asm register constraints for custom scalar types\n20fce70e5 - Fix another complex div edge case.\n5bacb5be9 - Fix null pointer dereference in Sparse-Dense products for Sparse vectors.\nd8c8ee6fb - Fix crash on construction of SparseMatrix with zero-length diagonal\n265496e86 - Fix heap overflow in BM_BatchContraction benchmark\neea4d31f5 - Simplify and modernize XprHelper.h\ndd826edb4 - Replace typedef with using in tensor contraction files\nabc3d6014 - Fix CUDA+Clang build warnings.\n0269c017a - Revise Tensor module README.md: fix bugs, add missing docs\nca94be70d - fix uninitialized variable in constexpr function\nb0ebf966a - Fix default rank-detection threshold in QR and LU decompositions\nd36a7db7b - Fix Eigen::array constructors.\n661cdb227 - Fix relative paths after move.\n57b1de233 - Fix row-major GEMV dropping rows when n8 heuristic disables main loop\n662d5c21f - Optimize SYMV, SYR, SYR2, and TRMV product kernels\nc66fc5286 - Add ULP accuracy measurement tool and documentation for vectorized math functions\nc20b6f5c4 - Restore EIGEN_EMPTY_STRUCT_CTOR as a no-op macro for backward compatibility\n77d917359 - Add ca-certificates to clang-tidy CI job\n444ae9761 - Clamp igamma/igammac output to [0,1] for numerical stability\neddb470a0 - Fix flaky array_cwise and sparse_basic tests\n25dba492e - Use stack-constructed variable for SVD block sweep.\nf64d1e0ac - Improve ConditionEstimator docs and tighten test bounds\n8525491eb - Add dedicated unit tests and benchmark for ConditionEstimator\ne730b1fe3 - Fix mixed products GEMM.\n3adfa9bd3 - Add const to non-mutating member functions across remaining modules\n13b61529f - Add const to non-mutating member functions in products/ and Serializer\naaca9e585 - Add missing const qualifiers in Eigen/src/Core/\n1b1b7e347 - Fix EIGEN_NO_AUTOMATIC_RESIZING not resizing empty destinations\n064d686c5 - Remove CXX11/ directory nesting for Tensor modules\n11eb66e1b - Remove pre-C++14 workarounds from unsupported/ tensor code\na95440de1 - Remove obsolete bench/ and btl/ directories\n6e2aff6b5 - Fix ambiguous static_cast in JacobiSVD blocking threshold\nd8ed4f688 - Fix GEBP half/quarter-packet loops for nr\u003e\u003d8 RHS packing on ARM64\n6b6d0d8c8 - Revert \"Fix ambiguous static_cast in JacobiSVD blocking threshold computation\"\nba2fc4e77 - Revert \"Fix GEBP half/quarter-packet loops for nr\u003e\u003d8 RHS packing on ARM64\"\n888d708dc - Fix GEBP half/quarter-packet loops for nr\u003e\u003d8 RHS packing on ARM64\ne567151ce - Fix ambiguous static_cast in JacobiSVD blocking threshold computation\na31de4778 - Blocked Jacobi SVD sweep with L2-cache-adaptive threshold\n647e0009b - Refactor BDCSVD D\u0026C code to reduce compilation time and memory footprint\n4fab38d79 - Make clang generic vector backend support 16, 32, and 64-byte vectors\nea25ea52b - Revert accidental changes from !2212 squash merge\n38f0f4275 - Update rmlarsen email address from @google.com to @gmail.com\nd0d70a952 - Consolidate complex math function boilerplate with shared macros\nc4c704e5d - Install libclang-rt-19-dev for asan-ubsan CI job\n61895c597 - Selectively add constexpr to Core expression template scaffolding\n34092d278 - Fix flaky tests: add iteration guards, yield in busy-waits, cap thread count\n28d090a49 - Refactor GenericPacketMathFunctions.h into smaller focused headers\n16da0279f - Add benchmarks for unsupported modules and extend supported benchmarks\nfa567f6bc - Add CUDA CI jobs with NVHPC (nvc++) as host and device compiler\n2cd9bb738 - Fix sparse product with entities that do not have direct access.\n00cc497d3 - Add clang-tidy, codespell, and sanitizer checks to CI pipeline\n241af1c0b - Add NVHPC (nvc++) compiler support and CI build/test jobs\nf3f2c676b - Fix direct access for sparse blocks.\nd537b51ed - Fix ComplexEigenSolver NaN with flush-to-zero arithmetic\n667cabe3a - Clean up comments in unsupported module\n78b76986b - Comment cleanup v3: trailing ??, informal language, FIXME/TODO colons\n112c2324b - Consolidate BF16/F16 wrapper macros and simplify arch math functions\nd5e67adbe - Clean up informal language, vague TODOs, and dead code in comments\n7d727d26b - Refactor GenericPacketMathFunctions.h into smaller focused headers\n9810969c0 - Suppress false-positive GCC and clang warnings in test builds\nad7f1fe70 - Improve clang vector extension backend\n1f49bf96c - Add new benchmarks for Core, LU, and QR operations\n8c35441f1 - Fix typos: misspellings, French variable names, and hyphenation\n44c613216 - Fix ~40 typos found by codespell across the codebase\nf52ad04bb - Fix ASAN-detected bugs in Diagonal::data() and array_cwise test\nd4077a6e9 - Reorganize benchmarks into subdirectories and clean up Eigen sources\n832b94097 - Update COPYING.README to clarify third-party license status\ne6accc73f - Fix comment typos, doubled words, grammar errors, and copy-paste mistakes\n0e424f405 - Remove dead code, commented-out blocks, and outdated comments\n18791a81b - Fix MSVC build: disable [[msvc::forceinline]] on generic lambdas\n95e8bc326 - Add EIGEN_LAMBDA_ALWAYS_INLINE macro for MSVC lambda inlining\na87ecfb17 - Use m_ prefix consistently for private/protected member variables\n270ea539f - Remove redundant EIGEN_STRONG_INLINE from trivial constexpr and \u003d default functions\ne0a8d6c9d - Fix compile warnings\n1dcea43c4 - Fix RowMajor performance for triangular/dense assignment\n374fe225b - Reduce GEMV and TRSM benchmark sizes for faster routine runs\n2c898e8b9 - Remove unused LhsPacketType typedef in gebp_peeled_loop\n4fdc82d69 - Fix mixed-type compilation error in row-major GEMV small-cols path\n4141d1fd2 - Fix -Wtautological-overlap-compare warning in row-major GEMV dispatch\n53e3408cb - Optimize GEMV kernels: row-major small-cols and template deduplication\n9c63d26de - Remove reference to nonexistent spmv.cpp in benchmarks\n5f09b3b63 - Fix missing template argument list in trsmKernelR for Clang 20/21\nc9eab4087 - Fix unused variable warning for phys_l1 on non-AVX512 builds\n3c86a013b - Vectorize generic trsmKernelR for non-AVX512 targets\n43a01f06a - update AVX and AVX512 to support gcc \u003c 10.1 and clang \u003c 10\n552ca8f15 - Simplify GEBP micro-kernel and improve blocking heuristics\ne953f1e50 - Modernize C++14 usage and minor optimizations in Core\nf69745b67 - Fix real x complex GEMM for backends where half \u003d\u003d full packet size\n073190be0 - Fix outdated documentation across multiple .dox files\nbdec88009 - Remove const from return-by-value types (issue #1087)\n3108f6360 - Migrate Eigen benchmarks to the Google benchmark framework\n740cac97b - Fix AVX double-precision trig and complex exp without AVX2\n50d6d92a7 - Optimize sparse-dense product by bypassing InnerIterator for compressed storage\nb6b2f31ba - Fix compiler warnings from GCC 13 and Clang 18\n113207a9d - Optimize JacobiSVD 2x2 kernel and hoist sweep threshold\ne6e5b5c4c - Fix pexp_complex for `complex\u003cdouble\u003e` (issue #3022)\n2b561f928 - Revert \"Specialized enable_borrowed_ranges for VectorwiseOp class range iteration\"\nd0654a201 - Specialized enable_borrowed_ranges for VectorwiseOp class range iteration\n1a2b80727 - Fix pdiv for complex packets involving infinites.\n9b709e826 - Diagonalview example typo\n23fcc1c6c - MatrixBase::diagonalView issue 604\n004d81a85 - Fix cblat3/zblat3 test program with gfortran.\n0ac2a2df9 - Prevent predux_half for DoublePacket from accidentally catching complex Packets of size \u003e\u003d 16\n4d05fcf8d - Fix packetmath tests on M* macs.\n752911927 - betainc edge case checks at start of calculation\nafb438053 - Fix RunQueue race condition on weak memory architectures (ARM64)\nc64829636 - Update HVX floating-point reduction to support V79 architecture.\nddfc68d39 - Fix clang vector backend type compatibility issues.\n93ff38884 - Fix relative tolerance scaling by multiplying with RHS norm in BiCGSTAB\n26c242ab5 - make EIGEN_BLAS macro names consistent and undef at end of file\n3d6f5fe8f - Tests: skip denorms in ARM ieee tests\nfdfdd4c96 - test suite: emit the function name when an ieee test fails\ne246f9cb6 - Use memset if !NumTraits\u003cScalar\u003e::RequireInitialization\nf46a2c561 - Fix bad static access for TensorDeviceGpu.\nc09151a2c - Wrote resizing documentation page\nf7772e394 - Gcc warnings\n918a5f1a6 - Fix warnings related to variable_if_dynamic.\n9a37aca9f - Fix assignment size assertion for EIGEN_NO_AUTOMATIC_RESIZING.\n251bff288 - CUDA 13 compatibility update for unit test gpu_basic\n0315fb319 - Change inline hint for `general_matrix_vector_product\u003c\u003e::run()` to gain performance\n7aea350ba - Fix more packetmath issues for RVV\n5d9beb81a - Initial version of reactivating RVV features like GeneralBlockPanelKernel\nd90a0534b - fix polynomialsolver test failures\n711118b74 - docs does not exists\nc30af8f3d - fix UB in random implementation and tests\nc5aa40675 - Fix `check_that_free_is_allowed` so that it properly checks `is_free_allowed` and not `is_malloc_allowed`\n5793499a5 - Fix AVX512FP16 build.\n2ac496ff8 - Revert !1953 and !1954\n9164d3f16 - Fix undefined behavior in packetmath.\n748e0a651 - Add missing semicolon\nfe973ab0c - Force early evaluation of boost expressions.\n976f15ebc - fix doc generation with doxygen 1.14 \u0026 1.15\n4f14da11d - Add basic support for packetmath for BF16 RVV\n21e4582d1 - Merge remote-tracking branch \u0027refs/remotes/origin/master\u0027\na7209fad7 - GemmKernel: Define static constexpr member variables out-of-class for C++14 compatibility\ncdc62b84c - Merge remote-tracking branch \u0027origin2/master\u0027\n26fe567dd - Fix FP16 for RVV so that it will compile for gcc\nafbf8173d - Merge remote-tracking branch \u0027origin2/master\u0027\n9b00db8cb - Simplify thread-safe initialization of GpuDeviceProperties.\n8cdc0fa67 - Fix naming of predux_half for RVV when LMUL \u003e 1\nf610edadc - Merge remote-tracking branch \u0027origin2/master\u0027\n5aefbab77 - Merge remote-tracking branch \u0027origin2/master\u0027\n3ff3d0378 - Merge remote-tracking branch \u0027origin2/master\u0027\n196eed3d6 - Merge branch \u0027master\u0027 of https://gitlab.com/libeigen/eigen\n\nPiperOrigin-RevId: 904604242\nChange-Id: I3b70435e8fee585af45ad5de8384531d716a57e8\n"
    },
    {
      "commit": "34de9d370db7d3d3e25d8f02330eaecf3875f54e",
      "tree": "079bb09d45e83a45a8cfc804d0cdd7f0bb68562a",
      "parents": [
        "9adff990ac72bd9cf32157cf0e9f57f6d3247b43"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Dec 03 11:18:01 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Dec 03 11:18:27 2025 -0800"
      },
      "message": "Update Eigen to commit:75bcd155c40cb48e647c87c3f29052360255bc9e\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n75bcd155c - Vectorize tan(x)\n01a919d13 - Fix AOCL cmake issues.\na73501cc7 - Added versioning for shared libraries.\ndb90c4939 - Add a ptanh_float implementation that is accurate to 1 ULP\n48eb5227c - Add BLAS function axpby.\na1eeb0220 - Expand CMake compatibility range for single-version specifications.\n8a1083e9b - Aocl integration updated\na6630c53c - Fix bug introduced in !2030\n\nPiperOrigin-RevId: 839836117\nChange-Id: I1dd1114e3a79200e9506a83c6c4db964d48deac2\n"
    },
    {
      "commit": "9adff990ac72bd9cf32157cf0e9f57f6d3247b43",
      "tree": "3e1f4a68b948645f47b580070ef1d4f8f4830cac",
      "parents": [
        "50853087ccc5fddec15b7fdb55a49f59728c75a7"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Thu Nov 20 16:47:52 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Nov 20 16:48:20 2025 -0800"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 834963038\nChange-Id: I8209614e03515300d12a032c799edffeb9bfea91\n"
    },
    {
      "commit": "50853087ccc5fddec15b7fdb55a49f59728c75a7",
      "tree": "508934b2f3f46c2164763ef791819c56bd0fdd0e",
      "parents": [
        "05b4f30bcae52b46c9ec7140c4269b1d51d5e7a8"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Nov 20 11:57:52 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Nov 20 11:58:21 2025 -0800"
      },
      "message": "Update Eigen to commit:49623d0c4e1af3c680845191948d10f6d3e92f8a\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n49623d0c4 - This patch adds support for RISCV\u0027s vector extension RVV1.0.\n8eb6551a8 - Add support for complex numbers in the generic clang backend\n1242d948d - Merge remote-tracking branch \u0027origin2/master\u0027\n8401a241c - Add summary to lldb pretty printing of Eigen::Matrix\ne75c29fd9 - EigenTesting.cmake: Quote argument to separate_arguments.\n38e2b9436 - Merge remote-tracking branch \u0027origin\u0027\n26dc851a7 - Merge remote-tracking branch \u0027origin\u0027\n332bfa95c - Merge remote-tracking branch \u0027origin/master\u0027\nadd5e7620 - Merge remote-tracking branch \u0027origin2/master\u0027\nf456635bf - Merge remote-tracking branch \u0027refs/remotes/origin/master\u0027\n745e019ed - Merge remote-tracking branch \u0027refs/remotes/origin/master\u0027\ne9b178bfe - Merge branch \u0027master\u0027 of https://gitlab.com/libeigen/eigen\n\nPiperOrigin-RevId: 834853880\nChange-Id: I1eff6869f000ef40e54f19a03fa56d56df7e4865\n"
    },
    {
      "commit": "05b4f30bcae52b46c9ec7140c4269b1d51d5e7a8",
      "tree": "667f7d1bcd154325e4fc0678a110de2e89c79c9f",
