Page MenuHomePhabricator

ftynse (Alex Zinenko)
User

Projects

User does not belong to any projects.

User Details

User Since
Jun 28 2018, 2:24 AM (125 w, 4 d)

Recent Activity

Today

ftynse updated the diff for D91982: [mlir] Add conversion from SCF parallel loops to OpenMP.

Address review

Mon, Nov 23, 2:32 PM · Restricted Project
ftynse committed rGf7d033f4d80f: [mlir] Support WsLoopOp in OpenMP to LLVM dialect conversion (authored by ftynse).
[mlir] Support WsLoopOp in OpenMP to LLVM dialect conversion
Mon, Nov 23, 2:28 PM
ftynse closed D91989: [mlir] Support WsLoopOp in OpenMP to LLVM dialect conversion.
Mon, Nov 23, 2:28 PM · Restricted Project
ftynse added a comment to D91905: Use MlirStringRef throughout the C API.

FYI: your email address wasn't picked up by git when landing - Author: George <>

Mon, Nov 23, 2:27 PM · Restricted Project
ftynse added reviewers for D91989: [mlir] Support WsLoopOp in OpenMP to LLVM dialect conversion: kiranchandramohan, SouraVX, kiranktp, chelini, wsmoses.
Mon, Nov 23, 11:59 AM · Restricted Project
ftynse requested review of D91989: [mlir] Support WsLoopOp in OpenMP to LLVM dialect conversion.
Mon, Nov 23, 11:58 AM · Restricted Project
ftynse updated the diff for D91982: [mlir] Add conversion from SCF parallel loops to OpenMP.

Address (most of) the review

Mon, Nov 23, 11:29 AM · Restricted Project
ftynse updated the diff for D91982: [mlir] Add conversion from SCF parallel loops to OpenMP.

Forgotten git add

Mon, Nov 23, 11:26 AM · Restricted Project
ftynse added reviewers for D91982: [mlir] Add conversion from SCF parallel loops to OpenMP: kiranchandramohan, SouraVX, kiranktp, wsmoses, chelini.
Mon, Nov 23, 11:00 AM · Restricted Project
ftynse updated the diff for D91982: [mlir] Add conversion from SCF parallel loops to OpenMP.

Sprinkle around a bit more documentation

Mon, Nov 23, 11:00 AM · Restricted Project
ftynse updated the diff for D91982: [mlir] Add conversion from SCF parallel loops to OpenMP.

Add cmake

Mon, Nov 23, 10:54 AM · Restricted Project
ftynse requested review of D91982: [mlir] Add conversion from SCF parallel loops to OpenMP.
Mon, Nov 23, 10:45 AM · Restricted Project
ftynse committed rG619630f997f1: [mlir] Temporarily disable flaky mlir-cpu-runner async tests (authored by ftynse).
[mlir] Temporarily disable flaky mlir-cpu-runner async tests
Mon, Nov 23, 7:53 AM
ftynse committed rG31a233d46367: [mlir] canonicalize away zero-iteration SCF for loops (authored by ftynse).
[mlir] canonicalize away zero-iteration SCF for loops
Mon, Nov 23, 6:04 AM
ftynse closed D91880: [mlir] canonicalize away zero-iteration SCF for loops.
Mon, Nov 23, 6:04 AM · Restricted Project
ftynse added inline comments to D91652: [mlir] Make attributes mutable in Python bindings.
Mon, Nov 23, 5:22 AM · Restricted Project
ftynse requested review of D91961: [mlir] move lib/Bindings/Python/Attributes.td to include/mlir/Bindings/Python.
Mon, Nov 23, 5:21 AM · Restricted Project
ftynse accepted D91738: Add userData to the diagnostic handler C API.

