Issues

• partitionedtreetn: let contract return a chosen set of output projectors in one call

  enhancement
  tensor4all/tensor4all-rs#1048 · Filter by author maxsroda opened on Oct 29, 2026

• core: per-call overhead of a two-tensor contraction is µs-scale (5 µs contract_pair, 20 µs session route vs 0.08 µs faer); the fixed cost is in tenferro's call path, TagSet/Index is <70 ns

  tensor4all/tensor4all-rs#1046 · Filter by author shinaoka opened on Oct 29, 2026
  

• tensorbackend: mat_mul on small Complex64 products is 2.3-3.2x slower than faer on borrowed slices; operand copies explain the m = 1 shapes, a kernel gap remains at m = 8

  enhancement
  tensor4all/tensor4all-rs#1040 · Filter by author simonb282 opened on Oct 29, 2026
  3

  

• core, treetn: IdxTensor::dense_host_values copies the whole payload where with_dense_slice borrows it (5x at chi = 32, 90x at chi = 141); read-only call sites in TreeTNCachedEvaluator

  enhancement
  tensor4all/tensor4all-rs#1039 · Filter by author simonb282 opened on Oct 29, 2026
  5

  

• macOS: default CPU backend = Apple Accelerate (faer stays the default elsewhere)

  tensor4all/tensor4all-rs#1035 · Filter by author shinaoka opened on Oct 29, 2026
  2

  

• Performance: measured per-call costs below the evaluator, and proposed catalog entries (tracking issue)

  enhancement
  tensor4all/tensor4all-rs#1027 · Filter by author simonb282 opened on Oct 28, 2026
  1

• core, treetci: a MultiIndexCache entry point for matrix-shaped requests with keys built from the two sides (1.6x per entry over one key per point; 2.1x with a fast hasher)

  enhancement
  tensor4all/tensor4all-rs#1026 · Filter by author simonb282 opened on Oct 28, 2026

• Tagged releases on a regular cadence (main is 330 commits past v0.2.0)

  enhancement
  tensor4all/tensor4all-rs#1025 · Filter by author simonb282 opened on Oct 28, 2026

• core, treetci: private fields for IdxTensor::indices and the TreeTCI2 state, with read accessors (invariants can be broken from outside)

  enhancement
  tensor4all/tensor4all-rs#1024 · Filter by author simonb282 opened on Oct 28, 2026
  1

• treetn, treetci, core: remaining public anyhow::Error in TreeTNOperationError and in the TCI callback (exceptions to #498 / AGENTS.md:118?)

  enhancement
  tensor4all/tensor4all-rs#1023 · Filter by author simonb282 opened on Oct 28, 2026

• treetn: reading one point per call through TreeTNCachedEvaluator costs 2.7-5.6x more per point than one batch (global-pivot walk)

  enhancement
  tensor4all/tensor4all-rs#1022 · Filter by author simonb282 opened on Oct 28, 2026

• treetn: TreeTNCachedEvaluator::evaluate_batched_typed is 24-32 % faster per point with a fast hasher, pre-sized maps and one hash per point

  enhancement
  tensor4all/tensor4all-rs#1021 · Filter by author simonb282 opened on Oct 28, 2026
  2

• core: MultiIndexCache::evaluate_batched costs about twice a one-key lookup per requested point (follow-up to #802 / #852)

  enhancement
  tensor4all/tensor4all-rs#1020 · Filter by author simonb282 opened on Oct 28, 2026

• core: direct_sum silently drops non-paired indices instead of rejecting mismatched operands

  tensor4all/tensor4all-rs#1018 · Filter by author tensor4all-ai-bot opened on Oct 27, 2026

• Post-closure gap: #1014 pin fix did not propagate to tests/feature-configurations/core-backend-cuda/Cargo.lock

  tensor4all/tensor4all-rs#1017 · Filter by author tensor4all-ai-bot opened on Oct 26, 2026

• py: QuanticsTCI.to_numpy re-contracts the full TT per grid point instead of using TTCache::evaluate_many

  tensor4all/tensor4all-rs#1010 · Filter by author tensor4all-ai-bot opened on Oct 21, 2026

• Proposal: enforce the additive-growth rules (file-size trigger, reuse-before-add, removal attestation) as non-blocking review checks

  enhancement
  tensor4all/tensor4all-rs#917 · Filter by author lingrui96 opened on Oct 11, 2026

• treetci: two passes per iteration with an iteration-counted stop window cost 1.1-1.6x TCI2's function evaluations on chains

  enhancement
  tensor4all/tensor4all-rs#899 · Filter by author lingrui96 opened on Oct 11, 2026
  1

• [Umbrella] Adopt tenferro's Engine API: bump tenferro, dual frontends (implicit + engine-explicit)

  tensor4all/tensor4all-rs#859 · Filter by author shinaoka opened on Oct 9, 2026
  12

• [treeaci] Umbrella: convergence, guard and timing residuals; retire chain ACI

  bug
  tensor4all/tensor4all-rs#854 · Filter by author lingrui96 opened on Oct 8, 2026
  2

• [Umbrella] CI performance: cold builds on lockfile change, cache budget, redundant jobs

  enhancementperformancetesting
  tensor4all/tensor4all-rs#848 · Filter by author lingrui96 opened on Oct 7, 2026
  1

• partitionedtreetn: merge-refine schedule follow-ups (regime benchmark, probe cost, #752 residue)

  tensor4all/tensor4all-rs#777 · Filter by author shinaoka opened on Sep 25, 2026
  2

• SRC/incremental-QR bookkeeping pays O(final_size²) copy cost: Matrix/IdxTensor have no capacity-reserving growth API

  tensor4all/tensor4all-rs#696 · Filter by author lingrui96 opened on Aug 30, 2026
  1

• Parallel SRC via interface sketching (combine arXiv:2606.23274 sub-TT partitioning with SRC)

  enhancement
  tensor4all/tensor4all-rs#691 · Filter by author shinaoka opened on Aug 27, 2026
  2

• TTCache: batched miss contraction in evaluate_many, integer keys for persistent caches, bounded LRU

  tensor4all/tensor4all-rs#627 · Filter by author shinaoka opened on Aug 14, 2026
  2