v1.1.0

Finch v1.1.0

Diff since v1.0.1

Merged pull requests:

• accelerate singleton levels (#675) (@willow-ahrens)
• CompatHelper: bump compat for UnsafeAtomics to 0.3, (keep existing compat) (#677) (@github-actions[bot])
• hotfix benchmarks (#680) (@willow-ahrens)
• Wma/benchmark improvements (#681) (@willow-ahrens)
• Update Project.toml (#684) (@willow-ahrens)
• Update Project.toml (#685) (@willow-ahrens)

Closed issues:

• Remove conditionals in sparsePoint, SparseBand, SparseInterval. (#604)
• Performance regression in Bellman Ford (#617)
• Add Galley to Benchmarks (#674)

v1.0.1

Finch v1.0.1

Diff since v1.0.0

Merged pull requests:

• Update README.md (#655) (@willow-ahrens)
• Update CONTRIBUTING.md (#656) (@willow-ahrens)
• Replace pyjulia with juliacall in docs (#657) (@gdalle)
• Update README.md (#658) (@willow-ahrens)
• run all the benchmarks by default. (#661) (@willow-ahrens)
• Add testsuite for parallel spmv using mutex (#668) (@Paramuths)
• Tensor Init Bug Fix (#670) (@kylebd99)
• Wma/fix sprand (#671) (@willow-ahrens)
• Kbd-loop-order-cost-model (#673) (@kylebd99)
• patch release (#676) (@willow-ahrens)

Closed issues:

• Post to discourse? (#599)
• fsprand distribution seems broken (#663)
• Performance of Matrix Chain Multiplication in Python (#669)

v1.0.0

Finch v1.0.0

Diff since v0.6.33

Merged pull requests:

• update docs (#640) (@willow-ahrens)
• Add Galley into Finch Codebase (#641) (@kylebd99)
• Update CI.yml (#642) (@willow-ahrens)
• Update CI.yml (#643) (@willow-ahrens)
• add Galley to precompile.jl (#646) (@kylebd99)
• add instance node (#649) (@kylebd99)
• #650 Galley Bug Fix (#651) (@kylebd99)
• Wma/docs3 (#652) (@willow-ahrens)
• Wma/rm links (#653) (@willow-ahrens)

Closed issues:

• Revise getting started guides and Readme (#603)
• swizzle generate dense tensor instead of sparse (#635)
• Finch.jlregressions (#644)
• Tag FinchLogicExecutor (#647)
• Galley crashes sometimes (#650)

v0.6.33

Finch v0.6.33

Diff since v0.6.32

Merged pull requests:

• add atomic element level (#515) (@willow-ahrens)
• fix 609 (#610) (@willow-ahrens)
• SparseList Follow Protocol Speedup (#613) (@kylebd99)
• Fixed missing Finch. before @closure (#616) (@Paramuths)
• Rename links (#619) (@willow-ahrens)
• Update ⚙ in tensor_formats.md for consistency (#620) (@AbdAlazezAhmed)
• update level names (#625) (@willow-ahrens)
• fix tensordot (#626) (@willow-ahrens)
• Wma/backwards fusion (#627) (@willow-ahrens)
• quick fix (#628) (@willow-ahrens)
• Wma/fix608 (#629) (@willow-ahrens)
• add benchmarks for special structures (#632) (@willow-ahrens)
• update version (#633) (@willow-ahrens)
• Introduce a de-parser to return Finch expressions to their readable macro input form (#634) (@willow-ahrens)
• Kbd small finch logic changes (#636) (@kylebd99)
• Add back tests for Galley to CI (#637) (@kylebd99)
• Version Bump to "0.6.33" (#638) (@kylebd99)

Closed issues:

• Documentation needed for separated memory level and Atomic Level (#531)
• When does Finch store fill values? (#605)
• Limit the scope of Looplet/Unfurl closures (#608)
• Copying swizzles (#609)
• Backwards Fusion Heuristic in Autoscheduler (#614)
• Matmul and reduce in lazy mode fails with ArgumentError (#615)
• Rename levels to match thesis (#618)
• Reorganize docs to center high-level interfaces (#621)
• FinchProgram => @finch invocation (#624)
• TTM @Einsum versus handwritten performance (#630)

v0.6.32

Finch v0.6.32

Diff since v0.6.31

Merged pull requests:

• fix eltype calculations, improve stability of nothrow divisions (#593) (@willow-ahrens)
• Wma/permute dims perf (#594) (@willow-ahrens)
• fixes #383 (#595) (@willow-ahrens)
• parallel benchmarks (#596) (@willow-ahrens)
• add follow protocol to sparse_list_level (#597) (@kylebd99)
• Wma/parallel fix (#598) (@willow-ahrens)
• Improve the Sparse level's sort procedure (#600) (@kylebd99)
• updates to SparseDictLevel (#601) (@willow-ahrens)
• Release 0.6.32 (#606) (@mtsokol)

Closed issues:

• Exception when negating a fiber (#383)
• Performance issue for row-major dense Tensors (#525)
• Follow protocol for SparseDict (#540)
• Multiple sparse hash outputs have collisions (#579)
• Dead code elimination does not work on macros generated by Finch (#582)
• Tracking issue for transposition performance (#584)
• floor_div mixed dtypes (#592)

v0.6.31

Finch v0.6.31

Diff since v0.6.30

Merged pull requests:

• Exponential Scansearch (#569) (@kylebd99)
• CompatHelper: add new compat entry for HDF5 in [weakdeps] at version 0.17, (keep existing compat) (#570) (@github-actions[bot])
• CompatHelper: add new compat entry for TensorMarket in [weakdeps] at version 0.2, (keep existing compat) (#571) (@github-actions[bot])
• CompatHelper: add new compat entry for NPZ in [weakdeps] at version 0.4, (keep existing compat) (#572) (@github-actions[bot])
• ambiguity error (#573) (@willow-ahrens)
• SparseByteMap follow fix (#581) (@kylebd99)
• Wma/gettersetter (#583) (@willow-ahrens)
• Add update protocol for SparseMatrixCSC and SparseVector (#585) (@willow-ahrens)
• perf improvement (#587) (@willow-ahrens)
• Wma/fix326 (#588) (@willow-ahrens)
• Wma/compiler docs (#589) (@willow-ahrens)
• fix eltype issues (#590) (@willow-ahrens)
• Release 0.6.31 (#591) (@mtsokol)

Closed issues:

• Vertical Loop Fusion (#155)
• Position Space Wrapper (#156)
• Post-PLDI Code Cleanup (#292)
• implement update protocol for SparseMatrixCSC and SparseVector (#320)
• all direct references to fields of the ctx should become getters/setters/etc (#324)
• Add an internal docs section describing virtual functions (#325)
• make the interface to looplets more uniform (#326)
• Abstract types with docs (#327)
• Add indexing tests (#332)
• Improvements to the loop ordering heuristic (#568)
• support non-boolean dtypes in any and all (#575)
• Incorrect eltype in tensordot for Complex inputs (#576)
• tensordot - input and output eltypes differences (#577)
• maximum, minimum always return float eltype (#578)
• SparseByteMap Follow Protocol Incorrect Results (#580)

v0.6.30

Finch v0.6.30

Diff since v0.6.29

Merged pull requests:

• add benchmarks to measure improvements to #565 (#566) (@willow-ahrens)
• Wma/fix565 (#567) (@willow-ahrens)

Closed issues:

• SDDMM speedup goal (#538)
• PlusOneVector performance regression (#565)

v0.6.29

Finch v0.6.29

Diff since v0.6.28

Merged pull requests:

• add heuristic loop order (#562) (@willow-ahrens)
• Wma/binsparse tensor (#564) (@willow-ahrens)

Closed issues:

• Add a loop ordering heuristic to the Compute interface. (#468)

v0.6.28

Finch v0.6.28

Diff since v0.6.27

Merged pull requests:

• indexing mistake in binsparse (#563) (@willow-ahrens)

v0.6.27

Finch v0.6.27

Diff since v0.6.26

Merged pull requests:

• fixing binsparse (#555) (@willow-ahrens)
• Wma/default >fill value (#556) (@willow-ahrens)
• add warning operators in Finch to allow floor division without errors (#557) (@willow-ahrens)
• Update Project.toml (#559) (@willow-ahrens)

Closed issues:

• RuleRewriteError in .// and rem. MethodError in prod (#474)
• Eager reshape implementation (#490)
• matrixmarket reader may be swapping row and column (#523)
• rename default to fill_value (#535)
• binsparse is out of date (#537)