I have worked with many C APIs in a number of languages and the attitude of "just do the memory management bookkeeping in the host language" has created some very unfortunate complexity in the bindings and more than a few bugs. This is a very low-cost change for the C bindings, and doesn't add much complexity for language bindings that don't care about this functionality (they can just pass NULL to the extra arguments). I think this is also a win from a self-documentation standpoint: before, I would need to go and figure out what are the different ways diagnostic handlers can be unregistered to understand the semantics. I would also need to worry about concurrency, specifically whether or not my handler is guaranteed to not be called after detach returns. I could read the code to figure this out (it turns out, the answer is "no") but I think it is better to minimize this sort of spelunking and just say "we will call your handler and when we are done calling your handler we will call your callback".

Mon, Nov 23, 5:07 AM · Restricted Project
ftynse added inline comments to D91652: [mlir] Make attributes mutable in Python bindings.
Mon, Nov 23, 5:05 AM · Restricted Project
ftynse committed rG1ec60862d702: [mlir] Avoid cloning ops in SCF parallel conversion to CFG (authored by ftynse).
[mlir] Avoid cloning ops in SCF parallel conversion to CFG
Mon, Nov 23, 5:01 AM
ftynse closed D91955: [mlir] Avoid cloning ops in SCF parallel conversion to CFG.
Mon, Nov 23, 5:01 AM · Restricted Project
ftynse requested review of D91959: [mlir] use STATUS instead of CHECK_* in MLIRDetectPythonEnv.cmake.
Mon, Nov 23, 4:59 AM · Restricted Project
ftynse updated the diff for D91880: [mlir] canonicalize away zero-iteration SCF for loops.

Address review

Mon, Nov 23, 4:57 AM · Restricted Project
ftynse accepted D91956: [mlir][Linalg] Drop symbol_source abstraction which does not pay for itself..
Mon, Nov 23, 4:30 AM · Restricted Project
ftynse added a reviewer for D91955: [mlir] Avoid cloning ops in SCF parallel conversion to CFG: pifon2a.
Mon, Nov 23, 4:23 AM · Restricted Project
ftynse requested review of D91955: [mlir] Avoid cloning ops in SCF parallel conversion to CFG.
Mon, Nov 23, 4:23 AM · Restricted Project
ftynse added inline comments to D91881: [MLIR][SPIRVToLLVM] Updated documentation on entry points and not supported ops.
Mon, Nov 23, 1:38 AM · Restricted Project
ftynse added a comment to D91905: Use MlirStringRef throughout the C API.

Sure, let's use MlirStringRef everywhere, this also indicates that we don't take ownership of the string. We added it much later than APIs that could be using it and nobody had time to reconsider those yet.

Mon, Nov 23, 1:36 AM · Restricted Project
ftynse accepted D91863: [mlir][Linalg] NFC - Factor out Linalg functionality for shape and loop bounds computation.
Mon, Nov 23, 1:17 AM · Restricted Project
ftynse abandoned D87096: [mlir] use storage uniquer for side effects.

Replaced with another approach - https://reviews.llvm.org/D91493.

Mon, Nov 23, 12:52 AM · Restricted Project

Fri, Nov 20

ftynse requested review of D91880: [mlir] canonicalize away zero-iteration SCF for loops.
Fri, Nov 20, 10:22 AM · Restricted Project
ftynse committed rGf5c5fd1c50bf: [MLIR] Correct block merge bug (authored by William S. Moses <gh@wsmoses.com>).
[MLIR] Correct block merge bug
Fri, Nov 20, 10:13 AM
ftynse closed D91745: [MLIR] Correct block merge bug.
Fri, Nov 20, 10:13 AM · Restricted Project
ftynse committed rG18d0f7d5c3b0: [mlir] add canonicalization patterns for trivial SCF 'for' and 'if' (authored by ftynse).
[mlir] add canonicalization patterns for trivial SCF 'for' and 'if'
Fri, Nov 20, 10:05 AM
ftynse closed D91865: [mlir] add canonicalization patterns for trivial SCF 'for' and 'if'.
Fri, Nov 20, 10:04 AM · Restricted Project
ftynse added inline comments to D91865: [mlir] add canonicalization patterns for trivial SCF 'for' and 'if'.
Fri, Nov 20, 9:51 AM · Restricted Project
ftynse updated the diff for D91865: [mlir] add canonicalization patterns for trivial SCF 'for' and 'if'.

