Page MenuHomePhabricator

ftynse (Alex Zinenko)
PhD

Projects

User does not belong to any projects.

User Details

User Since
Jun 28 2018, 2:24 AM (248 w, 18 h)

Recent Activity

Today

ftynse added a comment to D147223: [mlir] add readonly/consume annotations to transform named sequences.

Do you also have to update TransformDialect::verifyOperationAttribute?

Thu, Mar 30, 5:42 AM · Restricted Project, Restricted Project
ftynse requested review of D147223: [mlir] add readonly/consume annotations to transform named sequences.
Thu, Mar 30, 5:35 AM · Restricted Project, Restricted Project
ftynse accepted D147210: [mlir] Fix folding into tensor.pad op..
Thu, Mar 30, 2:28 AM · Restricted Project, Restricted Project

Yesterday

ftynse added a comment to D146961: [mlir] support external named transform libraries.

Has this failure been clearly root-caused to this patch? Because I have really hard time seeing how SCF bufferization tests are related to this change.

Wed, Mar 29, 1:17 AM · Restricted Project, Restricted Project

Tue, Mar 28

ftynse added inline comments to D146912: [MLIR][Analysis] Use Identifiers outside Presburger library.
Tue, Mar 28, 3:43 AM · Restricted Project, Restricted Project
ftynse committed rG92c694684052: [mlir] support external named transform libraries (authored by ftynse).
[mlir] support external named transform libraries
Tue, Mar 28, 2:47 AM · Restricted Project, Restricted Project
ftynse closed D146961: [mlir] support external named transform libraries.
Tue, Mar 28, 2:47 AM · Restricted Project, Restricted Project
ftynse added inline comments to D147035: [mlir] Add missing dependency to Bazel build fails..
Tue, Mar 28, 2:30 AM · Restricted Project, Restricted Project
ftynse accepted D147035: [mlir] Add missing dependency to Bazel build fails..

Feel free to commit such changes directly.

Tue, Mar 28, 2:30 AM · Restricted Project, Restricted Project
ftynse updated the diff for D146961: [mlir] support external named transform libraries.

Address review

Tue, Mar 28, 2:18 AM · Restricted Project, Restricted Project
ftynse accepted D147028: [mlir][doc] Fix typos.
Tue, Mar 28, 2:17 AM · Restricted Project, Restricted Project
ftynse accepted D147018: [mlir][python] Mark operator== overloads as const.
Tue, Mar 28, 1:44 AM · Restricted Project, Restricted Project

Mon, Mar 27

ftynse added a comment to D144469: [mlir] Implement pass utils for 1:N type conversions..

It's okay to reland.

Mon, Mar 27, 8:03 AM · Restricted Project, Restricted Project
ftynse committed rG5f34259609f6: [mlir] FuncToLLVM: fail gracefully in bare pointer multi-result covnersion (authored by ftynse).
[mlir] FuncToLLVM: fail gracefully in bare pointer multi-result covnersion
Mon, Mar 27, 7:54 AM · Restricted Project, Restricted Project
ftynse requested review of D146961: [mlir] support external named transform libraries.
Mon, Mar 27, 7:06 AM · Restricted Project, Restricted Project

Fri, Mar 24

ftynse committed rG8b513407407d: [mlir][Vector][Transforms] Improve the control over individual vector lowerings… (authored by nicolasvasilache).
[mlir][Vector][Transforms] Improve the control over individual vector lowerings…
Fri, Mar 24, 7:02 AM · Restricted Project, Restricted Project
ftynse closed D146755: [mlir][Vector][Transforms] Improve the control over individual vector lowerings and transforms.
Fri, Mar 24, 7:01 AM · Restricted Project, Restricted Project
ftynse updated the diff for D146755: [mlir][Vector][Transforms] Improve the control over individual vector lowerings and transforms.

Remove code repetition by turning pattern-based applyToEach into a trait.

Fri, Mar 24, 6:20 AM · Restricted Project, Restricted Project
ftynse commandeered D146755: [mlir][Vector][Transforms] Improve the control over individual vector lowerings and transforms.

Will do some deduplication here.

