Sparse Tensor Dialect Repository Structure

AI-generated Key Takeaways

• The Sparse Tensor Dialect, residing within the MLIR repository, provides infrastructure for representing and processing sparse tensors.

• It offers comprehensive support including header files, a runtime support library, Python bindings, and a C API for integration with external systems.

• The dialect includes a benchmark suite for performance evaluation and a robust testing framework encompassing file checks, end-to-end tests, unit tests, and CAPI/Python binding tests.

• Developers can leverage the Sparse Tensor Dialect to optimize computations involving sparse data structures within the MLIR ecosystem.

Sparse Tensor Dialect is located in MLIR repository.

Header Files

• mlir/include/mlir/Dialect/SparseTensor
• mlir/include/mlir/ExecutionEngine/SparseTensor

Library and C++ Files

• mlir/lib/Dialect/SparseTensor
• Sparse tensor runtime support library for execution engine
  • mlir/lib/ExecutionEngine/SparseTensor (Runtime)

Python Bindings and CAPI

• mlir/include/mlir-c/Dialect/SparseTensor.h
• mlir/lib/Bindings/Python/DialectSparseTensor.cpp (Passes)
• mlir/lib/CAPI/Dialect/SparseTensor.cpp (Passes)
• mlir/python/mlir/dialects/sparse_tensor.py

MLIR Benchmark

• mlir/benchmark/python/benchmark_sparse.py

Tests

• mlir/test/Dialect/SparseTensor (FileCheck tests)
• mlir/test/Integration/Dialect/SparseTensor (End-to-end tests)
• mlir/unittests/Dialect/SparseTensor (Unit tests)
• mlir/test/CAPI/sparse_tensor.c (CAPI tests)
• mlir/test/python/dialects/sparse_tensor/dialect.py (Python binding tests)