Address review.

Fri, Nov 20, 9:51 AM · Restricted Project
ftynse requested review of D91865: [mlir] add canonicalization patterns for trivial SCF 'for' and 'if'.
Fri, Nov 20, 5:43 AM · Restricted Project
ftynse accepted D91856: [mlir][std] Fold comparisons when the operands are equal.
Fri, Nov 20, 2:50 AM · Restricted Project
ftynse accepted D91854: [mlir][std] Canonicalize a dim(memref_reshape) into a load from the shape operand.
Fri, Nov 20, 2:30 AM · Restricted Project
ftynse added a comment to D91738: Add userData to the diagnostic handler C API.

Shared pointer logic looks fine: the lambda takes it by-copy and is stored internally in std::function, so the data remains live.

Fri, Nov 20, 1:48 AM · Restricted Project

Thu, Nov 19

ftynse committed rG9bb5bff57014: [mlir] Add an assertion on creating an Operation with null result types (authored by ftynse).
[mlir] Add an assertion on creating an Operation with null result types
Thu, Nov 19, 1:29 PM
ftynse closed D91770: [mlir] Add an assertion on creating an Operation with null result types.
Thu, Nov 19, 1:28 PM · Restricted Project
ftynse added inline comments to D91770: [mlir] Add an assertion on creating an Operation with null result types.
Thu, Nov 19, 10:36 AM · Restricted Project
ftynse updated the diff for D91770: [mlir] Add an assertion on creating an Operation with null result types.

Add a check in TypeRange

Thu, Nov 19, 10:35 AM · Restricted Project
ftynse accepted D91740: Make array pointers in the CAPI const.
Thu, Nov 19, 4:50 AM · Restricted Project
ftynse added inline comments to D91770: [mlir] Add an assertion on creating an Operation with null result types.
Thu, Nov 19, 1:58 AM · Restricted Project
ftynse requested review of D91770: [mlir] Add an assertion on creating an Operation with null result types.
Thu, Nov 19, 1:55 AM · Restricted Project
ftynse added a reviewer for D91745: [MLIR] Correct block merge bug: rriddle.
Thu, Nov 19, 1:08 AM · Restricted Project
ftynse added a comment to D91745: [MLIR] Correct block merge bug.

LGTM, forwarding to @rriddle as I am not very familiar with this code.

Thu, Nov 19, 1:08 AM · Restricted Project
ftynse committed rG58ce4a8b1190: [mlir][TableGen] Support intrinsics with multiple returns and overloaded… (authored by jikimjikim).
[mlir][TableGen] Support intrinsics with multiple returns and overloaded…
Thu, Nov 19, 1:00 AM
ftynse closed D91680: [mlir][TableGen] Support intrinsics with multiple returns and overloaded operands..
Thu, Nov 19, 12:59 AM · Restricted Project
ftynse accepted D91680: [mlir][TableGen] Support intrinsics with multiple returns and overloaded operands..

Thanks for fixing and iterating!

Thu, Nov 19, 12:52 AM · Restricted Project

Wed, Nov 18

ftynse added a comment to D91556: Add capturing of parameters to pass to omp::parallel.

Thanks, this looks better! I have a couple of further comments.

Wed, Nov 18, 3:46 AM · Restricted Project
ftynse requested changes to D91680: [mlir][TableGen] Support intrinsics with multiple returns and overloaded operands..
Wed, Nov 18, 3:32 AM · Restricted Project
ftynse added a comment to D91680: [mlir][TableGen] Support intrinsics with multiple returns and overloaded operands..

This broke some tests.

Wed, Nov 18, 3:32 AM · Restricted Project
ftynse committed rG052d24af2913: [mlir] Introduce support for parametric side-effects (authored by ftynse).
[mlir] Introduce support for parametric side-effects
Wed, Nov 18, 1:52 AM
ftynse closed D91493: [mlir] Introduce support for parametric side-effects.
Wed, Nov 18, 1:52 AM · Restricted Project

Tue, Nov 17

ftynse requested changes to D91556: Add capturing of parameters to pass to omp::parallel.
Tue, Nov 17, 3:43 PM · Restricted Project
ftynse accepted D91599: [mlir] Simplify code generated by ConvertToLLVMPattern::getStridedElementPtr()..

