llvm/llvm-project

Implement more Python DSL-based "exhaustive" sparse kernel generators (easy starter)

Open

#51.657 aperta il 26 ott 2021

Vedi su GitHub
 (7 commenti) (0 reazioni) (1 assegnatario)C++ (10.782 fork)batch import
bugzillagood first issuemlir:llvmmlir:sparsetensor

Metriche repository

Star
 (26.378 star)
Metriche merge PR
 (Merge medio 1g 2h) (1000 PR mergiate in 30 g)

Descrizione

Bugzilla Link 52315
Version unspecified
OS Linux
CC @joker-eph

Extended Description

Currently, we have an interesting "exhaustive" test generator that constructs, runs, and verifies many versions of SpMM (see https://github.com/llvm/llvm-project/blob/main/mlir/test/python/dialects/sparse_tensor/test_SpMM.py) [note that we don't actually exhaustively generate all versions to keep testing time down, but the code can in principle generate and test all versions).

We need more interesting kernels like this to be "exhaustively" tested(or, for practical reasons, pick a good subset of versions). This entry encapsulates adding a number of interesting kernels to this directory using the same style of DSL based testing.

Guida contributor