      "parents": [
        "cf56bbcf88571a0a90760b911462b80da33af97f"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Nov 13 11:40:40 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Nov 13 11:41:24 2025 -0800"
      },
      "message": "Update Eigen to commit:dcbaf2d608f306450f1e74949eb87e9a22a7ef4b\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\ndcbaf2d60 - Switch the inline hint to `EIGEN_ALWAYS_INLINE` for a few functions\na7674b70d - Improve packet op test coverage for IEEE special values.\n72bfca3d8 - cxx11_tensor_expr.cpp: delete extraneous semicolon\n\nPiperOrigin-RevId: 831944171\nChange-Id: I8aecde1b0e24ba41b1ffd9dc634b5d934a2eb21b\n"
    },
    {
      "commit": "cf56bbcf88571a0a90760b911462b80da33af97f",
      "tree": "cdb9c4a582a0571fd2902a08abe7a80d18e88cac",
      "parents": [
        "034dedf380e7cf78526038177f4fefc215b25ac1"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Nov 11 10:15:47 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Nov 11 10:16:18 2025 -0800"
      },
      "message": "Update Eigen to commit:9b511fe4fecf93c3438c43440c760dfa6f723b43\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n9b511fe4f - Fix cxx11_tensor_expr.cpp\n943fdc71c - Use more FMA in reciprocal iteration for precision\n1133aa82c - fix various compiler warnings\n8ae3b1aaa - Fix loongarch unsigned pabsdiff\n035cf6849 - Fix build of realview.cpp\n23a5482fc - Misc. packet math cleanups.\n4cb0776f8 - Add 5.0.1 release notes and a few unreleased features.\n8b85f5933 - Fix realview\nffcd7bdbd - Avoid breaking the build on older compilers.\nda867c31c - Fix defines in AVX512 custom TRSM kernel.\n\nPiperOrigin-RevId: 830969130\nChange-Id: Ifcafaaeed12734373fa9cc4742ce48267e7063c4\n"
    },
    {
      "commit": "034dedf380e7cf78526038177f4fefc215b25ac1",
      "tree": "495dbb4d065f917ba9f79d4575cbce2c4996b3a0",
      "parents": [
        "6f7da198637ab8f601b55f649a5150143e22d515"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Sat Nov 08 23:38:25 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Sat Nov 08 23:38:55 2025 -0800"
      },
      "message": "Update Eigen to commit:035cf684983197a13e579905d98198c365f44538\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n035cf6849 - Fix build of realview.cpp\n23a5482fc - Misc. packet math cleanups.\n4cb0776f8 - Add 5.0.1 release notes and a few unreleased features.\n8b85f5933 - Fix realview\nffcd7bdbd - Avoid breaking the build on older compilers.\nda867c31c - Fix defines in AVX512 custom TRSM kernel.\n\nPiperOrigin-RevId: 829998360\nChange-Id: I974aefe61d7a0071eb5e27d93f103b716e35ddf4\n"
    },
    {
      "commit": "6f7da198637ab8f601b55f649a5150143e22d515",
      "tree": "59172c6677ce62695a59366cf76eca514a62db1e",
      "parents": [
        "2d1422c771ec474cd2aaa68de5fadd89cf2ad315"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Sat Nov 08 22:49:30 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Sat Nov 08 22:50:10 2025 -0800"
      },
      "message": "Update Eigen to commit:035cf684983197a13e579905d98198c365f44538\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n035cf6849 - Fix build of realview.cpp\n23a5482fc - Misc. packet math cleanups.\n4cb0776f8 - Add 5.0.1 release notes and a few unreleased features.\n8b85f5933 - Fix realview\nffcd7bdbd - Avoid breaking the build on older compilers.\nda867c31c - Fix defines in AVX512 custom TRSM kernel.\n\nPiperOrigin-RevId: 829988986\nChange-Id: I796b10cf423c2e395939867aeda9847544da5660\n"
    },
    {
      "commit": "2d1422c771ec474cd2aaa68de5fadd89cf2ad315",
      "tree": "495dbb4d065f917ba9f79d4575cbce2c4996b3a0",
      "parents": [
        "ba8c11104b3a454236c955a44360c760fab4e99a"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Nov 07 11:22:39 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Nov 07 11:23:12 2025 -0800"
      },
      "message": "Update Eigen to commit:8a9bfb72d7074b63d8e6af12f78ef2fa0c61f4e3\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n8a9bfb72d - Rename preduce_half for HVX.\ned989c750 - Enable generic clang backend tests.\n3368ac6c6 - Don\u0027t set platform-specific vectorization macros for generic backend.\nfecfa7f27 - Fixes to make generic backend build with AVX512\nec93a6d09 - Add a generic Eigen backend based on clang vector extensions\n\nPiperOrigin-RevId: 829512412\nChange-Id: I088d23927b2ebdf35dd337f5187a1070464d2fb0\n"
    },
    {
      "commit": "ba8c11104b3a454236c955a44360c760fab4e99a",
      "tree": "6be6fb9cedea6063515f7459479aff7354473b56",
      "parents": [
        "c1524df5ff08f73f3c0355594b524cc3ad6d7a4c"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Thu Nov 06 21:23:45 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Nov 06 21:24:14 2025 -0800"
      },
      "message": "Update Eigen to commit:3368ac6c6934af3d7c16b05727084b4e08d07497\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n3368ac6c6 - Don\u0027t set platform-specific vectorization macros for generic backend.\nfecfa7f27 - Fixes to make generic backend build with AVX512\nec93a6d09 - Add a generic Eigen backend based on clang vector extensions\n\nPiperOrigin-RevId: 829252717\nChange-Id: I563971d8eb82645cb07f78e983e18fca732b7f44\n"
    },
    {
      "commit": "c1524df5ff08f73f3c0355594b524cc3ad6d7a4c",
      "tree": "9af3139d16518cb5d3da361609af48fdc363007e",
      "parents": [
        "b3296eaa66c31645942a8f7b5c097dbd6575e1d5"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Nov 06 17:09:16 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Nov 06 17:09:44 2025 -0800"
      },
      "message": "Update Eigen to commit:3368ac6c6934af3d7c16b05727084b4e08d07497\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n3368ac6c6 - Don\u0027t set platform-specific vectorization macros for generic backend.\nfecfa7f27 - Fixes to make generic backend build with AVX512\nec93a6d09 - Add a generic Eigen backend based on clang vector extensions\n\nPiperOrigin-RevId: 829167450\nChange-Id: I074af9be5dd099ec9109e8d36964c3274ba5d234\n"
    },
    {
      "commit": "b3296eaa66c31645942a8f7b5c097dbd6575e1d5",
      "tree": "6be6fb9cedea6063515f7459479aff7354473b56",
      "parents": [
        "3e28d6aebdad8722ad691eab6b3f73d0c8e847a8"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Nov 06 08:27:39 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Nov 06 08:28:35 2025 -0800"
      },
      "message": "Update Eigen to commit:7c7d84735e58b8397922070ec72eca66d6e811c7\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n7c7d84735 - Align temporary array in TensorSelectOp packet evaluator.\n142caf889 - Fix MKL enum conversion warning.\n9e5714b93 - Remove deprecated CUDA device properties.\n06f5cb487 - Use wrapper macro for multidimensional subscript feature test\n63fc0bc8c - Make TernarySelectOp immune to const differences.\n71703a981 - Make assume_aligned a no-op on ARM \u0026 ARM64 when msan is used, to work around a missing linker symbol.\nf95b4698f - Add support for C++23 multidimensional subscript operator\nb6fcddccf - Get rid of pblend packet op.\ned9a0e59b - Fix more bugs in !2052\na20fc40e4 - Revert \"simplify squaredNorm\"\n04eb06b35 - Fix doc references for nullary expressions.\nbfdbc031c - Fixes #2998.\n8716f109e - Implement assume_aligned using the standard API\nce70a507c - Enable more generic packet ops for double.\nfb5bb3e98 - simplify squaredNorm\nece9a4c0b - Always vectorize abs2() for non-complex types.\n60122df69 - Allow user to configure if `free` is allowed at runtime.\n923488391 - Fix SparseVector::insert(Index) assigning int to Scalar\nbe56fff1f - Fix ambiguous sqrt() overload caused by ADL\n2e91853ad - Fix a benign bug in ComplexQZ\n1a5eecd45 - Clarify range spanning major versions only works with 3.4.1.\nb4209fe98 - Eliminate use of std::cout in ArpackSelfAdjointEigenSolver.\nac3ef16f3 - Fix SparseVector::insertBack() with custom scalar types\n40da5b64c - CI enhancements: visual indication of flaky tests\n8e60d4173 - Support AVX for i686.\n\nPiperOrigin-RevId: 828966695\nChange-Id: I3a5adcf24e7c4510e3e8a4106c458456fcae8107\n"
    },
    {
      "commit": "3e28d6aebdad8722ad691eab6b3f73d0c8e847a8",
      "tree": "cc8c80f94ae81fc385f37eb387adc9e667ca0bbb",
      "parents": [
        "9aa7d4f33b4b56bfc1d92989e54a9e89f564b45c"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Tue Nov 04 16:54:54 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Nov 04 16:55:23 2025 -0800"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 828189485\nChange-Id: I2b46aabcbe0e77cda4cf3a65040b32af9e7d2763\n"
    },
    {
      "commit": "9aa7d4f33b4b56bfc1d92989e54a9e89f564b45c",
      "tree": "cbf50e7d5416ca382c4dd801a76f2c694fc4ac17",
      "parents": [
        "2fe79fe778163f0577efae55382720e5e3536d24"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Nov 04 14:59:44 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Nov 04 15:00:14 2025 -0800"
      },
      "message": "Update Eigen to commit:63fc0bc8c1fa0039a7ad535d1da3f565a9d1d18d\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n63fc0bc8c - Make TernarySelectOp immune to const differences.\n71703a981 - Make assume_aligned a no-op on ARM \u0026 ARM64 when msan is used, to work around a missing linker symbol.\nf95b4698f - Add support for C++23 multidimensional subscript operator\nb6fcddccf - Get rid of pblend packet op.\ned9a0e59b - Fix more bugs in !2052\na20fc40e4 - Revert \"simplify squaredNorm\"\n04eb06b35 - Fix doc references for nullary expressions.\nbfdbc031c - Fixes #2998.\n8716f109e - Implement assume_aligned using the standard API\nce70a507c - Enable more generic packet ops for double.\nfb5bb3e98 - simplify squaredNorm\nece9a4c0b - Always vectorize abs2() for non-complex types.\n60122df69 - Allow user to configure if `free` is allowed at runtime.\n923488391 - Fix SparseVector::insert(Index) assigning int to Scalar\nbe56fff1f - Fix ambiguous sqrt() overload caused by ADL\n2e91853ad - Fix a benign bug in ComplexQZ\n1a5eecd45 - Clarify range spanning major versions only works with 3.4.1.\nb4209fe98 - Eliminate use of std::cout in ArpackSelfAdjointEigenSolver.\nac3ef16f3 - Fix SparseVector::insertBack() with custom scalar types\n40da5b64c - CI enhancements: visual indication of flaky tests\n8e60d4173 - Support AVX for i686.\n\nPiperOrigin-RevId: 828148483\nChange-Id: Ib58c71600d7888d442aab2a03b21d494d20484c8\n"
    },
    {
      "commit": "2fe79fe778163f0577efae55382720e5e3536d24",
      "tree": "cc8c80f94ae81fc385f37eb387adc9e667ca0bbb",
      "parents": [
        "a8c2f0530aa63f6ebeffa03f8a15a7ddc3de85cc"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Wed Oct 22 09:13:52 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Oct 22 09:14:22 2025 -0700"
      },
      "message": "Update Eigen to commit:2aa2ff29002d070aa0f5aa86c06bf93e3bb7eff8\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2aa2ff290 - More ComplexQZ fixes.\nf5907c593 - Fix commit references in changelog.\ncc8748791 - Include required headers with relative paths in ComplexQZ\n73da4623b - Try disabling the cache again for ROCm.\nd739b9dc6 - Set the merge request template to be default.\ne77977e23 - Set the merge request template to be default.\n4c8744774 - Fixes #2987: delete unused variable steps\nd426838d0 - fix complexqz Wpedantic warnings\ndb02f9785 - Add a missing `#include \u003cversion\u003e` to Core.\nda55dd147 - Cleanup: Move 2x2 real SVD into the Jacobi module where it naturally belongs.\n99f851298 - ComplexQZ\ne1f1a608b - Fix DLL builds and c++ lapack declarations.\n3abaabb99 - Streamline merge request and bug templates.\n52358cb93 - Grammar fix \"must has\" --\u003e \"must have\".\n565db1c60 - Optimize ApplyOnTheRight for Jacobi rotations when FMA is available\n3bd0bfe0e - Disable ROCm job cache.\n\nPiperOrigin-RevId: 822616299\nChange-Id: Id0a3dbb302250cbce62bc67e21994f6b6de61987\n"
    },
    {
      "commit": "a8c2f0530aa63f6ebeffa03f8a15a7ddc3de85cc",
      "tree": "ae69ce7e3749f920c72aa4dd9f73f4605065f7c3",
      "parents": [
        "602155b5d303401d492ba792a78e4090dd7cdf45"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Tue Oct 21 10:38:16 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Oct 21 10:38:51 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 822182461\nChange-Id: I5f56f9b9d375075c0a66401587004bf920774fc8\n"
    },
    {
      "commit": "602155b5d303401d492ba792a78e4090dd7cdf45",
      "tree": "cf6e14490ed610ee94838fb2b89d823a37f08042",
      "parents": [
        "a33de0453f72732f452bd76348f31c06de7d2b3b"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Oct 16 16:55:10 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Oct 16 16:55:39 2025 -0700"
      },