Fri, Mar 24, 5:39 AM · Restricted Project, Restricted Project
ftynse accepted D146755: [mlir][Vector][Transforms] Improve the control over individual vector lowerings and transforms.
Fri, Mar 24, 4:50 AM · Restricted Project, Restricted Project
ftynse added inline comments to D146755: [mlir][Vector][Transforms] Improve the control over individual vector lowerings and transforms.
Fri, Mar 24, 3:38 AM · Restricted Project, Restricted Project

Thu, Mar 23

ftynse committed rG61f33def1375: [mlir][Vector] Make sure that vector.contract preserves extra attributes while… (authored by apaszke).
[mlir][Vector] Make sure that vector.contract preserves extra attributes while…
Thu, Mar 23, 3:32 AM · Restricted Project, Restricted Project
ftynse closed D146707: [mlir][Vector] Make sure that vector.contract preserves extra attributes while parsing.
Thu, Mar 23, 3:32 AM · Restricted Project, Restricted Project
ftynse accepted D146707: [mlir][Vector] Make sure that vector.contract preserves extra attributes while parsing.
Thu, Mar 23, 3:04 AM · Restricted Project, Restricted Project

Wed, Mar 22

ftynse added inline comments to D146611: [mlir][AffineMap] NFC - Refactor getProjectedMap and split into projectDims and projectSymbols.
Wed, Mar 22, 5:33 AM · Restricted Project, Restricted Project
ftynse added a comment to D146272: [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times.

This looks like an unrelated flang failure

Wed, Mar 22, 4:46 AM · Restricted Project, Restricted Project
ftynse accepted D146495: [MLIR][Affine] Fix assumption on int type in memref elt size method.
Wed, Mar 22, 3:24 AM · Restricted Project, Restricted Project
ftynse accepted D146616: [MLIR][LLVM] Copy byval attributes during inlining..
Wed, Mar 22, 3:21 AM · Restricted Project, Restricted Project
ftynse accepted D146611: [mlir][AffineMap] NFC - Refactor getProjectedMap and split into projectDims and projectSymbols.
Wed, Mar 22, 3:19 AM · Restricted Project, Restricted Project
ftynse accepted D146272: [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times.
Wed, Mar 22, 2:05 AM · Restricted Project, Restricted Project
ftynse accepted D146201: [mlir][Analysis][NFC] Split FlatAffineValueConstraints into multiple classes.

Mostly code motion, didn't go beyond the screening for obvious layering problems.

Wed, Mar 22, 1:55 AM · Restricted Project, Restricted Project

Tue, Mar 21

ftynse committed rGfb409a2822df: [mlir] Transform dialect: add named sequences (authored by ftynse).
[mlir] Transform dialect: add named sequences
Tue, Mar 21, 7:54 AM · Restricted Project, Restricted Project
ftynse closed D146433: [mlir] Transform dialect: add named sequences.
Tue, Mar 21, 7:54 AM · Restricted Project, Restricted Project
ftynse added inline comments to D146433: [mlir] Transform dialect: add named sequences.
Tue, Mar 21, 6:45 AM · Restricted Project, Restricted Project
ftynse updated the diff for D146433: [mlir] Transform dialect: add named sequences.

Address review.

Tue, Mar 21, 6:45 AM · Restricted Project, Restricted Project
ftynse added inline comments to D146495: [MLIR][Affine] Fix assumption on int type in memref elt size method.
Tue, Mar 21, 6:07 AM · Restricted Project, Restricted Project
ftynse accepted D146509: [mlir][Transform] NFC - Add more advanced debug spew to help hunt down potential misuses.
Tue, Mar 21, 5:58 AM · Restricted Project, Restricted Project
ftynse added a comment to D146272: [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times.

The gpu testcase is based on the pass-pipeline of cubin,the current compiler is not supported.

Tue, Mar 21, 5:29 AM · Restricted Project, Restricted Project

Mon, Mar 20

ftynse accepted D145932: [mlir] Support lowering of dialect attributes attached to top-level modules.

Nice, thanks!

Mon, Mar 20, 9:46 AM · Restricted Project, Restricted Project, Restricted Project
ftynse added a comment to D144657: [mlir] Add alloca address space handling to the data layout subsystem.

Sorry for delay, already-approved patches don't show up in the review stream :(

Mon, Mar 20, 9:38 AM · Restricted Project, Restricted Project
ftynse accepted D144657: [mlir] Add alloca address space handling to the data layout subsystem.
Mon, Mar 20, 9:37 AM · Restricted Project, Restricted Project
ftynse accepted D145467: [MLIR][Linalg] Generate unique LibraryCallName for LinalgOps..
Mon, Mar 20, 9:23 AM · Restricted Project, Restricted Project
ftynse requested review of D146433: [mlir] Transform dialect: add named sequences.
Mon, Mar 20, 9:22 AM · Restricted Project, Restricted Project
ftynse requested changes to D145467: [MLIR][Linalg] Generate unique LibraryCallName for LinalgOps..
Mon, Mar 20, 8:58 AM · Restricted Project, Restricted Project
ftynse added inline comments to D146272: [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times.
Mon, Mar 20, 2:27 AM · Restricted Project, Restricted Project

Fri, Mar 17

ftynse retitled D146272: [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times from [github][MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times #59503 to [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times.
Fri, Mar 17, 3:00 AM · Restricted Project, Restricted Project
ftynse requested changes to D146272: [MLIR] Lowering from GPU to LLVM fails when the same kernel is called multiple times.

