Page MenuHomePhabricator

aartbik (Aart Bik)
PhD

Projects

User does not belong to any projects.

User Details

User Since
Jan 10 2020, 10:13 AM (45 w, 3 d)

Aart J.C. Bik received his PhD from the Leiden University in 1996. From 1998 to 2007, he worked at Intel, where he was the lead compiler architect of automatic vectorization in the Intel C++/Fortran compilers. In 2007, he moved to Google, where he has worked on compilers for ART and the Dart VM. He is now a member of the TensorFlow codegen team.

Recent Activity

Today

aartbik added a reviewer for D91978: [mlir][sparse] add parallelization strategies to sparse compiler: ftynse.
Mon, Nov 23, 3:53 PM · Restricted Project
aartbik updated the summary of D91985: [mlir][sparse] generalize invariant expression handling in sparse compiler.
Mon, Nov 23, 2:54 PM · Restricted Project
aartbik added reviewers for D91985: [mlir][sparse] generalize invariant expression handling in sparse compiler: penpornk, mehdi_amini, tatianashp.
Mon, Nov 23, 11:16 AM · Restricted Project
aartbik added a comment to D91978: [mlir][sparse] add parallelization strategies to sparse compiler.

For ease of reviewing, I split out the invariant generalization in its own review: https://reviews.llvm.org/D91985

Mon, Nov 23, 11:16 AM · Restricted Project
aartbik updated the summary of D91978: [mlir][sparse] add parallelization strategies to sparse compiler.
Mon, Nov 23, 11:15 AM · Restricted Project
aartbik requested review of D91985: [mlir][sparse] generalize invariant expression handling in sparse compiler.
Mon, Nov 23, 11:14 AM · Restricted Project
aartbik added reviewers for D91978: [mlir][sparse] add parallelization strategies to sparse compiler: ezhulenev, penpornk, mehdi_amini, tatianashp.
Mon, Nov 23, 10:03 AM · Restricted Project
aartbik requested review of D91978: [mlir][sparse] add parallelization strategies to sparse compiler.
Mon, Nov 23, 10:01 AM · Restricted Project

Fri, Nov 20

aartbik added inline comments to D91896: [mlir] Add microbenchmark for linalg+async-parallel-for.
Fri, Nov 20, 6:09 PM · Restricted Project
aartbik added a comment to D91896: [mlir] Add microbenchmark for linalg+async-parallel-for.

Nice to see this in action! Right in time also for some parallel work in the sparse compiler!

Fri, Nov 20, 5:21 PM · Restricted Project
aartbik committed rGaf42550523d9: [mlir][sparse] refine optimization, add few more test cases (authored by aartbik).
[mlir][sparse] refine optimization, add few more test cases
Fri, Nov 20, 5:02 PM
aartbik closed D91818: [mlir][sparse] refine optimization, add few more test cases.
Fri, Nov 20, 5:02 PM · Restricted Project
aartbik updated the diff for D91818: [mlir][sparse] refine optimization, add few more test cases.

renamed A BC into S AB

Fri, Nov 20, 4:34 PM · Restricted Project
aartbik added inline comments to D91818: [mlir][sparse] refine optimization, add few more test cases.
Fri, Nov 20, 4:31 PM · Restricted Project
aartbik accepted D91321: [mlir][vector] Add transfer_op LoadToStore forwarding and deadStore optimizations.
Fri, Nov 20, 11:24 AM · Restricted Project

Thu, Nov 19

aartbik added a reviewer for D91818: [mlir][sparse] refine optimization, add few more test cases: penpornk.
Thu, Nov 19, 1:14 PM · Restricted Project
aartbik requested review of D91818: [mlir][sparse] refine optimization, add few more test cases.
Thu, Nov 19, 1:13 PM · Restricted Project
aartbik added inline comments to D91321: [mlir][vector] Add transfer_op LoadToStore forwarding and deadStore optimizations.
Thu, Nov 19, 10:54 AM · Restricted Project

Wed, Nov 18

aartbik committed rG9ad62f62b9ad: [mlir][sparse] remove a few rewriting failures (authored by aartbik).
[mlir][sparse] remove a few rewriting failures
Wed, Nov 18, 5:30 PM
aartbik closed D91748: [mlir][sparse] remove a few rewriting failures.
Wed, Nov 18, 5:29 PM · Restricted Project
aartbik added inline comments to D91748: [mlir][sparse] remove a few rewriting failures.
Wed, Nov 18, 5:27 PM · Restricted Project
aartbik added a reviewer for D91748: [mlir][sparse] remove a few rewriting failures: penpornk.
Wed, Nov 18, 3:49 PM · Restricted Project
aartbik requested review of D91748: [mlir][sparse] remove a few rewriting failures.
Wed, Nov 18, 3:48 PM · Restricted Project

Tue, Nov 17

aartbik committed rGeced4a8e6fe3: [mlir] [sparse] start of sparse tensor compiler support (authored by aartbik).
[mlir] [sparse] start of sparse tensor compiler support
Tue, Nov 17, 1:11 PM
aartbik closed D90994: [mlir] [sparse] start of sparse tensor compiler support.
Tue, Nov 17, 1:10 PM · Restricted Project
aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

alloc -> alloca, added a few more comments in top description

Tue, Nov 17, 12:18 PM · Restricted Project
aartbik added inline comments to D90994: [mlir] [sparse] start of sparse tensor compiler support.
Tue, Nov 17, 12:06 PM · Restricted Project
aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

addressed remaining comments (except builders)
also rebased with the new Affine utils that make for shorter code

Tue, Nov 17, 11:06 AM · Restricted Project
aartbik added a comment to D90994: [mlir] [sparse] start of sparse tensor compiler support.