      "message": "Update Eigen to commit:cd4f989f8f9288ab5aed1643ecb04c7be021021e\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\ncd4f989f8 - assume_aligned uses bytes not bits\nac7c192e1 - Add a bunch of useful scripts for planning releases.\n5bc944a3e - Fix jacobi svd for TriangularBase\ndbe9e6961 - Fix BLAS/LAPACK DLL usage on Windows.\nef3c5c1d1 - Add workaround for using std::fma for scalar multiply-add.\n5996176b8 - Fix alignment bug in avx pcast\u003cPacket4l, Packet4d\u003e\n4bd382df5 - Fix SSE PacketMath Compilation Error on QNX\n13bd14974 - fix errors in windows builds and tests\neea6587b0 - Fix scalar_inner_product_op when binary ops return a different type\n7eaf9ae68 - Add a method to SelfAdjointEigenSolver for computing the matrix exponential\n32b0f386b - Eliminate possible -Wstringop-overflow warning in .setZero()\n1edf360e3 - Fix line endings\nccde35bcd - Update dev version number.\na67f9dabb - tests: add missing link\ne6792039f - Update changelog to reflect 3.4.1 and 5.0.0 releases.\n4916887f2 - Update `geo_homogeneous` test, add eval() to PermutationMatrix.\n5c1029be1 - The \u0027CompressedStorageIterator\u003c\u003e\u0027 needs to satisfy the RandomAccessIterator\nf9f515fb5 - get rid of a bunch of windows jobs\n2e5447e62 - STYLE: Scripts with shebang should be executable\n8d7ebac6e - Disambiguate multiplication of a permutation matrix and a homogeneous vector\nbea7f7c58 - SparseMatrixBase: delete redundant/shadowed typedef\n7292c78e1 - blas: Fix parenthesis suggestion warning\ne524488eb - Convert Mercurial hgeol to gitattributes\ndbd25f632 - Fix select: return typed comparisons if vectorized\n027dc5bc8 - Extend the range of supported CMake package config versions\n4df215785 - Support matrix multiplication of homogeneous row vectors\n2d170aea1 - Define pcmp_le generically in terms of pcmp_eq and pcmp_lt.\nea869e183 - Add missing bool SSE2 PacketMath comparison\n6ef18340a - CMake: Explicit STATIC libs\n14477c5d4 - Replace deprecated std::is_trivial by an internal definition\nb2ec79a23 - Move smoketests to small GitLab runners.\n62fbd276e - Provide hints for deprecated functionality\nd38d669fd - Fix real schur exceptional shift\n4ac3e71f7 - CMake: Require at least C++14\na627f72cd - Add \"Version\" file and update version.\ne0a59e5a6 - Fix typo\n6cd6284f7 - Make the sparse matrix printer pretty\ne5f3fa2d6 - Add gemmtr implementation.\nf426eff94 - Add inline/device-function attributes to fma.\nda1a34a6b - Zero-out matrix for empty set of triplets.\n52fc978c6 - fixed typo `sparcity` -\u003e `sparsity`\n8a8fbc8f5 - Don\u0027t enable AVX for wasm.\n70d8d99d0 - Only build docs on push to main branch, not MRs.\n7f0cb638c - Specialize numext::madd for half/bfloat16.\n1e9d7ed7d - Add missing semicolon to `has_fma` definitions to fix GPU builds.\n5d4485e76 - Move more jobs to gitlab runners.\n2e8cc042a - Replace calls to numext::fma with numext:madd.\n52f570a40 - Move GPU ci jobs to gitlab-hosted runners.\n38b51d5b7 - Mitigate setConstant regression with custom scalars\nd2a70fe4e - Make permutation products aliasing by default.\n4ae564735 - Fix direct index aliased assignment.\n1a45d2168 - Fix use of FMA in triangular solver for boost multiprecision.\n05e74b1a4 - Tiny fix in `unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h`\nd36899812 - Fix MSVC error about missing std::bit_cast\nc487a4fe9 - Clean up most of testsuite on s390x\n\nPiperOrigin-RevId: 820448272\nChange-Id: Ie30f90f50d513e695d90b1295061104a87e8a762\n"
    },
    {
      "commit": "a33de0453f72732f452bd76348f31c06de7d2b3b",
      "tree": "022e00732a53e4d2f3431267c409470bb1aadc8d",
      "parents": [
        "0cfa9dc353c22d6544ab71d2843c7cdfe4b26167"
      ],
      "author": {
        "name": "Bogdan Graur",
        "email": "bgraur@google.com",
        "time": "Tue Sep 30 03:43:09 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Sep 30 03:43:47 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 813192026\nChange-Id: I6205d1bf30a2d04d5d21043120d0d3c4348025d6\n"
    },
    {
      "commit": "0cfa9dc353c22d6544ab71d2843c7cdfe4b26167",
      "tree": "9c98b3d0c5d74a2d949440c7f58059f2aea4e0a5",
      "parents": [
        "ebd59e832fec737951148f9bc97a07e4d970f438"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Aug 15 12:20:24 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Aug 15 12:20:51 2025 -0700"
      },
      "message": "Update Eigen to commit:4033cfcc1dd45b3cdf7285afd93556f2cfbe9425\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n4033cfcc1 - Fix dangling reference in VectorwiseOp::iterator: Episode II: The Dependent Typedef Strikes Back\ne9dfbad61 - Fix dangling reference in VectorwiseOp::iterator\n43a65a9cb - add RealView api\n\nPiperOrigin-RevId: 795566882\nChange-Id: I308f46d5fdbc5a41595a79cf9e45d700b63988ef\n"
    },
    {
      "commit": "ebd59e832fec737951148f9bc97a07e4d970f438",
      "tree": "00bd6e6ed1139939000c1e5ce96ff90deef31d53",
      "parents": [
        "ec4263e632e4d60cfbfa509f8dea77c7e3371749"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Aug 12 22:02:11 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Aug 12 22:02:42 2025 -0700"
      },
      "message": "Update Eigen to commit:954e21152e204b1960aca802eb9f16d054d70fd9\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n954e21152 - Include \u003climits\u003e in test main.h\ne15cd620a - Remove select class\n1c0048a08 - Fix inconsistency between ptrue and pcmp_* in HVX\nddce1d7d1 - Fixes #2952\n8b9dbcdaa - Fix numext::bit_cast() compilation failure in C++20\n975a5aba4 - Fix TODO: Use std::bit_cast or __builtin_bit_cast if available.\n\nPiperOrigin-RevId: 794397153\nChange-Id: I6acf902aeeafe4f69dd9fafefa9d823615d01970\n"
    },
    {
      "commit": "ec4263e632e4d60cfbfa509f8dea77c7e3371749",
      "tree": "226f835e96326f74177d415a77414fea3020e1e5",
      "parents": [
        "181ca0f3ef0f77064b4e08a7f24a4fbf89279313"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Aug 06 11:27:56 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Aug 06 11:28:32 2025 -0700"
      },
      "message": "Update Eigen to commit:4be7e6b4e0a82853e853c0c7c4ef72f395e1f497\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n4be7e6b4e - Fix pcmp_* for HVX to comply with the new definition of true \u003d Scalar(1)\nedcf4c135 - Remove fortran dependency for eigenblas.\ne4493233e - Fix EIGEN_OPTIMIZATION_BARRIER for clang-cl\nf5ead2d34 - Fix intel packet math header inclusion order\n1e65707aa - Suppress Warray-bounds warning in generic ploaduSegment, fix edge case for vectorized cast\nabeba8535 - Use proper float literals in SpecialFunctionsImpl.h.\nb5bef9dcb - Fix bug in Erfc introduced in !1862.\n97c7cc620 - Explicitly use the packet trait HasPow to control whether Pow is vectorized.\nefe5b6979 - Unconditionally include \u003cmemory\u003e. Some c++20 builds are currently broken because it is needed for std::assume_aligned.\n2cf66d4b0 - Use numext::fma in more places in SparseCore.\nd7fa5ebe0 - Fix API incompatibility for ILU in superLU support\ncedf1f4c1 - Fix typo: duplicated \u0027\\\u0027\u0027for\u0027\\\u0027\u0027 in docs\n302fc46bc - arm packet alignment requirements and aligned loads/stores\n430e35fbd - Fixed -Wshadow warning by renaming variables\nbd0cd1d67 - Fix self-adjoint products when multiplying by a compile-time vector.\n6854da2ea - Fix 1x1 selfadjoint matrix-vector product bug\nac1b29f82 - Set CMake POLICY CMP0177 to NEW\n849a33624 - Move default builds/tests to GitLab runners.\n8ac2fb077 - Use numext::fma for sparse x dense dot product.\ncc0be0043 - Fix docs build.\nf169c13d8 - Replace PPC g++-10 with g++14.\n7fa069ef9 - tensor documentation\n7c636dd5d - Move HIP/CUDA defines to Core.\n26616fe5b - Fix VSX packetmath psin and pcast tests.\na395ee162 - Fix a collection of random failures encountered when testing with Bazel.\n0bce653ef - Use QEMU for arm and ppc tests.\ndb8bd5b82 - Modify pselect and various masks to use Scalar(1) for true.\n6de0515fa - Create a changelog file.\n98fbf6ed7 - Decommission aarch64 ampere runner.\n\nPiperOrigin-RevId: 791772929\nChange-Id: Iccf81d31e1ae3e4e57b7fee8c55561ac83fc03d0\n"
    },
    {
      "commit": "181ca0f3ef0f77064b4e08a7f24a4fbf89279313",
      "tree": "b2d49493cd0cab4322dadaf666c17c70d7a850f0",
      "parents": [
        "0e8b00e7f96cf2ba8ca9b7d75eac6ad325fcc2a7"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Jun 23 13:33:25 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Jun 23 13:33:55 2025 -0700"
      },
      "message": "Update Eigen to commit:81044ec13df7608d0d9d86aff2ef9805fc69bed1\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n81044ec13 - Provide macro to explicitly disable alloca\nbcce88c99 - Faster emulated half comparisons\nac6955ebc - Remove MSVC warnings in FindCoeff.h\n67a898a07 - Fix unprotected SIZE in macro.\ncdf6a1f5e - Add OpenBLAS sbgemm.\nd228bcdf8 - Fix neon compilation bug\n994f3d107 - Fix neon packet math tests, add missing neon intrinsics\ncda19a625 - Make `Eigen::Map\u003cconst Vector\u003e::operator[]` return correct type\n\nPiperOrigin-RevId: 774907248\nChange-Id: Ic572084cfda6bbbe9bb0f5ec94da1ea07671e3e4\n"
    },
    {
      "commit": "0e8b00e7f96cf2ba8ca9b7d75eac6ad325fcc2a7",
      "tree": "ef12ad82e76803deaf5ac56cf6abd7fcedf34771",
      "parents": [
        "4c7a9cf07c8a33890fb87f33a6417e7c648c5564"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Jun 09 13:32:29 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Jun 09 13:33:08 2025 -0700"
      },
      "message": "Update Eigen to commit:d0b490ee091629068e0c11953419eb089f9e6bb2\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nd0b490ee0 - Optimize maxCoeff and friends\nc458d68fa - Fix compile warning about * with bool.\n3f00059be - Fix fuzzer range error for scalar parity check.\n\nPiperOrigin-RevId: 769272347\nChange-Id: Ic2aaaa6638027acaa5c47c8f9ab9fe0c3c0d8f78\n"
    },
    {
      "commit": "4c7a9cf07c8a33890fb87f33a6417e7c648c5564",
      "tree": "6ca2e9cd8f8a97a98c992f1e5478595c96d1a2fe",
      "parents": [
        "2bf945818c79300075c9dea26dc3874b41a69d42"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Jun 03 14:01:24 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Jun 03 14:01:56 2025 -0700"
      },
      "message": "Update Eigen to commit:21e89b930c6af56dbdaeea2a91d8b9d6fd2c208a\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n21e89b930 - Enable default behavior for pmin\u003cPropagateFast\u003e, predux_min, etc\n4fdf87bbf - clean up intel packet reductions\na7f183cad - Add factory/getters for quat coeffs in both orders\nd81aa18f4 - Explicitly construct the scalar for non-implicitly convertible types\n\nPiperOrigin-RevId: 766811637\nChange-Id: I6184e3c71bb418f61715e8c3d5fb29d357749417\n"
    },
    {
      "commit": "2bf945818c79300075c9dea26dc3874b41a69d42",
      "tree": "dfd6d67f814d754b0a48c0c0172caf25781ccdc7",
      "parents": [
        "22f6524ff35c9c5aa20dc320bdec8b424da36fa1"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu May 15 12:38:22 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu May 15 12:38:53 2025 -0700"
      },
      "message": "Update Eigen to commit:171bd08ca987987c3c50f0fa5dd8914bdd42dd3b\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n171bd08ca - fix 2849\ndb85838ee - Add DUCC FFT support\n6f1a14341 - Ensure info() implementation across all SolverBase derived types\nf3e7d64f3 - Fix: Correct Lapacke bindings for BDCSVD and JacobiSVD to match the updated API\n434a2fc4a - Fix obsolete comment in InverseImpl.h. We use PartialPivLU for the general case.\n\nPiperOrigin-RevId: 759257247\nChange-Id: I31deb8d802b88861782b34a089eecbdf71abfd47\n"
    },
    {
      "commit": "22f6524ff35c9c5aa20dc320bdec8b424da36fa1",
      "tree": "78f1253df4676922a4c6d60fba8c4449bac69320",
      "parents": [
        "d10e741dbbd25096637a48288609b247e06bc273"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu May 08 11:42:54 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu May 08 11:43:25 2025 -0700"
      },
      "message": "Update Eigen to commit:ae3aba99db4c829b4cc4d9fdd54321dedd814dc4\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nae3aba99d - Fix typo in CoreEvaluators.h\nee4f86f90 - Fix MSAN in vectorized casting evaluator\n6dbbf0a84 - CMake: only create uninstall target when eigen is top level\nfb2fca90b - Avoid unnecessary matrix copy in BDCSVD and JacobiSVD\nd6b23a225 - Fix unused local typedef warning in matrix exponential\n\nPiperOrigin-RevId: 756397904\nChange-Id: Ic6f355c0f282a639defe5f55dbbbc8ae01cd956f\n"
    },
    {
      "commit": "d10e741dbbd25096637a48288609b247e06bc273",
      "tree": "c0a6b5ca8cb9e1b70af468d535dc54c015b0e9fd",
      "parents": [
        "29dbcba5db6ba4ff3cb8ad783b9ea0abb0249bc3"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu May 01 13:21:16 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu May 01 13:21:47 2025 -0700"
      },