Please add a test. The IR in the bug is a good start for the test, but can be simplified further.

Fri, Mar 17, 3:00 AM · Restricted Project, Restricted Project

Thu, Mar 16

ftynse committed rG46c8422d833b: [mlir][transform] change RaggedArray internals (authored by ftynse).
[mlir][transform] change RaggedArray internals
Thu, Mar 16, 4:14 PM · Restricted Project, Restricted Project
ftynse closed D146239: [mlir][transform] change RaggedArray internals.
Thu, Mar 16, 4:14 PM · Restricted Project, Restricted Project
ftynse retitled D146239: [mlir][transform] change RaggedArray internals from [mlir][transform] WIP: change RaggedArray internals to [mlir][transform] change RaggedArray internals.
Thu, Mar 16, 10:17 AM · Restricted Project, Restricted Project
ftynse updated the diff for D146239: [mlir][transform] change RaggedArray internals.

Revise

Thu, Mar 16, 10:16 AM · Restricted Project, Restricted Project
ftynse planned changes to D146239: [mlir][transform] change RaggedArray internals.
Thu, Mar 16, 10:00 AM · Restricted Project, Restricted Project
ftynse requested review of D146239: [mlir][transform] change RaggedArray internals.
Thu, Mar 16, 10:00 AM · Restricted Project, Restricted Project

Wed, Mar 15

ftynse accepted D146151: [mlir][Affine] Use stable sorting to prevent non-determinism.
Wed, Mar 15, 9:12 AM · Restricted Project, Restricted Project

Tue, Mar 14

ftynse added a comment to D144988: [mlir][Transform][LinAlg] Add copy permutation for GPU memory hierarchy in the `transform.promote` op.

Could you please rebase?

Tue, Mar 14, 12:53 PM · Restricted Project, Restricted Project
ftynse accepted D144988: [mlir][Transform][LinAlg] Add copy permutation for GPU memory hierarchy in the `transform.promote` op.
Tue, Mar 14, 11:26 AM · Restricted Project, Restricted Project
ftynse requested changes to D145932: [mlir] Support lowering of dialect attributes attached to top-level modules.
Tue, Mar 14, 11:17 AM · Restricted Project, Restricted Project, Restricted Project
ftynse added inline comments to D145467: [MLIR][Linalg] Generate unique LibraryCallName for LinalgOps..
Tue, Mar 14, 10:33 AM · Restricted Project, Restricted Project
ftynse accepted D145948: [MLIR] : Add integer mul in scf to openmp conversion.
Tue, Mar 14, 10:28 AM · Restricted Project, Restricted Project
ftynse committed rGd9db5a5904fd: [mlir] relax value handle updates when operation is replaced (authored by ftynse).
[mlir] relax value handle updates when operation is replaced
Tue, Mar 14, 8:58 AM · Restricted Project, Restricted Project
ftynse closed D145254: [mlir] relax value handle updates when operation is replaced.
Tue, Mar 14, 8:57 AM · Restricted Project, Restricted Project

Mon, Mar 13

ftynse accepted D145977: [mlir][Transform] NFC - Various API cleanups and use RewriterBase in lieu of PatternRewriter.
Mon, Mar 13, 5:43 PM · Restricted Project, Restricted Project
ftynse accepted D145685: [mlir][Affine] Add helper functions to allow reordering affine.apply operands and decompose the ops into smaller components.
Mon, Mar 13, 5:38 PM · Restricted Project, Restricted Project
ftynse accepted D145784: [mlir][DialectUtils] Cleanup IndexingUtils and provide more affine variants while reusing implementations.
Mon, Mar 13, 5:29 PM · Restricted Project, Restricted Project
ftynse accepted D145901: [mlir] Use splitBlock instread of createBlock in GenericAtomicRMWLowering..
Mon, Mar 13, 10:02 AM · Restricted Project, Restricted Project

Sat, Mar 11

ftynse added a comment to D140357: [MLIR] Add readonly,writeonly,noalias to func arg attributes.