I addressed your remaining comments, except that I am pushing back (for now) on the insertion point vs. builder issue. Most obvious, since we don't have such a builder for while yet. But also, I am not convinced it will necessarily make for clearer code. Right now, every method leaves the insertion "cursor" at the right place for the next, even if that cursor can be something different (inside for, while, or if). I am not sure builders will improve on that, but could be convinced otherwise by trying once we actually have those abstractions. But let's not keep stalling this CL by trying to obtain a perfection that is also not met in other parts yet ;-)

Tue, Nov 17, 11:04 AM · Restricted Project

Mon, Nov 16

aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

broke up large genStmt() method into many genXXX() methods

Mon, Nov 16, 8:54 PM · Restricted Project
aartbik added inline comments to D90994: [mlir] [sparse] start of sparse tensor compiler support.
Mon, Nov 16, 8:42 PM · Restricted Project
aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

rebased with new built-in annotations

Mon, Nov 16, 10:57 AM · Restricted Project

Fri, Nov 13

aartbik committed rG9ddb464d37b0: [mlir] refactor common idiom into AffineMap method (authored by aartbik).
[mlir] refactor common idiom into AffineMap method
Fri, Nov 13, 7:18 PM
aartbik closed D91465: [mlir] refactor common idiom into AffineMap method.
Fri, Nov 13, 7:18 PM · Restricted Project
aartbik updated the diff for D91465: [mlir] refactor common idiom into AffineMap method.

addressed comments

Fri, Nov 13, 6:12 PM · Restricted Project
aartbik added a comment to D91465: [mlir] refactor common idiom into AffineMap method.

Thanks Medhi.

Fri, Nov 13, 6:09 PM · Restricted Project
aartbik added inline comments to D91321: [mlir][vector] Add transfer_op LoadToStore forwarding and deadStore optimizations.
Fri, Nov 13, 4:31 PM · Restricted Project
aartbik added inline comments to D84293: Add an assertion in SmallVector::push_back().
Fri, Nov 13, 3:13 PM · Restricted Project, Restricted Project
aartbik updated the summary of D91465: [mlir] refactor common idiom into AffineMap method.
Fri, Nov 13, 3:06 PM · Restricted Project
aartbik requested review of D91465: [mlir] refactor common idiom into AffineMap method.
Fri, Nov 13, 3:04 PM · Restricted Project

Thu, Nov 12

aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

fixed typo

Thu, Nov 12, 8:50 PM · Restricted Project
aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

addressed Alex' comments

Thu, Nov 12, 8:34 PM · Restricted Project
aartbik added a comment to D90994: [mlir] [sparse] start of sparse tensor compiler support.

Thanks Aart!

I did one pass, have a bunch of nits to make this look more MLIResque in style. I agree with Nicolas that we should think about how to connect this better to the infrastructure and avoid a monolithic codegen that will grow huge over time (the same rationale as for affine code generation could apply). My knowledge of the TACO paper is residual at this point, so it's not always easy to track what is happening in the code the high level. Maybe having a relatively short comment blob that explains the algorithm without referring to the paper could be a good way for us to crystallize the understanding on how this can be fitted into MLIR. That being said, I wouldn't oppose landing a version that is correct + tests, and then iterating on refactoring and generalization while keeping the tests green.

Thu, Nov 12, 8:26 PM · Restricted Project
aartbik added inline comments to D90994: [mlir] [sparse] start of sparse tensor compiler support.
Thu, Nov 12, 8:10 PM · Restricted Project
aartbik added inline comments to D90995: [mlir][vector] Support N-D vector in InsertMap/ExtractMap op.
Thu, Nov 12, 1:57 PM · Restricted Project
aartbik accepted D90995: [mlir][vector] Support N-D vector in InsertMap/ExtractMap op.
Thu, Nov 12, 10:28 AM · Restricted Project

Wed, Nov 11

aartbik committed rG08466596485d: [mlir][sparse] export sparse tensor runtime support through header file (authored by aartbik).
[mlir][sparse] export sparse tensor runtime support through header file
Wed, Nov 11, 9:04 PM
aartbik closed D91316: [mlir][sparse] export sparse tensor runtime support through header file.
Wed, Nov 11, 9:04 PM · Restricted Project
aartbik added a comment to D91316: [mlir][sparse] export sparse tensor runtime support through header file.

Do you have tests for all this?