      "message": "Update Eigen to commit:d2dce377670f1742f3f701b3585682d02f10bb92\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nd2dce3776 - Optimize slerp() as proposed by Gopinath Vasalamarri.\n66d8111ac - Use a more conservative method to detect non-finite inputs to cbrt.\nd6689a15d - Replace instances of EIGEN_CONSTEXPR macro\n33f5f5961 - Vectorize cbrt for float and double.\n533096090 - Enable packet segment in partial redux\n6266d430c - packet segment: also check DiagonalWrapper\ne39ad8bad - fix constexpr in CoreEvaluators.h\n7aefb9f4d - fix memset optimization for std::complex types\n73ca849a6 - fix packetSegment for ArrayWrapper / MatrixWrapper\n28c3b26d5 - masked load/store framework\ncebe09110 - Fix a potential deadlock because of Eigen thread pool\n11fd34cc1 - Fix the typing of the Tasks in ForkJoin.h\n2cd47d743 - Fixe Conversion Warning in Parallelizer\nb86004226 - Add postream for ostream-ing packets more reliably.\n02d9e1138 - Add missing pmadd for Packet16bf.\n9cc9209b9 - Fix cmake warning and default to j0.\ne0c99a8dd - By default, run ctests on all available cores in parallel.\n63a40ffb9 - Use fma\u003cfloat\u003e for fma\u003chalf\u003e and fma\u003cbfloat16\u003e if native fma is not available on the platform.\n44fb6422b - All triggering full CI if MR label containts all-tests\n3866cbfbe - Fix test for TensorRef of trace.\n6579e36eb - Allow Tensor trace to be passed to a TensorRef.\n8e32cbf7d - Reduce flakiness of test for Eigen::half.\nd935916ac - Add numext::fma and missing pmadd implementations.\n754bd24f5 - fix 2828\nac2165c11 - fix allFinite\n314396819 - Generalize the Eigen ForkJoin scheduler to use any ThreadPool interface.\n70f2aead9 - Use native _Float16 for AVX512FP16 and update vectorization.\n0259a52b0 - Use more .noalias()\n14f845a1a - Fix givens rotation.\n33b04fe51 - CMake: add install-doc target\n10e62ccd2 - Fix x86 complex vectorized fma\n\nPiperOrigin-RevId: 753703769\nChange-Id: I43bc1cf7c598ca3f306fffea9844e9c5b1a21b79\n"
    },
    {
      "commit": "29dbcba5db6ba4ff3cb8ad783b9ea0abb0249bc3",
      "tree": "4a7482ed8586264951507c08358a28bd92466d0b",
      "parents": [
        "7f2f981fe4c7df98a8de8813cc7ecf0d875cf8bf"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Wed Apr 09 12:44:31 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Apr 09 12:45:12 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 745695659\nChange-Id: Ifc595fe6199b3e4be51480325823ccf8f24b56c4\n"
    },
    {
      "commit": "7f2f981fe4c7df98a8de8813cc7ecf0d875cf8bf",
      "tree": "536fc53671fab3107f093e608601b43e3987c9e7",
      "parents": [
        "b5993254bc3b251f756e697ea29501d6535354d2"
      ],
      "author": {
        "name": "William Kong",
        "email": "weiweikong@google.com",
        "time": "Wed Apr 09 11:02:08 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Apr 09 11:02:42 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 745656995\nChange-Id: I6295a54f9eb6faa70a16fdcaa83771a94a7fab23\n"
    },
    {
      "commit": "b5993254bc3b251f756e697ea29501d6535354d2",
      "tree": "a9d598e2908fa8919136b7f67d2b24d30b12d34e",
      "parents": [
        "f5e2f3d38528a3f1ab8fc93b17a93e553700a669"
      ],
      "author": {
        "name": "William Kong",
        "email": "weiweikong@google.com",
        "time": "Mon Apr 07 10:59:22 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Apr 07 10:59:49 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 744780506\nChange-Id: I45d84a786b66e78e4de0e342cf951b5efb94b7cd\n"
    },
    {
      "commit": "f5e2f3d38528a3f1ab8fc93b17a93e553700a669",
      "tree": "6c2adeb732994b2f74233320c5766c163e040a18",
      "parents": [
        "85fc59bf36e6f6043cab2a62a27774704cff0ad1"
      ],
      "author": {
        "name": "William Kong",
        "email": "weiweikong@google.com",
        "time": "Fri Apr 04 15:29:58 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Apr 04 15:30:30 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 744088498\nChange-Id: I0a37cdca37408be60ff80f32523fcd03ef99d334\n"
    },
    {
      "commit": "85fc59bf36e6f6043cab2a62a27774704cff0ad1",
      "tree": "3fc528d73918acd20406f3c933c96aaeeacd3589",
      "parents": [
        "c8c2d6902f66cc8b61fd863c9c4573f157a07cfe"
      ],
      "author": {
        "name": "William Kong",
        "email": "weiweikong@google.com",
        "time": "Fri Apr 04 14:44:32 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Apr 04 14:45:11 2025 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 744075909\nChange-Id: Ibc63343d33da09049362cd4de5eb5a7a3442e1be\n"
    },
    {
      "commit": "c8c2d6902f66cc8b61fd863c9c4573f157a07cfe",
      "tree": "231050e82cdcb8b0727edc867131a818575cced0",
      "parents": [
        "48a4844c182dfc371767e41961e35f603f0bf90d"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Mar 10 09:23:44 2025 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Mar 10 09:24:26 2025 -0700"
      },
      "message": "Update Eigen to commit:464c1d097891a1462ab28bf8bb763c1683883892\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n464c1d097 - Format TensorDeviceThreadPool.h \u0026 use if constexpr for c++20.\n21223f6bb - Fix addition of different enum types.\n350544eb0 - Clean up TensorDeviceThreadPool.h\n43810fc1b - Fix extra semicolon in DeviceWrapper\nd28041ed5 - refactor AssignmentFunctors.h, unify with existing scalar_op\n9a8621403 - Optimize division operations in TensorVolumePatch.h\nbe5147b09 - Fix STL feature detection for c++20.\n179a49684 - Fix CMake BOOST warning\n\nPiperOrigin-RevId: 735405240\nChange-Id: Ie55bcb8e50a5323c276230c5a89ac630c547bfb1\n"
    },
    {
      "commit": "48a4844c182dfc371767e41961e35f603f0bf90d",
      "tree": "1c584a01c214b2225061457b7f9f5cf94b7f198e",
      "parents": [
        "506d025d1c04800d9aaf569cc2bc2f4a9026efb5"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Feb 26 16:19:32 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Feb 26 16:19:56 2025 -0800"
      },
      "message": "Update Eigen to commit:dd56367554cdf0662cc1d4a8e462e8c3e8656d08\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\ndd5636755 - Fix docs job for nightlies\nd79bac0d3 - Fix boolean scatter and random generation for tensors.\n9935396b1 - Specify constructor template arguments for ConstexprTest struct\n72adf891d - Slightly simplify ForkJoin code, and make sure the test is actually run.\n6aebfa9ac - Build docs on push, and don\u0027\\\u0027\u0027t expire\nbddaa99e1 - Fix bitwise operation error when compiling as C++26\ne42dceb3a - Fix implicit copy-constructor warning in TensorRef.\n5fc6fc988 - Initialize matrix in bicgstab test\n0ae7b5901 - Make assignment constexpr\n4dda5b927 - fix Warray-bounds in inner product\n\nPiperOrigin-RevId: 731494675\nChange-Id: Id59938322667f1bd2b3348ce69f1bc143608e9e4\n"
    },
    {
      "commit": "506d025d1c04800d9aaf569cc2bc2f4a9026efb5",
      "tree": "e99b5864961fb0a3a14e3fa66d716c8a789135bb",
      "parents": [
        "12566d89f20d46ce979eb92f933fc5acbd2831fd"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Feb 20 18:43:56 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Feb 20 18:44:31 2025 -0800"
      },
      "message": "Update Eigen to commit:66f7f51b7e069d0a03a21157fa60b24aece69aeb\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n66f7f51b7 - Disable fno-check-new on clang.\n151f6127d - Fix Warray-bounds warning for fixed-size assignments\n1d8b82b07 - Fix power builds for no VSX and no POWER8.\neb3f9f443 - refactor AssignmentEvaluator\n9c211430b - Fix TensorRef details\n22cd7307d - Remove assumption of std::complex for complex scalar types.\n6b4881ad4 - Eliminate type-punning UB in Eigen::half.\n420d891de - Add missing mathjax/latex configuration.\nbecefd59e - Returns condition number of zero if matrix is not invertible.\n809d266b4 - Fix numerical issues with BiCGSTAB.\nef475f277 - Add missing graphviz to doc build.\na0591cbc9 - Fix doxygen-generated pages\n715deac18 - Add EIGEN_CI_CTEST_ARGS to allow for custom timeout.\n\nPiperOrigin-RevId: 729330781\nChange-Id: I16aadc1cb1ddf4816604ce120e933f9660da17f8\n"
    },
    {
      "commit": "12566d89f20d46ce979eb92f933fc5acbd2831fd",
      "tree": "f3962a0fa4fc0d4b46784a4b9d68642eb3b2ac8c",
      "parents": [
        "33de8fe7e3a3c4a1100a1de0842497f2286047e9"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Feb 06 14:11:39 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Feb 06 14:12:26 2025 -0800"
      },
      "message": "Update Eigen to commit:4c38131a16803130b66266a912029504f2cf23cd\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n4c38131a1 - Fix android hardware_destructive_inference_size issue.\n\nPiperOrigin-RevId: 724069969\nChange-Id: I0e8b46df0be1c65d24129440d7e8ae5c68796d6d\n"
    },
    {
      "commit": "33de8fe7e3a3c4a1100a1de0842497f2286047e9",
      "tree": "539cbd4204dbef268c594cbe619b30376a647cba",
      "parents": [
        "45b6e3d6cdec442ce0e5d44b5bf2c91ba57fd017"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Wed Feb 05 13:28:27 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Feb 05 13:29:01 2025 -0800"
      },
      "message": "Update Eigen to commit:4c2611d27cb9a5fe12e18c8953097871db1d3abf\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n4c2611d27 - Update check for std::hardware_destructive_interference_size\nc079ee5e4 - Fix tensor documentation.\n74264c391 - Add missing return statements for ppc.\n3ebe898b5 - Build and deploy nightly docs.\n\nPiperOrigin-RevId: 723630092\nChange-Id: I4d429697d4943480a550b3fb1e4fad838116564a\n"
    },
    {
      "commit": "45b6e3d6cdec442ce0e5d44b5bf2c91ba57fd017",
      "tree": "18f9c05447b0a062b487da978c61aea5b5737b25",
      "parents": [
        "c4608705bc18561ade4111dcb7b48d09691d54a9"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Wed Feb 05 10:12:03 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Feb 05 10:12:35 2025 -0800"
      },
      "message": "Update Eigen to commit:b73bb766a57f872b4fe376c4818cd9b17dc1a761\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nb73bb766a - Increase max alignment to 256.\nb1e74b1cc - Fix all the doxygen warnings.\n9589cc4e7 - Fix loongarch64 emulated tests.\n2926b2e0a - added functions to fetch L and U Factors from IncompleteLUT\n\nPiperOrigin-RevId: 723556843\nChange-Id: Iee75792d75189f9f65b667b25348ca593d5ec121\n"
    },
    {
      "commit": "c4608705bc18561ade4111dcb7b48d09691d54a9",
      "tree": "be0f101d68fc07374740c84693773f8aa0dba881",
      "parents": [
        "0b49b85c5b29e467fab2abd5431d81c4c6a03125"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Jan 30 16:17:12 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Jan 30 16:17:58 2025 -0800"
      },
      "message": "Update Eigen to commit:b6849f675d4237398c4a3ae806ddaf4ca461ed01\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nb6849f675 - Change the midpoint chosen in Eigen::ForkJoinScheduler.\n1b2e84e55 - Fix minor typos in `ForkJoin.h`\n872c179f5 - Fix UTF-8 encoding errors introduced by #1801\n\nPiperOrigin-RevId: 721553862\nChange-Id: I44149f69a94bfdc8b3df291521bdf3effc69878e\n"
    },
    {
      "commit": "0b49b85c5b29e467fab2abd5431d81c4c6a03125",
      "tree": "8d9abf694411c0e72ea4a0de88856280b620d2dc",
      "parents": [
        "8d535dea750dfc9a623c6f74c35231a6f52b80ce"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Jan 29 13:14:34 2025 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Jan 29 13:15:16 2025 -0800"
      },
      "message": "Update Eigen to commit:2a35a917be47766a895be610bedd66006980b7e6\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2a35a917b - Fix syntax error in NonBlockingThreadPool.h\na056b9311 - improve Simplicial Cholesky analyzePattern\n5d866a7a7 - Fix potential data race on `spin_count_` NonBlockingThreadPool member variable\nbc67025ba - Clean up and fix the documentation of ForkJoin.h\ndc1126e76 - Fix threadpool for c++14.\ncd511a09a - Fix initialization order and remove unused variables in NonBlockingThreadPool.h.\nf679843dc - Block doc non square\nf9705adab - Fix typo introduced in the refactor of NonBlockingThreadPool\nb75895a8b - Try to fix loongarch\n4a6ac97d1 - Add a ForkJoin-based ParallelFor algorithm to the ThreadPool module\ne98683846 - Add LoongArch64 architecture LSX support.(build/test )\nc486af5ad - Change Eigen::aligned_allocator to not inherit from std::allocator\nabac563f5 - Update documentation to clarify cross product for complex numbers.\n2e76277bd - Zero-initialize test arrays to avoid uninitialized reads.\nad13df7ea - Fix std::fill_n reference.\n9836e8d03 - Fix read of uninitialized threshold in SparseQR\n7bb23b1e3 - CI: don\u0027\\\u0027\u0027t add ToolChain PPA\n7bb8c58e7 - Fix the missing CUDA device qualifier\n24e0c2a12 - use omp_get_max_threads if setNbThreads is not set\na32db4396 - Add missing `#include \u003cnew\u003e`\nc01ff4531 - Enable fill_n and memset optimizations for construction and assignment\naf59ada0a - Use alpine for deploying nightly tag.\n4a9e32ae0 - matrix equality operator\n00776d1ba - Remove branch name from nightly tag job.\n\nPiperOrigin-RevId: 721084675\nChange-Id: I7361407852d7898dcb174fe12fa4448812aedc04\n"
    },
    {
      "commit": "8d535dea750dfc9a623c6f74c35231a6f52b80ce",
      "tree": "0efea24d5ad12812cd9ecd1c981ec061089544dc",
      "parents": [
        "650354fb88d6e9f8e3c705534bb14e7a510f429f"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Wed Dec 18 20:43:19 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Dec 18 20:43:50 2024 -0800"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 707764415\nChange-Id: I74a76c6582de01b5a43351b5113b724a4008817f\n"
    },
    {
      "commit": "650354fb88d6e9f8e3c705534bb14e7a510f429f",
      "tree": "62f511d17da6654610402c4a815af2d825925a67",
      "parents": [
        "13b53a8016b75e9e102eeb6d091e11f725c624b7"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Dec 09 13:16:44 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Dec 09 13:17:21 2024 -0800"
      },
