[mlir][sparse] recognize ReLu operation during sparsification (#92016)

This is a proof of concept recognition of the most basic forms of ReLu
operations, used to show-case sparsification of end-to-

[mlir][sparse] recognize ReLu operation during sparsification (#92016)

This is a proof of concept recognition of the most basic forms of ReLu
operations, used to show-case sparsification of end-to-end PyTorch
models. In the long run, we must avoid lowering such constructs too
early (with this need for raising them back).

See discussion at

https://discourse.llvm.org/t/min-max-abs-relu-recognition-starter-project/78918

show more ...

[mlir][sparse] add merger support on Batch LevelType. (#83186)

Reapply "[mlir][sparse] remove LevelType enum, construct LevelType from LevelFormat and Properties" (#81923) (#81934)

Revert "[mlir][sparse] remove LevelType enum, construct LevelType from LevelF…" (#81923)

Reverts llvm/llvm-project#81799 ; this broke the mlir gcc7 bot.

[mlir][sparse] remove LevelType enum, construct LevelType from LevelF… (#81799)

…ormat and properties instead.

[mlir][sparse] rename DimLevelType to LevelType (#73561)

The "Dim" prefix is a legacy left-over that no longer makes sense, since
we have a very strict "Dimension" vs. "Level" definition for sparse

[mlir][sparse] rename DimLevelType to LevelType (#73561)

The "Dim" prefix is a legacy left-over that no longer makes sense, since
we have a very strict "Dimension" vs. "Level" definition for sparse
tensor types and their storage.

show more ...

[mlir][sparse] code cleanup (remove dead code related to filter loop). (#72573)

[mlir][sparse] cleanup merger test, add header (#70279)

[mlir][sparse] minor cleanup of merger unit test

Removed some of the warning supression needed for
the multi-arg macro logic by making number of arguments
the same everywhere. Also removes some verb

[mlir][sparse] minor cleanup of merger unit test

Removed some of the warning supression needed for
the multi-arg macro logic by making number of arguments
the same everywhere. Also removes some verbose comments
and obvious TODOs.

Reviewed By: wrengr

Differential Revision: https://reviews.llvm.org/D157327

show more ...

[mlir][sparse] admit un-sparsifiable operations if all its operands are loaded from dense input

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D153998

[mlir][sparse] merger extension to support sparsifying arith::CmpI/CmpF operation

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D152761

[mlir][sparse] Preliminary code changes for ExprId, LatPointId, LatSetId newtypes

This commit contains several code changes which are ultimately required for converting the varions `Merger` identifi

[mlir][sparse] Preliminary code changes for ExprId, LatPointId, LatSetId newtypes

This commit contains several code changes which are ultimately required for converting the varions `Merger` identifiers from typedefs to newtypes. The actual implementation of the newtypes themselves has been split off into separate commits, in hopes of simplifying the review process.

Depends On D146561

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146684

show more ...

[mlir][sparse] Fixing -Wignored-reference-qualifiers in MergerTest.cpp

These warnings were introduced by D146561.

Reviewed By: aartbik, Peiming

Differential Revision: https://reviews.llvm.org/D147

[mlir][sparse] Fixing -Wignored-reference-qualifiers in MergerTest.cpp

These warnings were introduced by D146561.

Reviewed By: aartbik, Peiming

Differential Revision: https://reviews.llvm.org/D147090

show more ...

[mlir][sparse] Removing shared_ptr from the MergerTest.cpp unit test

This is a preliminary change to make way for converting the Merger's identifier types from mere typedefs to actual types (which c

[mlir][sparse] Removing shared_ptr from the MergerTest.cpp unit test

This is a preliminary change to make way for converting the Merger's identifier types from mere typedefs to actual types (which causes some issues that this patch fixes).

Depends On D146676

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146561

show more ...

[mlir][sparse] fix crash when using pure constant index in indexing mapping (fixes #61530)

To address https://github.com/llvm/llvm-project/issues/61530

Reviewed By: aartbik, wrengr

Differential Re

[mlir][sparse] fix crash when using pure constant index in indexing mapping (fixes #61530)

To address https://github.com/llvm/llvm-project/issues/61530

Reviewed By: aartbik, wrengr

Differential Revision: https://reviews.llvm.org/D146563

show more ...

[mlir][sparse] Making `TensorExp::Kind` a nested enum-class

This improves namespacing, and follows the pattern used for "Kind" enums elsewhere in MLIR.

Reviewed By: aartbik

Differential Revision:

[mlir][sparse] Making `TensorExp::Kind` a nested enum-class

This improves namespacing, and follows the pattern used for "Kind" enums elsewhere in MLIR.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146086

show more ...

[mlir][sparse] Cleaning up names in {Merger,LoopEmitter,CodegenEnv}.{h,cpp}

This change does a bunch of renaming to clear up confusions in these files. In particular, this change:

* Renames variab

[mlir][sparse] Cleaning up names in {Merger,LoopEmitter,CodegenEnv}.{h,cpp}

This change does a bunch of renaming to clear up confusions in these files. In particular, this change:

* Renames variables and methods to clarify the "dim"/"lvl" distinction, and changes them to use the `Dimension`/`Level` types as appropriate.
* Introduces new typedefs
* `ExprId`, `LatPointId`, `LatSetId`: to clarify the interning design of the Merger.
* `LoopId`, `LoopOrd`: to clarify the distinction between arbitrary names for loop-variables, vs numeric identifiers based on the actual order of loop generation.
* `TensorId`
* (Future CLs will change these from typedefs to structs/classes, so that the typechecker can help avoid mixups.)
* Updates documentation to match the new terminology
* Adds additional assertions
* Adds `const` to local variables along the way

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D145756

show more ...

[mlir][sparse] support affine expression on sparse dimensions (analysis implementation)

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D138171

Apply clang-tidy fixes for readability-identifier-naming in MergerTest.cpp (NFC)

[mlir][sparse] code refactoring, move <tid, loop id> -> dim map to Merger.

To address unresolved comments in D136185

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D136780

[mlir][sparse] add some APIs for merger to query the tensor id for output tensor and synthetic tensor.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D136630

[mlir][sparse] Removing the DimLvlType and DimLevelFormat types

This removes another massive source of redundancy, and instead has the Merger.{h,cpp} reuse the SparseTensorEnums library.

Depends On

[mlir][sparse] Removing the DimLvlType and DimLevelFormat types

This removes another massive source of redundancy, and instead has the Merger.{h,cpp} reuse the SparseTensorEnums library.

Depends On D136005

Reviewed By: Peiming

Differential Revision: https://reviews.llvm.org/D136123

show more ...

[mlir][sparse] Favors synthetic tensor over other undefined tensors

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D135385

[mlir][sparse] Favors defined dimension when optimize lattice points.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D135337

[mlir][sparse] Implement sparse_tensor.select

The region within sparse_tensor.select is used as the runtime criteria
for whether to keep the existing value in the sparse tensor.

While the sparse el

[mlir][sparse] Implement sparse_tensor.select

The region within sparse_tensor.select is used as the runtime criteria
for whether to keep the existing value in the sparse tensor.

While the sparse element is provided to the comparison, indices may also
be used to decide on whether to keep the original value. This allows, for
example, to only keep the upper triangle of a matrix.

Reviewed by: aartbik

Differential Revision: https://reviews.llvm.org/D134761

show more ...