Thanks, looks much simpler!

Tue, Nov 17, 3:37 PM · Restricted Project
ftynse accepted D91581: [mlir] Add support for referencing a SymbolRefAttr in a SideEffectInstance.
Tue, Nov 17, 3:31 PM · Restricted Project
ftynse accepted D90994: [mlir] [sparse] start of sparse tensor compiler support.
Tue, Nov 17, 12:53 PM · Restricted Project
ftynse added inline comments to D90994: [mlir] [sparse] start of sparse tensor compiler support.
Tue, Nov 17, 11:41 AM · Restricted Project
ftynse requested review of D91652: [mlir] Make attributes mutable in Python bindings.
Tue, Nov 17, 11:23 AM · Restricted Project
ftynse committed rGf3dab16dc721: [mlir] Add a _get_default_loc_context utility to Python bindings (authored by ftynse).
[mlir] Add a _get_default_loc_context utility to Python bindings
Tue, Nov 17, 8:56 AM
ftynse closed D91623: [mlir] Add a _get_default_loc_context utility to Python bindings.
Tue, Nov 17, 8:55 AM · Restricted Project
ftynse accepted D91594: [mlir] Remove unused ConvertToLLVMPattern::linearizeSubscripts()..
Tue, Nov 17, 5:57 AM · Restricted Project
ftynse accepted D91549: [mlir] Simplify std.alloc lowering to LLVM..

Nice!

Tue, Nov 17, 5:45 AM · Restricted Project
ftynse added a comment to D91410: [llvm][clang][mlir] Add checks for the return values from Target::createXXX to prevent protential null deref.

LGTM for the MLIR part

Tue, Nov 17, 5:36 AM · Restricted Project, Restricted Project, Restricted Project
ftynse accepted D91283: [mlir][Python] Make DenseElementsAttr loading be int size agnostic..

Thanks for fixing!

Tue, Nov 17, 5:35 AM · Restricted Project
ftynse accepted D91621: [mlir][std] Canonicalize extract_element(tensor_cast)..
Tue, Nov 17, 5:33 AM · Restricted Project
ftynse accepted D91616: [mlir][std] Fold dim(dynamic_tensor_from_elements, %cst).
Tue, Nov 17, 5:32 AM · Restricted Project
ftynse committed rG88f25bda1376: [mlir] Allow for using interface class name in ODS interface definitions (authored by ftynse).
[mlir] Allow for using interface class name in ODS interface definitions
Tue, Nov 17, 5:29 AM
ftynse closed D91499: [mlir] Allow for using interface class name in ODS interface definitions.
Tue, Nov 17, 5:29 AM · Restricted Project
ftynse requested review of D91623: [mlir] Add a _get_default_loc_context utility to Python bindings.
Tue, Nov 17, 5:19 AM · Restricted Project
ftynse committed rGef8e859c0b3c: [mlir] Fix Python tests after "module_terminator" migrated to ODS (authored by ftynse).
[mlir] Fix Python tests after "module_terminator" migrated to ODS
Tue, Nov 17, 5:17 AM
ftynse added inline comments to D91542: [mlir] Add basic support for attributes in ODS-generated Python bindings.
Tue, Nov 17, 3:27 AM · Restricted Project
ftynse committed rGc5a6712f8cee: [mlir] Add basic support for attributes in ODS-generated Python bindings (authored by ftynse).
[mlir] Add basic support for attributes in ODS-generated Python bindings
Tue, Nov 17, 2:48 AM
ftynse closed D91542: [mlir] Add basic support for attributes in ODS-generated Python bindings.
Tue, Nov 17, 2:47 AM · Restricted Project
ftynse updated the diff for D91542: [mlir] Add basic support for attributes in ODS-generated Python bindings.

Address review.