      "message": "Update Eigen to commit:7f2377859377da6f22152015c28b12c04752af77\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n7f2377859 - Add tag to commit instead of branch\nc30b35a31 - Force tag to update to latest head.\na26ba6734 - Add LICENSE file in correct place so it is picked up by gitlab.\n08c31c3ba - try alpine for formatting\n1ac1af62e - Update deploy job\n7b6623af3 - Fix special packetmath erfc flushing for ARM32.\nfd48fbb26 - Update rocm docker again again.\na885340ba - Update rocm docker again.\n45a8478d0 - Update rocm docker image in CI.\nde4afcf41 - Add a deploy phase to the CI that tags the latest nightly pipeline if it passes.\n5e8916050 - move constructor / move assignment doc strings\n\nPiperOrigin-RevId: 704393684\nChange-Id: Ic0d81e61335b2fcca8b0cce28ca1097e7a319cc5\n"
    },
    {
      "commit": "13b53a8016b75e9e102eeb6d091e11f725c624b7",
      "tree": "8af03240c108e1d66b65c87755375b302c81982d",
      "parents": [
        "27f0866278ac7acf5a3d119d9126c5a27ba0888f"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Dec 04 22:00:17 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Dec 04 22:01:00 2024 -0800"
      },
      "message": "Update Eigen to commit:77a073aaa821300fc129ef24b3b052be2e670a3e\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n77a073aaa - fix checkformat ci stage\n41e46ed24 - fix IOFormat alignment\na0d32e40d - fix map fill logic\n\nPiperOrigin-RevId: 702977029\nChange-Id: I04f6b6f1391e7005165977b32535f9d8b691e8fe\n"
    },
    {
      "commit": "27f0866278ac7acf5a3d119d9126c5a27ba0888f",
      "tree": "6a59865bfefff82b2863815d45a1f2f405c61e9e",
      "parents": [
        "e2dcd159fbe07c00b6c8e50616eedde1a6325eaa"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Nov 27 12:06:23 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Nov 27 12:06:55 2024 -0800"
      },
      "message": "Update Eigen to commit:d34b100c137ac931379ae5e1b888f16a9c8d6c72\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nd34b100c1 - Fix UB in setZero\n\nPiperOrigin-RevId: 700770909\nChange-Id: Ie2c8b90d462a0c47301f8fd0d013aeca222087bc\n"
    },
    {
      "commit": "e2dcd159fbe07c00b6c8e50616eedde1a6325eaa",
      "tree": "ae1b79ccd80df51d5b45b7323ce18ce1ef773b00",
      "parents": [
        "c9dc4f953ed0d763955fa6b0e33adf47a60c8f56"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Nov 27 09:49:54 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Nov 27 09:50:25 2024 -0800"
      },
      "message": "Update Eigen to commit:f19a6803c87cb8486c3147cd180edb02c9ff9a68\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nf19a6803c - Refactor special case handling in pow(x,y) and revert to repeated squaring for \u003cfloat,int\u003e\n5064cb7d5 - Add test for using pcast on scalars.\n1ea61a5d2 - Improve pow(x,y): 25% speedup, increase accuracy for integer exponents.\n8ad4344ca - optimize setConstant, setZero\n\nPiperOrigin-RevId: 700731159\nChange-Id: I762ac07c755af7d2e5d030e3a09291fedc604d1a\n"
    },
    {
      "commit": "c9dc4f953ed0d763955fa6b0e33adf47a60c8f56",
      "tree": "26be2814a83513f6f961c299c78d0ff9e4382629",
      "parents": [
        "b67c47efeb80cdabc78a2bc413659b2f2c59ef1a"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Nov 20 23:55:05 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Nov 20 23:55:35 2024 -0800"
      },
      "message": "Update Eigen to commit:5610a13b7767a94e063342277e0b9de4de13be57\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n5610a13b7 - Simplify and speed up pow() by 5-6%\n6c6ce9d06 - Enable vectorized erf\u003cdouble\u003e(x) for SSE and AVX, which was accidentally removed in merge request 1750.\n\nPiperOrigin-RevId: 698667782\nChange-Id: I9900a2805733da008786803884383647f84347b0\n"
    },
    {
      "commit": "b67c47efeb80cdabc78a2bc413659b2f2c59ef1a",
      "tree": "7a8104f63964a65e9c5538cfb7bca6e4490ccd7e",
      "parents": [
        "191bc95ae6eec66b0704a9baff7156bd93b5c4bf"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Nov 19 13:44:50 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Nov 19 13:45:28 2024 -0800"
      },
      "message": "Update Eigen to commit:e7c799b7c984f9b8bea27967bb04a97c52e62582\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\ne7c799b7c - Prevent premature overflow to infinity in exp(x). The changes also provide a 3-4% speedup.\n00af47102 - Revert https://gitlab.com/libeigen/eigen/-/commit/040180078db70b8673932d7e5615920d64ceeaf5\n8ee6f8475 - Speed up exp(x) by 30-35%.\n93ec5450c - disable fill_n optimization for msvc\n0af6ab4b7 - Remove unnecessary check for HasBlend trait.\nd5eec781b - Get rid of redundant computation for large arguments to erf(x).\n2fc63808e - Fix C++20 constexpr test compilation failures\n5133c836c - Vectorize erf(x) for double.\nd6e3b528b - Update Assign_MKL.h to cast disparate enum type to int, so it can be compared...\n040180078 - Ensure that destructor\u0027\\\u0027\u0027s needed by lldb make it into binary in non-inlined fashion\n0fb2ed140 - Make element accessors constexpr\n8b4efc8ed - check_size_for_overflow: use numeric limits instead of c99 macro\n\nPiperOrigin-RevId: 698131745\nChange-Id: Ia044ff8444a6d2266afdd7ad05556afd3be2366d\n"
    },
    {
      "commit": "191bc95ae6eec66b0704a9baff7156bd93b5c4bf",
      "tree": "874c64fc4ef4eab945521130ed0d0bb515b97e52",
      "parents": [
        "209760957eca6047b21726cadd43b3ffbdf4b65b"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Nov 13 14:08:03 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Nov 13 14:08:46 2024 -0800"
      },
      "message": "Update Eigen to commit:489dbbc6516f6729fa8e5c8bddc9ea446d5ef032\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n489dbbc65 - make fixed_size matrices conform to std::is_standard_layout\n283d871a3 - Add missing EIGEN_DEVICE_FUNCTION decorations.\n0d366f653 - Vectorize erfc(x) for double and improve erfc(x) for float.\n8adf43640 - more avx predux_any\nbc424f617 - add missing avx predux_any functions\ne52ac76ca - use EIGEN_CPLUSPLUS instead of checking cpp version\n122be167c - Revert \"make fixed-size objects trivially move assignable\"\nd49021212 - Tensor Roll / Circular Shift / Rotate\nbb73be8a2 - make fixed-size objects trivially move assignable\n\nPiperOrigin-RevId: 696264736\nChange-Id: I84aad26ecbe5fd7afee89965e0d6c6f05aca48e6\n"
    },
    {
      "commit": "209760957eca6047b21726cadd43b3ffbdf4b65b",
      "tree": "e6e17d1cadcaca4ef5c0ee1c87a941f8f078c6f0",
      "parents": [
        "e71e37af08060693a32ac4d0ca7c25b8a9073f4f"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Nov 05 13:56:10 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Nov 05 13:56:50 2024 -0800"
      },
      "message": "Update Eigen to commit:bb73be8a2e8da00c1cf5d26acb9fe47a87b8047f\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nbb73be8a2 - make fixed-size objects trivially move assignable\n\nPiperOrigin-RevId: 693468343\nChange-Id: Ic28e61d86a09b47f93ebc1edb70b370ca21c3061\n"
    },
    {
      "commit": "e71e37af08060693a32ac4d0ca7c25b8a9073f4f",
      "tree": "b6ad6030c5ad45d2490d1fe3237acd11bbb52283",
      "parents": [
        "cb436af996b7c2531dead7746905380bc216900f"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Nov 05 10:21:27 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Nov 05 10:22:05 2024 -0800"
      },
      "message": "Update Eigen to commit:bb73be8a2e8da00c1cf5d26acb9fe47a87b8047f\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nbb73be8a2 - make fixed-size objects trivially move assignable\n\nPiperOrigin-RevId: 693394913\nChange-Id: Id039c61e4f79f73a7c80787606f9789e20825f41\n"
    },
    {
      "commit": "cb436af996b7c2531dead7746905380bc216900f",
      "tree": "e6e17d1cadcaca4ef5c0ee1c87a941f8f078c6f0",
      "parents": [
        "eae6db0ca08e15a56faa194349561649dde541dd"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Nov 04 11:22:06 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Nov 04 11:22:40 2024 -0800"
      },
      "message": "Update Eigen to commit:7fd305ecae2410714cde018cb6851f49138568c8\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n7fd305eca - Fix GPU builds.\nc8267654f - Don\u0027t use __builtin_alloca_with_align with nvc++\n84c446df2 - Fix macro redefinition warning in FFTW test\na9584d8e3 - Fix clang6 failures.\ndd4c2805d - Fix clang6 failures.\n9e962d9c5 - Fix OOB access in triangular matrix multiplication.\n695e49d1b - Fix NVCC builds for CUDA 10+.\ndae09773f - Don\u0027t pass matrices by value.\nc23ec3420 - Add tests for sizeof() with one dynamic dimension.\n58b252e5b - Fix typo in PacketMath.h\n6c04d0cd6 - Add missing exp2 definition for Altivec.\nb15ebb1c2 - add nextafter for bfloat16\n53b83cddf - Include \u003ctype_traits\u003e in main.h for std::is_trivial*\n37563856c - Fix stack allocation assert\n3f067c485 - Add exp2() as a packet op and array method.\n4e5136d23 - make fixed size matrices and arrays trivially_default_constructible\n\nPiperOrigin-RevId: 693033345\nChange-Id: I9316de3727feb01f71e21e186576a4fa1a14e06d\n"
    },
    {
      "commit": "eae6db0ca08e15a56faa194349561649dde541dd",
      "tree": "e97cf23539c0e5f3f4e74dc8e52a351f5715964c",
      "parents": [
        "db3836c322b66a38eacde8400509ad1e1025f669"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Oct 15 10:21:29 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Oct 15 10:22:12 2024 -0700"
      },
      "message": "Update Eigen to commit:b396a6fbb2e173f52edb3360485dedf3389ef830\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nb396a6fbb - Add free-function swap.\n820e8a45f - add compile time info to reverse in place\nb55dab7f2 - Fix DenseBase::tail for Dynamic template argument\ne0cbc55d9 - Update README.md\n\nPiperOrigin-RevId: 686151157\nChange-Id: Ib5c5a685c6277a21aa6fd75f5468de4c42461ddd\n"
    },
    {
      "commit": "db3836c322b66a38eacde8400509ad1e1025f669",
      "tree": "39141f32a8b66473f27fab3276f773ae78128f92",
      "parents": [
        "a75b2b0c7974572ce274e8207ba5161001a0b87d"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Oct 10 11:42:30 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Oct 10 11:42:59 2024 -0700"
      },
      "message": "Update Eigen to commit:7eea0a9213e801ad9479a6499fd0330ec1db8693\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n7eea0a921 - Vectorize erfc() for float\n\nPiperOrigin-RevId: 684520203\nChange-Id: Iff78fdb6f65b8a2720f0b51d67f560445c0e42bf\n"
    },
    {
      "commit": "a75b2b0c7974572ce274e8207ba5161001a0b87d",
      "tree": "e8987e6f2610b524391652ed214f253b4a65553f",
      "parents": [
        "9a6317647091156dd66e61fcdaec866be67f327d"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Oct 09 12:35:25 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Oct 09 12:36:12 2024 -0700"
      },
      "message": "Update Eigen to commit:78f3c654eef75ff15cb27c0816f702e4828a11ad\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n78f3c654e - Don\u0027\\\u0027\u0027t use constexpr with half.\n6d7af238f - Adjust array_cwise for 32-bit arm.\n74dcfbbd0 - Use ppolevl for polynomial evaluation in more places.\na097f728f - Avoid producing erf(x) \u003d NaN for large |x|.\n44b16f48c - Improve speed and accuracy or erf()\n12068cbcd - Fix inverse evaluator for running on CUDA device.\n4e8e5e740 - Add max_digits10 in NumTraits for mpreal types.\n8e8c31908 - Add missing EIGEN_DEVICE_FUNC annotations.\n7ad7c1d5c - fix implicit conversion warning (again)\nd052b7f86 - add extra debugging info to float_pow_test_impl, clean up array_cwise tests\nba5183f98 - fix warning in EigenSolver::pseudoEigenvalueMatrix()\n3ffb4e50d - fix implicit conversion in TensorChipping\nb6b8b54e5 - Fixed issue #2858: removed unneeded call to _mm_setzero_si128\n2a3465102 - Refactor code to use constexpr for data() functions.\n\nPiperOrigin-RevId: 684133454\nChange-Id: I0ff7c2f21ba567663489e36035a1eb726c64c1e5\n"
    },
    {
      "commit": "9a6317647091156dd66e61fcdaec866be67f327d",
      "tree": "a35fde36c1d5b23664d2b6e63329d37bc3d4429b",
      "parents": [
        "7e3951f60de2e468d1d51edda265497155b48a3d"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Oct 01 10:38:56 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Oct 01 10:40:20 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 681078766\nChange-Id: I2d62f47fc9796c0785b23c6a0933501f58c273bc\n"
    },
    {
      "commit": "7e3951f60de2e468d1d51edda265497155b48a3d",
      "tree": "b3482a13b1f6e7eac3b184f88ed1b42dab4ed499",
      "parents": [
        "2601a6cf4d62562240ab8d1cb795273b178d672b"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Sep 30 10:11:46 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Sep 30 10:12:25 2024 -0700"
      },
      "message": "Update Eigen to commit:2d4c9b400cca33d2f5cf316efc7151236244edb1\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2d4c9b400 - make fixed size matrices and arrays trivially_copy_constructible and trivially_move_constructible\n132f281f5 - Fix generic ceil for SSE2.\n84282c42f - optimize new dot product\nfb477b8be - Better dot products\n134b526d6 - Update NonBlockingThreadPool.h plain asserts to use eigen_plain_assert\n072ec9d95 - Fix a bug for pcmp_lt_or_nan and Add sqrt support for SVE\n931538979 - Fix bug in bug fix for atanh.\n\nPiperOrigin-RevId: 680623801\nChange-Id: I0493caa2a8dd2b403773c440cbe312fbade0b50e\n"
