[all-commits] [llvm/llvm-project] 16b8f4: [mlir][sparse] add a "release" operation to sparse...

[Aart Bik via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 16b8f4ddae1cb36ac16c6eb451613c032e4064f6
      https://github.com/llvm/llvm-project/commit/16b8f4ddae1cb36ac16c6eb451613c032e4064f6
  Author: Aart Bik <ajcbik at google.com>
  Date:   2021-10-05 (Tue, 05 Oct 2021)

  Changed paths:
    M mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
    M mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
    M mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp
    M mlir/test/Dialect/SparseTensor/conversion.mlir
    M mlir/test/Dialect/SparseTensor/invalid.mlir
    M mlir/test/Dialect/SparseTensor/roundtrip.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/lit.local.cfg
    R mlir/test/Integration/Dialect/SparseTensor/CPU/sparse-constant_to_sparse_tensor.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_cast.mlir
    A mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_constant_to_sparse_tensor.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_conversion.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_filter_conv2d.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_flatten.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_matvec.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_mttkrp.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_out_simple.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_quantized_matmul.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_reductions.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sampled_matmul.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sampled_mm_fusion.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_scale.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_spmm.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_storage.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sum.mlir

  Log Message:
  -----------
  [mlir][sparse] add a "release" operation to sparse tensor dialect

We have several ways to materialize sparse tensors (new and convert) but no explicit operation to release the underlying sparse storage scheme at runtime (other than making an explicit delSparseTensor() library call). To simplify memory management, a sparse_tensor.release operation has been introduced that lowers to the runtime library call while keeping tensors, opague pointers, and memrefs transparent in the initial IR.

*Note* There is obviously some tension between the concept of immutable tensors and memory management methods. This tension is addressed by simply stating that after the "release" call, no further memref related operations are allowed on the tensor value. We expect the design to evolve over time, however, and arrive at a more satisfactory view of tensors and buffers eventually.

Bug:
http://llvm.org/pr52046

Reviewed By: bixia

Differential Revision: https://reviews.llvm.org/D111099