[mlir][sparse] Converting SparseTensorCOO to use standard C++-style iterators.
authorwren romano <2998727+wrengr@users.noreply.github.com>
Fri, 7 Oct 2022 21:21:02 +0000 (14:21 -0700)
committerwren romano <2998727+wrengr@users.noreply.github.com>
Tue, 11 Oct 2022 21:03:37 +0000 (14:03 -0700)
commit90fd13b0a1120a7fa50a21d4d46af61285b7a964
treed5fb328d1b37f08b4c1ce04620983b3ee0050624
parent1079662d2fff7ae799503d910b299c5108d105fd
[mlir][sparse] Converting SparseTensorCOO to use standard C++-style iterators.

This differential comprises three related changes: (1) it gives SparseTensorCOO standard C++-style iterators; (2) it removes the old iterator stuff from SparseTensorCOO; and (3) it introduces SparseTensorIterator which behaves like the old SparseTensorCOO iterator stuff used to.

The SparseTensorIterator class is needed because the MLIR codegen cannot easily use the C++-style iterators (hence why SparseTensorCOO had the old iterator stuff).  Distinguishing SparseTensorIterator from SparseTensorCOO also helps improve API hygiene since these two classes are used for distinct purposes.  And having SparseTensorIterator as its own class enables changing the underlying implementation in the future, without needing to worry about updating all the codegen tests etc.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D135485
mlir/include/mlir/ExecutionEngine/SparseTensor/COO.h
mlir/include/mlir/ExecutionEngine/SparseTensorUtils.h
mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp
mlir/lib/ExecutionEngine/SparseTensorUtils.cpp
mlir/test/Dialect/SparseTensor/sparse_concat.mlir
mlir/test/Dialect/SparseTensor/sparse_reshape.mlir