    },
    {
      "commit": "2601a6cf4d62562240ab8d1cb795273b178d672b",
      "tree": "f622182658d45601c19fb5b589773d9529ad14a4",
      "parents": [
        "bc745443ad61a46b67cebc6bff9b7ef0ecd6c399"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Sep 03 15:00:36 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Sep 03 15:01:10 2024 -0700"
      },
      "message": "Update Eigen to commit:f33af052e0e60d4aa367328e7d9dffc9dedca6d8\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nf33af052e - Fix bug for atanh(-1).\n66927f780 - Fix out-of-range arguments to _mm_permute_pd.\nbbdabebf4 - Vectorize atanh\u003cdouble\u003e. Make atanh(x) standard compliant for |x| \u003e\u003d 1.\n26e2c4f61 - Add nvc++ support\nc59332d74 - Detect \"effectively inner/outer\" chipping in TensorChipping\n648bce6ca - SSE/AVX Complex FMA\nc21a80be3 - BDCSVD: Suppress Wmaybe-uninitialized\n9d3d37c5b - Complex Numtraits::HasSign and nmsub test\nc5189ac65 - Fix GeneralizedEigenSolver::eigenvectors() not appearing in documentation\n3b5a1b415 - sve instrinsics with \"_x\" suffix will be faster than \"_z\" suffix\n98f1ac5e6 - Fix breakage in GPU build.\n231308f69 - TensorVolumePatchOp: Suppress Wmaybe-uninitialized caused by unreachable code\n\nPiperOrigin-RevId: 670718238\nChange-Id: I313776ca7eadd10d3deb6006c80e732eb0ff1fdf\n"
    },
    {
      "commit": "bc745443ad61a46b67cebc6bff9b7ef0ecd6c399",
      "tree": "7c189ff9641eefd2a94dfe08d7d44fcb77d80a8c",
      "parents": [
        "8b9a5c24a2d0966e756511a73cd7cf944bb546d7"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Aug 29 12:00:24 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Aug 29 12:01:00 2024 -0700"
      },
      "message": "Update Eigen to commit:2bf8fe1489102bad4be90ac3c8397a63b8dd398f\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2bf8fe148 - NEON Complex Intrinsics\nf91f8e9ab - Consolidate float and double implementations of patan().\n87239e058 - vectorize squaredNorm() for complex types\n32d95bb09 - Add vectorized implementation of tanh\u003cdouble\u003e\ncc240eea2 - Speed up and improve accuracy of tanh.\n\nPiperOrigin-RevId: 669015363\nChange-Id: I9cbbb78b3ce2b439571fc925beb702f69a96ac00\n"
    },
    {
      "commit": "8b9a5c24a2d0966e756511a73cd7cf944bb546d7",
      "tree": "3805987301c77d349dc31d6b00ae917d353b39ab",
      "parents": [
        "ee287b2673e4098cb0d298b222f8cbeb576dbf0c"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Mon Aug 26 11:25:50 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Aug 26 11:26:25 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 667652516\nChange-Id: I63b15c0d1c298068a10f916d860f38517ef12af3\n"
    },
    {
      "commit": "ee287b2673e4098cb0d298b222f8cbeb576dbf0c",
      "tree": "8f7fa607d25ecfc6ae41d52c3f544a02d77b0b72",
      "parents": [
        "ca0ee8864adfe0ab1225b549299604aba47a3713"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu Aug 15 22:36:49 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Aug 15 22:37:22 2024 -0700"
      },
      "message": "Update Eigen to commit:92e373e6f553dd842337c8467f4ddb669e0e9199\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n92e373e6f - Speed up StableNorm for non-trivial sizes and improve consistency between aligned and unaligned inputs.\n\nPiperOrigin-RevId: 663605055\nChange-Id: Ifa7a1d92774470b421c87722a96c57949e49ac4c\n"
    },
    {
      "commit": "ca0ee8864adfe0ab1225b549299604aba47a3713",
      "tree": "84804a6b17a97a6037d8d018f052c5d1774b0feb",
      "parents": [
        "cd3b0ab2e985e5fe26b4ca4a94be12ba76687e76"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Aug 14 11:18:07 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Aug 14 11:19:10 2024 -0700"
      },
      "message": "Update Eigen to commit:ab310943d62982a45734f52fed782ec57c9aa6a0\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nab310943d - Add a yield instruction in the two spinloops of the threaded matmul implementation.\n99ffad197 - A few cleanups to threaded product code and test.\n59498c96f - SSE/AVX use fmaddsub for complex products\n\nPiperOrigin-RevId: 662989342\nChange-Id: I5ca079b01b427f8ba6a32dc6b2684eeda7ac1fd5\n"
    },
    {
      "commit": "cd3b0ab2e985e5fe26b4ca4a94be12ba76687e76",
      "tree": "4dc5ae68940317521b5cc0bb2b879a112c79a389",
      "parents": [
        "84ee861c84d73aea0fb6bdfad491161e1646cbb6"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Aug 05 14:08:34 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Aug 05 14:09:16 2024 -0700"
      },
      "message": "Update Eigen to commit:1dcae7cefcd479ec1f881dd4831665ab292ee1e7\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n1dcae7cef - Revert \"BDCSVD fix -Wmaybe-uninitialized\"\nd14b0a4e5 - Remove C++23 check around has_denorm deprecation suppression\n24db46050 - hlog symbol lookup should not restricted to global namespace\n767e60e29 - Fix Woverflow warnings in PacketMathFP16\n802568322 - Fix conversion of `Eigen::half` to `_Float16` in AVX512 code\nec18dd09c - fix pi in kissfft\n2b7b7aac5 - Speed up complex * complex matrix multiplication.\nb3e3b7b0e - Remove implicit this capture in lambdas\ne44db2109 - Optimize ThreadPool spinning\nc593e9e94 - Fix typos\nfd98cc49f - Avoid atomic false sharing in RunQueue\n0b646f3f3 - Update file .clang-format\n1dcb07bb2 - Update file eigen_navtree_hacks.js\nddb163ffb - Update file .clang-format\n3f06651fd - BDCSVD fix -Wmaybe-uninitialized\n6331da95e - fixing a lot of typos\nc29c80012 - Fix formatting in README.md\n\nPiperOrigin-RevId: 659676252\nChange-Id: I700b8fc2e3899bed3cca0dbf0b0c67f3b61fb8a0\n"
    },
    {
      "commit": "84ee861c84d73aea0fb6bdfad491161e1646cbb6",
      "tree": "93af184756a869beccbd879766745b2e743529c9",
      "parents": [
        "2571273647cdf6673639ebe9b3aa3237c1e7a822"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Mon Aug 05 12:16:40 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Aug 05 12:17:15 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 659640141\nChange-Id: I5b3579480d96954f871f1706668811a5adeb001f\n"
    },
    {
      "commit": "2571273647cdf6673639ebe9b3aa3237c1e7a822",
      "tree": "ed54f96b0dd3b69fd25b498d0ea866b99fe64db6",
      "parents": [
        "7cb6aea553cad957edce4dc7e72389dc0e8b684b"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Thu Jul 25 16:41:19 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Jul 25 16:41:51 2024 -0700"
      },
      "message": "[tsl] Remove platform:types from mutex dependencies\n\nPiperOrigin-RevId: 656140625\nChange-Id: Iaa124ee08a8519b0aeaeaf304c2d80908f854eab\n"
    },
    {
      "commit": "7cb6aea553cad957edce4dc7e72389dc0e8b684b",
      "tree": "1c9b8262099d7d63a1c012b750ed8656e0621cc2",
      "parents": [
        "5757c099abbb7d8c3797986f947915f335245cf7"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Thu Jul 25 14:47:15 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Jul 25 14:47:54 2024 -0700"
      },
      "message": "[tsl] Remove platform:types from mutex dependencies\n\nPiperOrigin-RevId: 656100657\nChange-Id: I406d7afa795d308992d8b7ba7e1068039ed06972\n"
    },
    {
      "commit": "5757c099abbb7d8c3797986f947915f335245cf7",
      "tree": "ed54f96b0dd3b69fd25b498d0ea866b99fe64db6",
      "parents": [
        "e8f5b20687e5dcf99b900a8f91a472273f9006f8"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Wed Jul 24 19:24:09 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Jul 24 19:24:42 2024 -0700"
      },
      "message": "[tsl] Remove dependency on platform:types from platform:mutex\n\nPiperOrigin-RevId: 655781413\nChange-Id: If6169c4bdd60edea8a5452f00eedb1cd1d70de29\n"
    },
    {
      "commit": "e8f5b20687e5dcf99b900a8f91a472273f9006f8",
      "tree": "e59e5dd0c35f621931ec3eb34e3b44a74f2d84e4",
      "parents": [
        "6bad1ea30a0871dbe7ee7dff2e8e4ea853b9d781"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Mon Jul 22 10:25:16 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Jul 22 10:25:55 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 654807362\nChange-Id: Idfc14428e292550424db2436e38b3b30ce6a6ac6\n"
    },
    {
      "commit": "6bad1ea30a0871dbe7ee7dff2e8e4ea853b9d781",
      "tree": "70cabf8780bfbcfacef9b29e424aaf85411dafb6",
      "parents": [
        "2bbd4a870ad457f7745d83d77f82cfc135ab4ea4"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Sat Jul 20 15:33:34 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Sat Jul 20 15:34:35 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 654364452\nChange-Id: I7189202a6135d5504068bbacd24be42adb689f7d\n"
    },
    {
      "commit": "2bbd4a870ad457f7745d83d77f82cfc135ab4ea4",
      "tree": "45c6f804388afd542508dc02dc1ae5d802da06f6",
      "parents": [
        "3cb0cd4d07a52122b1e4ee8cd906945794c577d8"
      ],
      "author": {
        "name": "Eugene Zhulenev",
        "email": "ezhulenev@google.com",
        "time": "Sat Jul 20 15:33:02 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Sat Jul 20 15:33:29 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 654364383\nChange-Id: I4c964c0eb6ee2518f5211e0638a559a2ecb41bd8\n"
    },
    {
      "commit": "3cb0cd4d07a52122b1e4ee8cd906945794c577d8",
      "tree": "edcd4eebef84ad9a5982a8c49e3e822afc22fdcc",
      "parents": [
        "14a5b325090e92735def50002492aee46e4e47fc"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Jul 03 14:20:02 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Jul 03 14:20:32 2024 -0700"
      },
      "message": "Update Eigen to commit:33d0937c6bdf5ec999939fb17f2a553183d14a74\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n33d0937c6 - Add async support for \u0027chip\u0027 and \u0027extract_volume_patches\u0027\n\nPiperOrigin-RevId: 649197338\nChange-Id: Ib7bb58b482c3554553bfa116fefe4ab64bfe7e4e\n"
    },
    {
      "commit": "14a5b325090e92735def50002492aee46e4e47fc",
      "tree": "ff4abae35c35ca48e6fa0c5b198d720be47aa87f",
      "parents": [
        "0c691a3fff584bf93f87f7f83fb457a19c361ffa"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Jun 21 12:28:06 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Jun 21 12:29:34 2024 -0700"
      },
      "message": "Update Eigen to commit:d791d48859c6fc7850c9fd5270d2b236c818068d\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nd791d4885 - Fix AVX512FP16 build failure\n2fae4d7a7 - Revert \"fix scalar pselect\"\nb430eb31e - AVX512F double-\u003eint64_t cast\n02bcf9b59 - fix scalar pselect\n392b95bdf - allow pointer_based_stl_iterator to conform to the contiguous_iterator concept if we are in c++20\n27f817625 - fixing warning C5054: operator \u0027\u003d\u003d\u0027: deprecated between enumerations of different types\n\nPiperOrigin-RevId: 645467923\nChange-Id: Icf1fd74b44ed4aea311a3d57854c5d5814e89e42\n"
    },
    {
      "commit": "0c691a3fff584bf93f87f7f83fb457a19c361ffa",
      "tree": "19589c0ba0a51b1c5d76b1788011ebe619ccdef7",
      "parents": [
        "9dfc0afdf1dbe8217c8421c061834d195ae40fae"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Mon Jun 03 13:32:25 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Mon Jun 03 13:32:56 2024 -0700"
      },
      "message": "Update Eigen to commit:eac6355df2fb3e306a3078a236ddc6d9f19fba03\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\neac6355df - Fix warnings created by other warnings fix\n7029a2e97 - Vectorize allFinite()\ne60522703 - Fix warnings\n\nPiperOrigin-RevId: 639894674\nChange-Id: I8214bac2a26c6ce4371b8ba1e540121ff46ad904\n"
    },
    {
      "commit": "9dfc0afdf1dbe8217c8421c061834d195ae40fae",
      "tree": "5da7ea8f023bd71a60f14e9fafe83ead12d5c4a2",
      "parents": [
        "440a71a78fb63be4f4e46dbf757a49ccf87c2aab"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu May 30 11:52:37 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu May 30 11:53:15 2024 -0700"
      },
      "message": "Update Eigen to commit:38b9cc263bbaeb03ce408a4e26084543a6c0dedb\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n38b9cc263 - Fix warnings about repeated deinitions of macros.\nf02f89bf2 - Don\u0027t redefine EIGEN_DEFAULT_IO_FORMAT in main.h.\n9148c47d6 - Vectorize isfinite and isinf.\n5a9f66fb3 - Fix Thread tests\nc4d84dfdd - Fix compilation failures on constexpr matrices with GCC 14\n99adca8b3 - Incorporate Threadpool in Eigen Core\nd165c7377 - Format EIGEN_STATIC_ASSERT() as a statement macro\nf78dfe36b - use built in alloca with align if available\nb9b1c8661 - Suppress C++23 deprecation warnings for std::has_denorm and std::has_denorm_loss\n3d2e738f2 - fix performance-no-int-to-ptr\nde8013fa6 - Fix ubsan failure in array_for_matrix\n5e4f3475b - Remove call to deprecated method initParallel() in SparseDenseProduct.h\n59cf0df1d - SparseMatrix::insert add checks for valid indices\nc0fe6ce22 - Fixed a clerical error at documentation of class Matrix.\nafb17288c - Fix gcc6 compile error.\n4d1d14e06 - Change predux on PowerPC for Packet4i to NOT saturate the sum of the elements (like other architectures).\nff174f792 - fix issue: cmake package does not set include path correctly\n\nPiperOrigin-RevId: 638732445\nChange-Id: I550ab37d222c05c774ae2959fe6b7a6bbb61e01a\n"
    },
    {
      "commit": "440a71a78fb63be4f4e46dbf757a49ccf87c2aab",
