Issues

• faer-ffi: let the team handle own the pool instead of rebuilding it per call

  tensor4all/faer#333 · Filter by author shinaoka opened on Nov 8, 2026
  

• tiled copy_across_layouts should be chosen on the working-set size, not on the layout alone

  perf
  tensor4all/faer#323 · Filter by author shinaoka opened on Nov 7, 2026
  

• Row-major parity: run every dense factorization natively on row-major input via transpose duality (no whole-matrix copy), at column-major performance

  enhancement
  tensor4all/faer#314 · Filter by author shinaoka opened on Nov 7, 2026
  4

  

• docs(tracking): point to this library from upper layers' agent docs (tenferro, tensor4all-rs, tlinalg, Colmat.jl, Python)

  documentation
  tensor4all/faer#313 · Filter by author shinaoka opened on Nov 7, 2026
  

• API request: run-time values for SchurParams/GeneralizedSchurParams shift and deflation controls

  tensor4all/faer#302 · Filter by author terasakisatoshi opened on Nov 7, 2026
  2

• the #282 tiled triangular copy was adopted on a measurement that does not meet the adoption protocol

  tensor4all/faer#290 · Filter by author shinaoka opened on Nov 7, 2026
  2

• Measure the SIMD LU leaf that landed in #276 (unmeasured in 0.31.0)

  tensor4all/faer#279 · Filter by author shinaoka opened on Nov 5, 2026
  3

  

• Held-team saved Schur eigenvectors allocate repeatedly for n=257 ordinary F-layout input

  tensor4all/faer#271 · Filter by author terasakisatoshi opened on Nov 5, 2026

• Saved Schur eigenvector scratch cannot cover the internal range-scaling Mat

  tensor4all/faer#270 · Filter by author terasakisatoshi opened on Nov 5, 2026
  1

• Owning dense constructors cannot consume caller scratch (LLT allocation MWE)

  tensor4all/faer#268 · Filter by author terasakisatoshi opened on Nov 5, 2026
  2

• Householder factor upgrade splits multiple blocks at a block count, producing wrong coefficients

  tensor4all/faer#265 · Filter by author terasakisatoshi opened on Nov 5, 2026
  

• Real QZ AED scratch omits three Sylvester permutation arrays (n=18)

  tensor4all/faer#260 · Filter by author terasakisatoshi opened on Nov 4, 2026

• Real QZ loses orthogonality and coupled reconstruction accuracy with a valid maximum AED window

  tensor4all/faer#257 · Filter by author terasakisatoshi opened on Nov 4, 2026
  1

• Generalized Hessenberg reduction returns nonfinite factors for finite subnormal inputs

  tensor4all/faer#256 · Filter by author terasakisatoshi opened on Nov 4, 2026

• gevd: make the dlaln2 conformance verifiable (differential test + documented deviations)

  tensor4all/faer#234 · Filter by author shinaoka opened on Nov 4, 2026

• Handover: thread the reductions (#178 / #179 / #180), order, bars and protocol

  tensor4all/faer#198 · Filter by author shinaoka opened on Nov 2, 2026
  

• perf(aarch64): Apple Silicon (M5 Max) measurement and ordered gap list

  tensor4all/faer#186 · Filter by author shinaoka opened on Nov 1, 2026

• sparse: block sparse matrices and rank>2 sparse tensors for O(N) DFT kernels

  tensor4all/faer#185 · Filter by author shinaoka opened on Nov 1, 2026

• eigh loses its single thread lead by four threads and is 1.4x to 1.6x slower at eight

  tensor4all/faer#180 · Filter by author shinaoka opened on Oct 31, 2026
  4

  

• svd_values is slower than MKL at every thread count, and worse as threads grow

  tensor4all/faer#179 · Filter by author shinaoka opened on Oct 31, 2026
  4

  

• Reassess current tall-skinny QR scaling and identify any remaining bottleneck

  tensor4all/faer#178 · Filter by author shinaoka opened on Oct 31, 2026
  57

  

• Pfaffian: the symmetric swaps are the serial wall, not the trailing update

  tensor4all/faer#176 · Filter by author shinaoka opened on Oct 31, 2026
  4

• Pfaffian: quiet-host Team32 scaling and actual AVX-512 validation

  tensor4all/faer#173 · Filter by author shinaoka opened on Oct 31, 2026
  

• Umbrella: parallel structure of the factorizations (equal-area partitions, static team ownership, look-ahead, block width per thread count)

  tensor4all/faer#168 · Filter by author shinaoka opened on Oct 30, 2026
  10

  

• Port OpenBLAS test/ and ctest/ BLAS-level checks into lightweight unit tests (follow-up to #31)

  tensor4all/faer#166 · Filter by author shinaoka opened on Oct 30, 2026