Tue, Nov 17, 2:41 AM · Restricted Project

Mon, Nov 16

ftynse updated the diff for D91542: [mlir] Add basic support for attributes in ODS-generated Python bindings.

Customize handling of unit attributes

Mon, Nov 16, 10:10 AM · Restricted Project
ftynse updated the diff for D91542: [mlir] Add basic support for attributes in ODS-generated Python bindings.

Simplify cmake further

Mon, Nov 16, 9:17 AM · Restricted Project
ftynse requested review of D91542: [mlir] Add basic support for attributes in ODS-generated Python bindings.
Mon, Nov 16, 7:34 AM · Restricted Project
ftynse accepted D91523: [mlir] Require std.alloc() ops to have canonical layout during LLVM lowering..
Mon, Nov 16, 5:01 AM · Restricted Project

Sun, Nov 15

ftynse requested review of D91499: [mlir] Allow for using interface class name in ODS interface definitions.
Sun, Nov 15, 8:50 AM · Restricted Project
ftynse requested review of D91493: [mlir] Introduce support for parametric side-effects.
Sun, Nov 15, 6:44 AM · Restricted Project

Fri, Nov 13

ftynse committed rGc9e9cc3fe700: [MLIR] Allow setting "CodeView" flag in LLVMIR translation on MSVC. (authored by scotttodd).
[MLIR] Allow setting "CodeView" flag in LLVMIR translation on MSVC.
Fri, Nov 13, 8:31 AM
ftynse closed D91365: [MLIR] Allow setting "CodeView" flag in LLVMIR translation on MSVC..
Fri, Nov 13, 8:31 AM · Restricted Project
ftynse accepted D91430: [mlir] NFC: tiny fix in comment..
Fri, Nov 13, 8:29 AM · Restricted Project
ftynse accepted D91365: [MLIR] Allow setting "CodeView" flag in LLVMIR translation on MSVC..

Thanks!

Fri, Nov 13, 1:56 AM · Restricted Project

Thu, Nov 12

ftynse accepted D91374: [MLIR] Fix standard -> LLVM conversion to fail for unsupported memref element type..
Thu, Nov 12, 3:19 PM · Restricted Project
ftynse added a comment to D90994: [mlir] [sparse] start of sparse tensor compiler support.

Thanks Aart!

Thu, Nov 12, 3:14 PM · Restricted Project
ftynse added a comment to D91365: [MLIR] Allow setting "CodeView" flag in LLVMIR translation on MSVC..

There's potential for a more generic solution here. I tried passing a ArrayRef<llvm::Module::ModuleFlagEntry> through: https://github.com/ScottTodd/llvm-project/commit/e276941578e8afd1f77380eceb9d76cd1b11b41c , but I think something using attributes on the Operation might be better. For projects not using memrefs, the "malloc" and "free" functions could also be optional.

Thu, Nov 12, 10:34 AM · Restricted Project
ftynse accepted D91340: [mlir][gpu] Only transform mapped parallel loops to GPU..

Please fix the test

Thu, Nov 12, 8:20 AM · Restricted Project
ftynse committed rGf9265de8c634: [mlir] Generate Op builders for Python bindings (authored by ftynse).
[mlir] Generate Op builders for Python bindings
Thu, Nov 12, 2:29 AM
ftynse closed D91190: [mlir] Generate Op builders for Python bindings.
Thu, Nov 12, 2:29 AM · Restricted Project

Tue, Nov 10

ftynse requested review of D91190: [mlir] Generate Op builders for Python bindings.
Tue, Nov 10, 10:27 AM · Restricted Project
ftynse accepted D91166: [mlir][llvm] Expose getters for alias and align attribute names.
Tue, Nov 10, 9:22 AM · Restricted Project
ftynse accepted D90823: Add Python binding to run a PassManager on a MLIR Module.
Tue, Nov 10, 3:07 AM · Restricted Project
ftynse added inline comments to D90960: [mlir] ODS-backed python binding generator for custom op classes.
Tue, Nov 10, 1:58 AM · Restricted Project