There isn't sufficient context to justify the change in the commit description, and the change itself doesn't seem to do anything else than lifting function argument attributes from LLVM IR to the Func dialect level. Yes, there is SPIR-V, but the change does not include the lowering to SPIR-V or even a description of how it can be done. Blindly copying LLVM IR is acceptable in the LLVM dialect, but not at the higher-level abstractions such as Func. It is unclear to me what these attributes would even mean in the full generality of what MLIR type system allows. For example, what is a writeonly tensor? How do two SPIR-V types noalias? How do even two memrefs noalias when the default conversion to LLVM produces two pointers that precisely do alias each other?

Sat, Mar 11, 10:05 AM · Restricted Project, Restricted Project
ftynse added a reviewer for D145829: [mlir][llvm] Add fastmathFlags to SelectOp: gysit.
Sat, Mar 11, 9:40 AM · Restricted Project, Restricted Project
ftynse accepted D145829: [mlir][llvm] Add fastmathFlags to SelectOp.

Could you check if the translation works in the opposite direction, too?

Sat, Mar 11, 9:39 AM · Restricted Project, Restricted Project

Sat, Mar 4

ftynse added a comment to D145288: [mlir][scf] Relax type requirement on for.

Do we need to also update the lowering to LLVM? It may assume type is always index and end up using a different bitwidth.

Sat, Mar 4, 12:18 AM · Restricted Project, Restricted Project

Fri, Mar 3

ftynse requested review of D145254: [mlir] relax value handle updates when operation is replaced.
Fri, Mar 3, 9:30 AM · Restricted Project, Restricted Project

Wed, Mar 1

ftynse added a comment to D144988: [mlir][Transform][LinAlg] Add copy permutation for GPU memory hierarchy in the `transform.promote` op.

LGTM when the last comment is addressed, but I'd like an opinion from @nicolasvasilache on the promotion logic.

Wed, Mar 1, 5:28 PM · Restricted Project, Restricted Project
ftynse added a comment to D145038: [mlir][llvm] Delete LLVMIntrinsicGen.cpp..

We can have test for this, but we probably don't want this to be on the _main_ build path.

Wed, Mar 1, 12:41 PM · Restricted Project, Restricted Project
ftynse accepted D143352: [mlir][python] Add generic operation parse APIs.
Wed, Mar 1, 12:28 PM · Restricted Project, Restricted Project
ftynse added a comment to D144883: [MLIR] Add convertModule functionallity to the LLVMTranslationInterface and ModuleTranslation.

After a little bit of time going over the suggested change of ditching convertModuleOperation and using the existing convertOperation infrastructure, I can see two ways of doing it:

  • Teach the default LLVMTranslationInterface::convertOperation to ignore builtin.module and gpu.module and return success, but this also means removing the error output for not finding a dialect interface for an operation inside of the ModuleTranslation::convertOperation function, otherwise we can hoist this ignore from LLVMTranslationInterface::convertOperation to ModuleTranslation::convertOperation and keep the error emission.
  • Create and register a LLVMTranslationInterface for gpu and builtin dialects that return success and do nothing else.

I think you were meaning the latter, I am happy to do either, or even another alternative if I have misunderstood what you mean!

Wed, Mar 1, 12:21 PM · Restricted Project, Restricted Project
ftynse added inline comments to D144988: [mlir][Transform][LinAlg] Add copy permutation for GPU memory hierarchy in the `transform.promote` op.
Wed, Mar 1, 12:05 PM · Restricted Project, Restricted Project
ftynse accepted D145053: Make it possible to create DenseElementsAttrs with arbitrary shaped types in Python bindings.
Wed, Mar 1, 11:46 AM · Restricted Project, Restricted Project
ftynse added a comment to D145038: [mlir][llvm] Delete LLVMIntrinsicGen.cpp..

Note that many other tools are not directly "used" upstream, mlir-reduce, mlir-parser-fuzzer, etc. They are nevertheless useless.

Wed, Mar 1, 11:40 AM · Restricted Project, Restricted Project
ftynse requested changes to D145038: [mlir][llvm] Delete LLVMIntrinsicGen.cpp..

I don't see why this needs to be deleted. It was used to generate the initial ODS for some intrinsics and is still useful for adding new target-specific ones. It is not used in the build process because the complexity of adding a stage to the build process does not justify it.

Wed, Mar 1, 11:36 AM · Restricted Project, Restricted Project

Feb 28 2023

ftynse added a comment to D144883: [MLIR] Add convertModule functionallity to the LLVMTranslationInterface and ModuleTranslation.

This allows existing dialect modules such as gpu.module to continue to not do dialect specific lowering, allowing them to opt in or out based on their dialects needs.