      "tree": "c39c249ce480f4331f2cbbf4f02c90f7f33b0f9b",
      "parents": [
        "ea8b62b13e508decf55995661435d8d059feb70c"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed May 08 17:25:30 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed May 08 17:26:07 2024 -0700"
      },
      "message": "Update Eigen to commit:e16d70bd4e9cdebd2fbdae63b1a4d86493fbbde6\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\ne16d70bd4 - Fix FFT when destination does not have unit stride.\n99c18bce6 - Msvc muluh\n8e4797178 - Bit shifting functions\n9700fc847 - Reorganize CMake and minimize configuration for non-top-level builds.\n\nPiperOrigin-RevId: 631975359\nChange-Id: I537052ef6b9ed6e3ecf90e2aebf5c3898ffc919c\n"
    },
    {
      "commit": "ea8b62b13e508decf55995661435d8d059feb70c",
      "tree": "bb35c566adedf0c04b291ebb26c55ec909293b8e",
      "parents": [
        "ad3071076bc0f1d4156e58f29187d28f30eb85b7"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Thu May 02 15:13:03 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu May 02 15:13:48 2024 -0700"
      },
      "message": "Update Eigen to commit:c1d637433e3b3f9012b226c2c9125c494b470ae6\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nc1d637433 - Judge unitary-ness relative to scaling.\n9000b3767 - Fix new generic nearest integer ops on GPU.\n0ee5c90aa - Eigen transpose product\nfb95e90f7 - Add truncation op\nd5524fc57 - Remove unnecessary semicolons.\nae5280aa8 - Fix more hard-coded magic bounds.\na5e147305 - Fix undefined behavior for generating inputs to the predux_mul test.\ndcceb9afe - Unbork avx512 preduce_mul on MSVC.\n\nPiperOrigin-RevId: 630192447\nChange-Id: Ia1ed5f493b17bb3e8e5d0df50b5d2a966c6bbdd3\n"
    },
    {
      "commit": "ad3071076bc0f1d4156e58f29187d28f30eb85b7",
      "tree": "9beb3367704f8fed00896e16278ce1118cd360b2",
      "parents": [
        "a0bee8690129638e46876e324e2ac5793f7e6eee"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Apr 26 12:08:57 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Apr 26 12:09:31 2024 -0700"
      },
      "message": "Update Eigen to commit:42aa3d17cd27ecdfa34b81100314749f72c00256\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n42aa3d17c - Slightly adjust error bound for nonlinear tests.\n1c8c734c8 - Fix sin/cos on PPC.\n34967b0b5 - Revert \"fix transposed matrix product bug\"\n9cec679ef - Don\u0027t let the PPC runner try to cross-compile.\n574bc8820 - fix transposed matrix product bug\n112ad8b84 - Revert part of !1583, which may cause underflow on ARM.\n8cafbc473 - Fix unused variable warnings in TensorIO\n4de870b6e - fix autodiff enum comparison warnings\n2265242aa - Update CI scripts.\nee9d57347 - Fix `tridiagonalization_inplace_selector::run()` when called from CUDA\n1550c9954 - Eigen select\n\nPiperOrigin-RevId: 628474184\nChange-Id: I9f5a69fd4ad6d7c65a5deccfc669047656d84536\n"
    },
    {
      "commit": "a0bee8690129638e46876e324e2ac5793f7e6eee",
      "tree": "30fa19bcda4ac233ad141f542d0d0eb8b570bda1",
      "parents": [
        "ce8159acabc40467a6644f263ba99185fcbf52af"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Apr 19 11:34:11 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Apr 19 11:34:39 2024 -0700"
      },
      "message": "Update Eigen to commit:5635d37f46acc2985aa66d9385665a76c3b9d5c7\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n5635d37f4 - more pblend optimizations\nf0795d35e - Fix new psincos for ppc and arm32.\nad452e575 - Fix compilation problems with PacketI on PowerPC.\nfcaf03ef7 - fix pendantic compiler warnings\nb5feca5d0 - Fix build for pblend and psin_double, pcos_double when AVX but not AVX2 is supported.\n888fca0e2 - Simd sincos double\n6ad2ccea4 - Eigen pblend\n9099c5eac - Handle missing AVX512 intrinsic\n\nPiperOrigin-RevId: 626425279\nChange-Id: I008ca76bc6c357da4c37c0adfa654b053ac9c18b\n"
    },
    {
      "commit": "ce8159acabc40467a6644f263ba99185fcbf52af",
      "tree": "31ebfb2ab1e435a417abc5585ddad9f1c92432aa",
      "parents": [
        "0403750c06a3bb328076182a6bab6761e23d3265"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Apr 16 10:32:08 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Apr 16 10:32:47 2024 -0700"
      },
      "message": "Update Eigen to commit:122befe54cc0c31273d9e1caef80b49ad834bf4c\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n122befe54 - Fix \"unary minus operator applied to unsigned type, result still unsigned\" on MSVC and other stupid warnings\ndcdb0233c - Refactor indexed view to appease MSVC 14.16.\n5226566a1 - Speed up pldexp_generic.\n3c6521ed9 - Add constexpr to accessors in DenseBase, Quaternions and Translations\n3c9109238 - Add support for Packet8l to AVX512.\n\nPiperOrigin-RevId: 625378611\nChange-Id: Ib21f5d585439de99cba53737d4ceaebe323f5cbe\n"
    },
    {
      "commit": "0403750c06a3bb328076182a6bab6761e23d3265",
      "tree": "438d003173b8ca2ffec2bbcf02188401be3d1020",
      "parents": [
        "bf93d0dc1943f035ac7ea32780d87b23e5492da5"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Tue Apr 09 13:58:59 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Apr 09 13:59:37 2024 -0700"
      },
      "message": "Update Eigen to commit:2620cb930b7ad87ed1d77a26319739f4b1c86d33\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2620cb930 - Update file Geometry_SIMD.h\nb2c9ba2be - Fix preprocessor condition on when to use fast float logistic implementation.\n283d69294 - Guard AVX2 implementation of psignbit in  PacketMath.h\nbe54cc8de - Fix preverse for PowerPC.\nc5b234196 - Fix unused variable warning in TensorIO.h\n86aee3d9c - Fix long double random\n776d86d8d - AVX: guard Packet4l definition\ne63d9f6cc - Fix random again\nf75e2297d - AVX2 - double-\u003eint64_t casting\n13092b5d0 - Fix usages of Eigen::array to be compatible with std::array.\n77833f932 - Allow symbols to be used in compile-time expressions.\nd26e19714 - Add missing cwiseSquare, tests for cwise matrix ops.\n35bf6c8ed - Add SimplicialNonHermitianLLT and SimplicialNonHermitianLDLT\n4dccaa587 - Use truncation rather than rounding when casting Packet2d to Packet2l.\n7b5d32b7c - Sparse move\nc8d368bda - More fixes for 32-bit.\nde304ab96 - Fix using ScalarPrinter redefinition for gcc.\nc54303848 - Undef macro in TensorContractionGpu.h that causes buildbreakages.\nd8aa4d6ba - Fix another instance of Packet2l on win32.\n9f77ce4f1 - Add custom formatting of complex numbers for Numpy/Native.\n5570a2786 - cross3_product vectorization\n0b3df4a6e - Remove \"extern C\" in CholmodSupport.\na39ade4cc - Protect use of alloca.\nb86641a4c - Add support for casting between double and int64_t for SSE and AVX2.\nd88393258 - Fix Packet*l for 32-bit builds.\nd792f13a6 - Make more Matrix functions constexpr\nd3cd31265 - Remove slow index check in Tensor::resize from release mode.\n386e2079e - Fix Jacobi module doc.\n8b101ade2 - Fix CwiseUnaryView for MSVC.\n0951ad2a8 - Don\u0027t hide rbegin/rend for GPU.\n24f8fdeb4 - Fix CwiseUnaryView const access (Attempt 2).\n285da30ec - Fix const input and c++20 compatibility in unary view.\n126ba1a16 - Add Packet2l for SSE.\n1d4369c2f - Fix CwiseUnaryView.\n352ede96e - Fix incomplete cholesky.\nf1adb0ccc - Split up cxx11_tensor_gpu to reduce timeouts.\n17f3bf898 - Fix pexp test for ARM.\n6da34d9d9 - Allow aligned assignment in TRMV.\n3e8e63eb4 - Fix packetmath plog test on Windows.\n5ffb307af - Fix deprecated anonymous enum-enum conversion warnings\n55dd48747 - Revert \"fix unaligned access in trmv\"\n38fcedaf8 - Fix pexp complex test edge-cases.\n251ec4208 - Return 0 volume for empty AlignedBox\n64edfbed0 - Fix static_assert for c++14.\n3f3144f53 - fix unaligned access in trmv\n23f6c2685 - Rip out make_coherent, add CoherentPadOp.\nedaf9e16b - Fix triangular matrix-vector multiply uninitialized warning.\n98620b58c - Eliminate FindCUDA cmake warning.\ncc941d69a - Update error about c++14 requirement.\n6893287c9 - Add degenerate checks before calling BLAS routines.\nfa201f1bb - Fix QR colpivoting warnings and test failure.\nb33491070 - delete shadowed typedefs\na962a2759 - Fix MSVC GPU build.\na2f8eba02 - Speed up sparse x dense dot product.\n7a88cdd6a - Fix signed integer UB in random.\na6dc930d1 - Speed up SparseQR.\nfeaafda30 - Change array_size result from enum to constexpr.\n8a73c6490 - Remove \"using namespace Eigen\" from blas/common.h.\n6ed4d80cc - Fix crash in IncompleteCholesky when the input has zeros on the diagonal.\n3859e8d5b - Add method signDeterminant() to QR and related decompositions.\ndb6b9db33 - Make header guards in GeneralMatrixMatrix.h and Parallelizer.h consistent:...\nb56e30841 - Enable direct access for IndexedView.\n90087b990 - Fix use of uninitialized memory in kronecker_product test.\n6b365e74d - Fix GPU build for ptanh_float.\nb14c5d0fa - Fix real schur and polynomial solver.\n8a4118746 - fix exp complex test: use int instead of index\n960892ca1 - JacobiSVD: get rid of m_scaledMatrix, m_adjoint, hopefully fix some compiler warnings\n18a161bf1 - fix pexp_complex_test\nbe06c9ad5 - Implement float pexp_complex\n4d419e220 - Rename generic_fast_tanh_float to ptanh_float and move it to...\n\nPiperOrigin-RevId: 623273359\nChange-Id: I8711925b6cac355eaac66253080b16a637ae5bdc\n"
    },
    {
      "commit": "bf93d0dc1943f035ac7ea32780d87b23e5492da5",
      "tree": "5226b6207d50c4abdd9b143d8187be75991adfb0",
      "parents": [
        "941ca8d83f776b9a07153d3abef2877907aa0555"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Apr 05 18:15:52 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Apr 05 18:16:30 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 622331418\nChange-Id: I0d1bdfeece05539a90f17f1aa16f1889fdb87bdf\n"
    },
    {
      "commit": "941ca8d83f776b9a07153d3abef2877907aa0555",
      "tree": "438d003173b8ca2ffec2bbcf02188401be3d1020",
      "parents": [
        "f6c457cdc27f232d7cf7f4f9b801aa8c3f9f18c6"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Apr 05 13:30:15 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Apr 05 13:32:20 2024 -0700"
      },
      "message": "Update Eigen to commit:2620cb930b7ad87ed1d77a26319739f4b1c86d33\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2620cb930 - Update file Geometry_SIMD.h\nb2c9ba2be - Fix preprocessor condition on when to use fast float logistic implementation.\n283d69294 - Guard AVX2 implementation of psignbit in  PacketMath.h\nbe54cc8de - Fix preverse for PowerPC.\nc5b234196 - Fix unused variable warning in TensorIO.h\n86aee3d9c - Fix long double random\n776d86d8d - AVX: guard Packet4l definition\ne63d9f6cc - Fix random again\nf75e2297d - AVX2 - double-\u003eint64_t casting\n13092b5d0 - Fix usages of Eigen::array to be compatible with std::array.\n77833f932 - Allow symbols to be used in compile-time expressions.\nd26e19714 - Add missing cwiseSquare, tests for cwise matrix ops.\n35bf6c8ed - Add SimplicialNonHermitianLLT and SimplicialNonHermitianLDLT\n4dccaa587 - Use truncation rather than rounding when casting Packet2d to Packet2l.\n7b5d32b7c - Sparse move\nc8d368bda - More fixes for 32-bit.\nde304ab96 - Fix using ScalarPrinter redefinition for gcc.\nc54303848 - Undef macro in TensorContractionGpu.h that causes buildbreakages.\nd8aa4d6ba - Fix another instance of Packet2l on win32.\n9f77ce4f1 - Add custom formatting of complex numbers for Numpy/Native.\n5570a2786 - cross3_product vectorization\n0b3df4a6e - Remove \"extern C\" in CholmodSupport.\na39ade4cc - Protect use of alloca.\nb86641a4c - Add support for casting between double and int64_t for SSE and AVX2.\nd88393258 - Fix Packet*l for 32-bit builds.\nd792f13a6 - Make more Matrix functions constexpr\nd3cd31265 - Remove slow index check in Tensor::resize from release mode.\n386e2079e - Fix Jacobi module doc.\n8b101ade2 - Fix CwiseUnaryView for MSVC.\n0951ad2a8 - Don\u0027\\\u0027\u0027t hide rbegin/rend for GPU.\n24f8fdeb4 - Fix CwiseUnaryView const access (Attempt 2).\n285da30ec - Fix const input and c++20 compatibility in unary view.\n126ba1a16 - Add Packet2l for SSE.\n1d4369c2f - Fix CwiseUnaryView.\n352ede96e - Fix incomplete cholesky.\nf1adb0ccc - Split up cxx11_tensor_gpu to reduce timeouts.\n17f3bf898 - Fix pexp test for ARM.\n6da34d9d9 - Allow aligned assignment in TRMV.\n3e8e63eb4 - Fix packetmath plog test on Windows.\n5ffb307af - Fix deprecated anonymous enum-enum conversion warnings\n55dd48747 - Revert \"fix unaligned access in trmv\"\n38fcedaf8 - Fix pexp complex test edge-cases.\n251ec4208 - Return 0 volume for empty AlignedBox\n64edfbed0 - Fix static_assert for c++14.\n3f3144f53 - fix unaligned access in trmv\n23f6c2685 - Rip out make_coherent, add CoherentPadOp.\nedaf9e16b - Fix triangular matrix-vector multiply uninitialized warning.\n98620b58c - Eliminate FindCUDA cmake warning.\ncc941d69a - Update error about c++14 requirement.\n6893287c9 - Add degenerate checks before calling BLAS routines.\nfa201f1bb - Fix QR colpivoting warnings and test failure.\nb33491070 - delete shadowed typedefs\na962a2759 - Fix MSVC GPU build.\na2f8eba02 - Speed up sparse x dense dot product.\n7a88cdd6a - Fix signed integer UB in random.\na6dc930d1 - Speed up SparseQR.\nfeaafda30 - Change array_size result from enum to constexpr.\n8a73c6490 - Remove \"using namespace Eigen\" from blas/common.h.\n6ed4d80cc - Fix crash in IncompleteCholesky when the input has zeros on the diagonal.\n3859e8d5b - Add method signDeterminant() to QR and related decompositions.\ndb6b9db33 - Make header guards in GeneralMatrixMatrix.h and Parallelizer.h consistent:...\nb56e30841 - Enable direct access for IndexedView.\n90087b990 - Fix use of uninitialized memory in kronecker_product test.\n6b365e74d - Fix GPU build for ptanh_float.\nb14c5d0fa - Fix real schur and polynomial solver.\n8a4118746 - fix exp complex test: use int instead of index\n960892ca1 - JacobiSVD: get rid of m_scaledMatrix, m_adjoint, hopefully fix some compiler warnings\n18a161bf1 - fix pexp_complex_test\nbe06c9ad5 - Implement float pexp_complex\n4d419e220 - Rename generic_fast_tanh_float to ptanh_float and move it to...\n\nPiperOrigin-RevId: 622268063\nChange-Id: Ia444528ed7190f328160d8784de910a3722b5aed\n"
