Phase 0: disposition of open issues targeting the retiring crates

[shinaoka]

Part of #199 (Phase 0).

The retirement makes the einsum2/opteinsum improvement backlog obsolete here. For each issue: close as superseded by #199, or transfer the idea to tenferro if it still applies to the absorbed implementation.

• strided-einsum2: allow multiple GEMM backends to coexist and support runtime selection #86 (multiple GEMM backends, runtime selection): transfer decision to tenferro (provider selection is CpuBackend policy there).
• einsum2 batched GEMM has 3.5x overhead vs Julia for small-m contractions #115 (small-m batched GEMM overhead): transfer to tenferro with the benchmark.
• Adopt OMEinsum.jl's pre-permutation dimension fusion in einsum2 #116 (OMEinsum pre-permutation dimension fusion): transfer to tenferro-einsum or close.
• Reduce code duplication in contiguous.rs prepare functions #120, Extract common pipeline from einsum2 entry points in lib.rs #121, Merge BackendConfig and BgemmBackend into a single trait #122, Simplify axis classification: reduce 6 groups to 4 + lazy trace #126, Extract transpose-flip helper in bgemm_blas.rs #127 (einsum2 internal refactors): close as superseded.
• BufferPool: use best-fit (>=) instead of exact-size matching #123, BufferPool: make it user-controllable via eval() API #124 (BufferPool best-fit, user control): close as superseded (pooling is owned by tenferro-cpu's BufferPool).
• Use cblas_dgemm_batch when available instead of manual batch loop #125 (cblas_dgemm_batch): transfer to tenferro.
• perf: avoid finalize copy in einsum2 by deferring output permutation (lazy permute) #128 (lazy output permutation): transfer to tenferro.
• Track Faer typed MaybeUninit overwrite API for strided GEMM #198 (faer typed MaybeUninit overwrite API tracking): transfer to tenferro or close with a pointer from the absorption issue.
• Nested parallelism: Strided thread splitting vs BLAS internal threading #6 (nested parallelism, Strided splitting vs BLAS threading): split; the BLAS-internal-threading half moves to tenferro, the strided splitting half stays here.

[shinaoka]

Cross-repository disposition for #198: no tenferro transfer is required because tenferro's active Faer dot path uses initialized output storage. The supporting audit is tensor4all/tenferro-rs#1591 (comment).

The remaining #198 disposition can therefore be resolved locally alongside this transfer/closure ledger and #202.

[shinaoka]

Phase 0 disposition update

Audited strided-rs main at 0082be119828af37da535c1cf5ba3c9b91092a1c and tenferro-rs origin/main at b724a3065a129c7eab4e9a8e3241537e1c5a1bdd.

Closed as already implemented: #122, #123, #124, #126, #127.
Closed as superseded/not planned under the retirement decision in #199/#202: #120 and #121.
The tenferro counterpart #1592 was also closed after recording the explicit “nothing to port” decision; current tenferro-einsum owns omeco planning, repeated-label/single-tensor paths, and runtime-owned caching.

Still intentionally open for a separate disposition: #86, #115, #116, #125, #128, and #6; #198 remains blocked on the external Faer typed overwrite contract. Phase 2 removal remains #202, and this comment does not close or redefine that work.

[shinaoka]

2026-08-10 棚卸し

obsolete ledger — close as not planned。retirement 前提は #226 で撤回された。#198 も upstream tracker として継続する。各 issue の現行 disposition は #230 に移したため、本 ledger は実装せず close する。

Umbrella: #230。実装 PR の base は umbrella/issue-burndown-2026-08。