Feb 28 2023, 8:23 AM · Restricted Project, Restricted Project

Feb 27 2023

ftynse accepted D144851: [mlir][llvm] Add AliasAnalysis and AccessGroup interfaces..
Feb 27 2023, 11:48 AM · Restricted Project, Restricted Project
ftynse accepted D144875: [mlir][llvm] Rename LLVMOpsInterfaces.td to LLVMInterfaces.td (NFC)..
Feb 27 2023, 11:41 AM · Restricted Project, Restricted Project
ftynse accepted D144836: Fix bug in addAffineParallelOpDomain upper bound constraint.
Feb 27 2023, 11:39 AM · Restricted Project, Restricted Project
ftynse accepted D144856: [mlir][Linalg] Refactor transform.structured.pad to separate out hoisting.
Feb 27 2023, 11:38 AM · Restricted Project, Restricted Project
ftynse added a comment to D144883: [MLIR] Add convertModule functionallity to the LLVMTranslationInterface and ModuleTranslation.

Why does it have to be a separate interface function? Module is just the builtin.module operation, it looks like we can just call convertOperation for it and teach the default translation to do nothing with it.

Feb 27 2023, 11:29 AM · Restricted Project, Restricted Project
ftynse committed rG9a29d8753884: [mlir][sparse] Add checking parent op of SortOp (authored by sotoshigoto).
[mlir][sparse] Add checking parent op of SortOp
Feb 27 2023, 7:37 AM · Restricted Project, Restricted Project
ftynse closed D143874: [mlir][sparse] Add checking parent op of SortOp.
Feb 27 2023, 7:37 AM · Restricted Project, Restricted Project
ftynse committed rGb46e78c7cb11: [mlir][NFC] Cleanup Passes documentation (authored by sotoshigoto).
[mlir][NFC] Cleanup Passes documentation
Feb 27 2023, 7:36 AM · Restricted Project, Restricted Project
ftynse closed D142868: [mlir][NFC] Cleanup Passes documentation.
Feb 27 2023, 7:35 AM · Restricted Project, Restricted Project
ftynse committed rG115711c19cd2: [mlir][LinAlg][Transform][GPU] Add GPU memory hierarchy to the transform. (authored by tavakkoliamirmohammad).
[mlir][LinAlg][Transform][GPU] Add GPU memory hierarchy to the transform.
Feb 27 2023, 7:34 AM · Restricted Project, Restricted Project
ftynse closed D144666: [mlir][LinAlg][Transform][GPU] Add GPU memory hierarchy to the transform.promote op.
Feb 27 2023, 7:34 AM · Restricted Project, Restricted Project

Feb 25 2023

ftynse accepted D144666: [mlir][LinAlg][Transform][GPU] Add GPU memory hierarchy to the transform.promote op.
Feb 25 2023, 6:41 AM · Restricted Project, Restricted Project

Feb 24 2023

ftynse requested changes to D144666: [mlir][LinAlg][Transform][GPU] Add GPU memory hierarchy to the transform.promote op.

Nice, almost there, thanks you!

Feb 24 2023, 11:54 AM · Restricted Project, Restricted Project

Feb 23 2023

ftynse requested changes to D144666: [mlir][LinAlg][Transform][GPU] Add GPU memory hierarchy to the transform.promote op.
Feb 23 2023, 2:31 PM · Restricted Project, Restricted Project

Feb 21 2023

ftynse added a comment to D144330: [MLIR][RISCV] Emit target-abi info for RISC-V.

Could we have a test?

Feb 21 2023, 1:33 PM · Restricted Project, Restricted Project
ftynse requested review of D144515: [mlir] make some structured transform ops use types.
Feb 21 2023, 1:05 PM · Restricted Project, Restricted Project
ftynse accepted D144479: [mlir][LLVM] Don't make `LLVM_IntPtrBase` a `BuildableType` to allow the use of opaque pointers.
Feb 21 2023, 10:16 AM · Restricted Project, Restricted Project

Feb 16 2023

ftynse added a comment to D144108: [mlir][LinAlg][Transform] Add a transform op for conv2d to im2col.

These patterns look very similar, any chance they can be generalized somehow?

Feb 16 2023, 12:37 AM · Restricted Project, Restricted Project

Feb 15 2023

ftynse accepted D144087: [mlir][transform] Add transform.get_result op.
Feb 15 2023, 3:41 AM · Restricted Project, Restricted Project
ftynse added inline comments to D144087: [mlir][transform] Add transform.get_result op.
Feb 15 2023, 3:41 AM · Restricted Project, Restricted Project
ftynse accepted D142634: [mlir] Add vectorize_nd_extract attribute to masked_vectorize.
Feb 15 2023, 12:09 AM · Restricted Project, Restricted Project