    },
    {
      "commit": "f6c457cdc27f232d7cf7f4f9b801aa8c3f9f18c6",
      "tree": "5226b6207d50c4abdd9b143d8187be75991adfb0",
      "parents": [
        "24b67b333c27d568c44af32af4f8d5a99e032dfe"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Tue Mar 26 11:14:22 2024 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Tue Mar 26 11:15:05 2024 -0700"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 619251887\nChange-Id: Ia819717250bc88294af24451d3340822e72c429f\n"
    },
    {
      "commit": "24b67b333c27d568c44af32af4f8d5a99e032dfe",
      "tree": "ae0ac7d36f7554c309080cfce80a7dc48756bacb",
      "parents": [
        "09e46ef6377d541dc306b8e6eaae2b5ea6d654dc"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Feb 22 11:37:01 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Feb 22 11:37:44 2024 -0800"
      },
      "message": "Remove int return values from LAPACK functions.\n\nPiperOrigin-RevId: 609443878\nChange-Id: I4a23ce01977c42315de900c5d77254c33dffba92\n"
    },
    {
      "commit": "09e46ef6377d541dc306b8e6eaae2b5ea6d654dc",
      "tree": "d3dde863bf830146e65f07cb08539427306faec8",
      "parents": [
        "615fd9f22a073ab1205d7e544a97f0ad7076b53e"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Feb 22 11:35:20 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Feb 22 11:35:52 2024 -0800"
      },
      "message": "Update Eigen to commit:2a9055b50ed22101da7d77e999b90ed50956fe0b\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n2a9055b50 - Fix random for custom scalars that don\u0027t have constexpr digits().\n500a3602f - Use traits\u003cMatrix\u003e::Options instead of Matrix::Options.\n0b9ca1159 - Fix deflation in BDCSVD.\nf40ad38fd - Fix failure on ARM with latest compilers.\na24bf2e9a - Disable float16 packet casting if native AVX512 f16 is available.\n5361dea83 - Remove return int types from BLAS/LAPACK functions.\n7e655c9a5 - Fixes 2780\n6ea33f95d - Eliminate warning about writing bytes directly to non-trivial type.\n06b45905e - Remove r_cnjg due to conflicts with f2c.\n9229cfa82 - Fix division by zero  UB in packet size logic.\n186f8205d - Apply clang-format to lapack/blas directories\n4eac211e9 - Fix C++20 error, Arithmetic between different enumeration types\nd1d87973f - Fix segfault in CholmodBase::factorize() for zero matrix\n7b87b2191 - Fix UB in bool packetmath test.\n431e4a913 - Fix the fuzz\n3ab8f4825 - fix tests when scalar is bfloat16, half\n3ebaab8a6 - Fix PPC rand and other failures.\nebd13c3b1 - fix skew symmetric test\n128c8abf4 - Fix gcc-6 bug in the rand test.\nd626762e3 - improve random\na9ddab3e0 - Fix a bunch of ODR violations.\n\nPiperOrigin-RevId: 609443252\nChange-Id: Iddb4afa1b81a173975c2ef3771fc4b08683fdbdc\n"
    },
    {
      "commit": "615fd9f22a073ab1205d7e544a97f0ad7076b53e",
      "tree": "aeed8f2c944dd050300a65ffa3107c63f5290ad8",
      "parents": [
        "01de8baeaf828c458935f1bf7b21cdde4c479ba5"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Jan 31 11:55:58 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Jan 31 11:56:28 2024 -0800"
      },
      "message": "Update Eigen to commit:7fd7a3f946e5ac152d28dad388cff8bfa1026925\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n7fd7a3f94 - Implement plog_complex\n043442e21 - Fix preshear transformation.\n69ee52ed1 - Remove Skyline.\n0f0c76dc2 - Use stableNorm in ComplexEigenSolver.\ndd71c23e2 - Remove MoreVectorization.\n62b547403 - Fix busted formatting in Eigen::Tensor README.md.\nd8d0b60b5 - Fix CI for clang-6 when cross-compiled.\n\nPiperOrigin-RevId: 603117689\nChange-Id: Ibcf2eda9792eb9140a58d262a2252f5ee9dcd558\n"
    },
    {
      "commit": "01de8baeaf828c458935f1bf7b21cdde4c479ba5",
      "tree": "1ad33b5528a6c8b190489b83ce5209b7987c11be",
      "parents": [
        "2434cfd1f4e6e30b715ae8e8b879bc3cf7326ad2"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Jan 25 21:28:23 2024 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Jan 25 21:29:54 2024 -0800"
      },
      "message": "Update Eigen to commit:f3912891504ad965d5fe8bd11d7346faa69b4026\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nf39128915 - Fix bug in checking subnormals.\n5a90fbcea - Update documentation of lapack second/dsecnd.\n907982024 - Remove simple relicense script.\n851b40afd - LAPACK CPU time functions.\na73970a86 - Fix arm32 issues.\n580812201 - Formatting.\n92f9544f6 - Remove explicit specialization of member function.\n2692fb2b7 - Fix compile-time error caused by chip static asserts\n2c6b61c00 - Add half and quarter vector support to HVX architecture\n05a457534 - Chipping Asserts v2\nfc92fe312 - SPQR: Fix build error, Index/StorageIndex mismatch.\neede526b7 - [Compressed Storage] Use smaller type of Index \u0026 StorageIndex for determining maximum size during resize.\n772057c55 - Revert \"Add asserts for .chip\"\n6163dbe2b - Allow specifying a temporary directory for fileio outputs.\n6b6bb9d34 - Fix unused warnings in failtest.\nf6e41e643 - Revert \"Clean up stableNorm\"\nb1ae206ea - Add asserts for .chip\nb0f906419 - add missing constexpr qualifier\n34fd46a9b - Update CI with testing framework from eigen_ci_cross_testing.\nb2814d53a - Fix stableNorm when input is zero-sized.\nc29a41011 - check pointers before freeing\n48e0c827d - [ROCm] MI300 related test support\n538577301 - Fix TensorForcedEval in the case of the evaluator being copied.\n3f3bc6d86 - Improve documentation of SparseLU\nd33174d5a - Doc: Fix Basic slicing examples minor issues\n19b119288 - Add factor getters to Cholmod LLT/LDLT\na1a96fafd - Clean up stableNorm\n3026f1f29 - Fix various asan errors.\na2cf99ec6 - Fix GPU+clang+asan.\nfee5d60b5 - Fix MSAN failures.\n9697d481c - Fix up clang-format CI.\n85efa8329 - Set up clang-format in CI\n2c4541f73 - fix msvc clz\n75e273afc - Add internal ctz/clz implementation.\n\nPiperOrigin-RevId: 601659527\nChange-Id: I52089e61fd81692bb2147f7d15ad98eb56da27c1\n"
    },
    {
      "commit": "2434cfd1f4e6e30b715ae8e8b879bc3cf7326ad2",
      "tree": "61dda5d4edf4b97c0d27f85f2438cf98611b074a",
      "parents": [
        "2a10258d45ce0d23630971b7050643d0828d10de"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Dec 06 12:02:41 2023 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Dec 06 12:03:14 2023 -0800"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 588501530\nChange-Id: I44fd1429ce59639367931c04ea6ab40edacd698f\n"
    },
    {
      "commit": "2a10258d45ce0d23630971b7050643d0828d10de",
      "tree": "a7154c0cee51fb5c3d54b61938a130de71971946",
      "parents": [
        "2201fb98dc9e6f2f32faea749f3dc71ec7ecc2c2"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Dec 06 11:00:58 2023 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Dec 06 11:01:27 2023 -0800"
      },
      "message": "No public description\n\nPiperOrigin-RevId: 588482241\nChange-Id: I740534b73e33f8759a023c4188c5fdd2c8731abc\n"
    },
    {
      "commit": "2201fb98dc9e6f2f32faea749f3dc71ec7ecc2c2",
      "tree": "3502356f7d9aa33ad867ad767a2487f5a5876ddd",
      "parents": [
        "7e3d01b7599f6fc0fee8c5d7e7fefb1127138584"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Fri Dec 01 15:43:08 2023 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Dec 01 15:43:31 2023 -0800"
      },
      "message": "Update Eigen to commit:9ea520fc4510b49408f1445b603b1f5dad267c2c\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n9ea520fc4 - Ensure that mc is not smaller than Traits::nr\ndd8c71e62 - Fix typecasting for arm32\nb2cb49e28 - Static asserts to check for matching NumDimensions\n283dec7f2 - Update file GeneralMatrixVector.h\n66b9f4ed5 - Fix (u)int64_t-\u003efloat conversion on arm\nd1b03fb5c - Gemv microoptimization\n\nPiperOrigin-RevId: 587143778\nChange-Id: Ia7e9e827a65966b5642968ef5b1097db5b7e61cf\n"
    },
    {
      "commit": "7e3d01b7599f6fc0fee8c5d7e7fefb1127138584",
      "tree": "dd50bf59bdc4c246d136fa125553f6a09e497bd6",
      "parents": [
        "94941edc5fb0546f5f466b4b23bd7995244d31fa"
      ],
      "author": {
        "name": "Rasmus Munk Larsen",
        "email": "rmlarsen@google.com",
        "time": "Wed Nov 15 16:15:37 2023 -0800"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Wed Nov 15 16:16:13 2023 -0800"
      },
      "message": "Update Eigen to commit:3cf6bb6f1ce9cd3cd6bdc43cdcfae7f4822ff2c3\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n3cf6bb6f1 - Fix a bug in commit 76e8c0455396446f8166c798da5efe879e010bdc:\n32165c6f0 - Fix Wshorten-64-to-32 warning in gemm parallelizer\nb33dbb576 - Fix implicit narrowing warning in Parallelizer.h.\n3a9635b20 - Link pthread for product_threaded test\nf78c37f0a - traits\u003cRef\u003e::match: use correct strides\n516d08a49 - Fix typo in Parallelizer.h\n76e8c0455 - Generalize parallel GEMM implementation in Core to work with ThreadPool in addition to OpenMP.\n4d54c43d6 - Fix typo to allow nomalloc test to pass on AVX512.\na25f02d73 - Fix int overflow causing cxx11_tensor_gpu_1 to fail.\n6f9ad7da6 - fix Wshorten-64-to-32 warnings in div_ceil\n1aac9332c - TensorReduction: replace divup with div_ceil\n5de0f2f89 - Fixes #2735: Component-wise cbrt\n48b254a4b - Disable denorm deprecation warnings in MSVC C++23.\n176821f2f - Avoid building docs if cross-compiling or not top level.\n\nPiperOrigin-RevId: 582839371\nChange-Id: I1a6247e188e09c8f9c56c5bbda59d4714573f5fd\n"
    },
    {
      "commit": "94941edc5fb0546f5f466b4b23bd7995244d31fa",
      "tree": "30d3870fdbddb7828c32ddd48dc5facfa454c5e8",
      "parents": [
        "ea4e852e36ee0e87745db1b2600e4936ce08ccf0"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Fri Oct 20 10:16:10 2023 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Oct 20 10:16:42 2023 -0700"
      },
      "message": "Update Eigen to commit:aa6964bf3a34fd607837dd8123bc42465185c4f8\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\naa6964bf3 - Work around MSVC issue in Block XprType.\n877c2d1e9 - fix typo in comment\n0c9526912 - Pass div_ceil arguments by value.\nd9839718a - [ROCm] Replace HIP_PATH with ROCM_PATH for rocm 6.0\n5bdf58b8d - Eliminate use of _res.\na96545777 - Consolidate multiple implementations of divup/div_up/div_ceil.\n\nPiperOrigin-RevId: 575252840\nChange-Id: Ief064e2e85f2d17f6267a36093fcdfbd81d37dad\n"
    },
    {
      "commit": "ea4e852e36ee0e87745db1b2600e4936ce08ccf0",
      "tree": "5b650636d331da4d4c29c1f8e7e116ce827750b7",
      "parents": [
        "c57c9d5429314d4487aa992163fb87167b805324"
      ],
      "author": {
        "name": "Googler",
        "email": "noreply@google.com",
        "time": "Fri Oct 20 02:27:59 2023 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Fri Oct 20 02:28:38 2023 -0700"
      },
      "message": "Update Eigen to commit:aa6964bf3a34fd607837dd8123bc42465185c4f8\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\naa6964bf3 - Work around MSVC issue in Block XprType.\n877c2d1e9 - fix typo in comment\n0c9526912 - Pass div_ceil arguments by value.\nd9839718a - [ROCm] Replace HIP_PATH with ROCM_PATH for rocm 6.0\n5bdf58b8d - Eliminate use of _res.\na96545777 - Consolidate multiple implementations of divup/div_up/div_ceil.\n\nPiperOrigin-RevId: 575157496\nChange-Id: I3154dd9ac5d8713a1630caf4f502e3c082bbbbc0\n"
    },
    {
      "commit": "c57c9d5429314d4487aa992163fb87167b805324",
      "tree": "30d3870fdbddb7828c32ddd48dc5facfa454c5e8",
      "parents": [
        "acbc4b2460cc9650541d7cd956f871d14f49ec4f"
      ],
      "author": {
        "name": "C. Antonio Sanchez",
        "email": "cantonios@google.com",
        "time": "Thu Oct 19 20:55:49 2023 -0700"
      },
      "committer": {
        "name": "Copybara-Service",
        "email": "copybara-worker@google.com",
        "time": "Thu Oct 19 20:56:41 2023 -0700"
      },
      "message": "Update Eigen to commit:aa6964bf3a34fd607837dd8123bc42465185c4f8\n\nCHANGELOG\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\naa6964bf3 - Work around MSVC issue in Block XprType.\n877c2d1e9 - fix typo in comment\n0c9526912 - Pass div_ceil arguments by value.\nd9839718a - [ROCm] Replace HIP_PATH with ROCM_PATH for rocm 6.0\n5bdf58b8d - Eliminate use of _res.\na96545777 - Consolidate multiple implementations of divup/div_up/div_ceil.\n\nPiperOrigin-RevId: 575086623\nChange-Id: I9361807d4c9bdd69dd7876f7068538919bde690f\n"
    }
  ],
  "next": "acbc4b2460cc9650541d7cd956f871d14f49ec4f"
}