Wed, Nov 11, 8:36 PM · Restricted Project
aartbik updated the diff for D91316: [mlir][sparse] export sparse tensor runtime support through header file.

addressed missing "for"

Wed, Nov 11, 8:34 PM · Restricted Project
aartbik added inline comments to D91316: [mlir][sparse] export sparse tensor runtime support through header file.
Wed, Nov 11, 8:32 PM · Restricted Project
aartbik added inline comments to D91316: [mlir][sparse] export sparse tensor runtime support through header file.
Wed, Nov 11, 6:57 PM · Restricted Project
aartbik added reviewers for D91316: [mlir][sparse] export sparse tensor runtime support through header file: penpornk, tatianashp.
Wed, Nov 11, 6:51 PM · Restricted Project
aartbik requested review of D91316: [mlir][sparse] export sparse tensor runtime support through header file.
Wed, Nov 11, 6:50 PM · Restricted Project
aartbik committed rGe1dbc25ee231: [mlir][sparse] integrate sparse annotation into generic linalg op (authored by aartbik).
[mlir][sparse] integrate sparse annotation into generic linalg op
Wed, Nov 11, 5:27 PM
aartbik closed D91224: [mlir][sparse] integrate sparse annotation into generic linalg op.
Wed, Nov 11, 5:27 PM · Restricted Project
aartbik updated the diff for D91224: [mlir][sparse] integrate sparse annotation into generic linalg op.

minor cleanup

Wed, Nov 11, 4:39 PM · Restricted Project
aartbik updated the diff for D91224: [mlir][sparse] integrate sparse annotation into generic linalg op.

addresses comments, no more magic strings

Wed, Nov 11, 4:31 PM · Restricted Project
aartbik added inline comments to D91224: [mlir][sparse] integrate sparse annotation into generic linalg op.
Wed, Nov 11, 3:43 PM · Restricted Project

Tue, Nov 10

aartbik added reviewers for D91224: [mlir][sparse] integrate sparse annotation into generic linalg op: penpornk, tatianashp.
Tue, Nov 10, 6:12 PM · Restricted Project
aartbik requested review of D91224: [mlir][sparse] integrate sparse annotation into generic linalg op.
Tue, Nov 10, 6:10 PM · Restricted Project
aartbik added a comment to D90995: [mlir][vector] Support N-D vector in InsertMap/ExtractMap op.

few more nits

Tue, Nov 10, 10:29 AM · Restricted Project

Mon, Nov 9

aartbik added inline comments to D90995: [mlir][vector] Support N-D vector in InsertMap/ExtractMap op.
Mon, Nov 9, 11:05 AM · Restricted Project
aartbik added reviewers for D90994: [mlir] [sparse] start of sparse tensor compiler support: gkestor, ftynse.
Mon, Nov 9, 10:36 AM · Restricted Project
aartbik added a comment to D90994: [mlir] [sparse] start of sparse tensor compiler support.

Thanks for pushing on this Aart!
I realize this may sound like a lot of things that have to come together to get to a similar functionality but written in a more progressive and composable fashion.

Mon, Nov 9, 10:34 AM · Restricted Project
aartbik added inline comments to D90994: [mlir] [sparse] start of sparse tensor compiler support.
Mon, Nov 9, 9:24 AM · Restricted Project

Fri, Nov 6

aartbik updated the diff for D90994: [mlir] [sparse] start of sparse tensor compiler support.

fixed lint issue

Fri, Nov 6, 8:07 PM · Restricted Project
aartbik added reviewers for D90994: [mlir] [sparse] start of sparse tensor compiler support: tatianashp, penpornk, reidtatge, mehdi_amini, silvas, ThomasRaoux, stellaraccident.
Fri, Nov 6, 6:05 PM · Restricted Project
aartbik requested review of D90994: [mlir] [sparse] start of sparse tensor compiler support.
Fri, Nov 6, 5:59 PM · Restricted Project

Tue, Nov 3

aartbik accepted D90603: [mlir] Add lowering to CFG for WhileOp.
Tue, Nov 3, 10:25 AM · Restricted Project

Mon, Nov 2

aartbik accepted D90604: [mlir] Add a simpler lowering pattern for WhileOp representing a do-while loop.

LGTM once the others are approved

Mon, Nov 2, 5:44 PM · Restricted Project
aartbik added inline comments to D90603: [mlir] Add lowering to CFG for WhileOp.
Mon, Nov 2, 5:43 PM · Restricted Project

Thu, Oct 29

aartbik accepted D90255: [mlir] Add a generic while/do-while loop to the SCF dialect.

lowering next? ;-)

Thu, Oct 29, 10:08 AM · Restricted Project
aartbik added inline comments to D90352: Introduce a Bazel build configuration.
Thu, Oct 29, 10:05 AM · Restricted Project

Tue, Oct 27

aartbik added a comment to D90255: [mlir] Add a generic while/do-while loop to the SCF dialect.

Thanks for adding the while!

Tue, Oct 27, 12:31 PM · Restricted Project

Mon, Oct 26

aartbik accepted D90164: [mlir] NFC: properly align IR in comments.
Mon, Oct 26, 10:13 AM · Restricted Project

Oct 23 2020

aartbik accepted D89744: [mlir][Vector] Introduce UnrollVectorOptions to control vector unrolling..
Oct 23 2020, 11:27 AM · Restricted Project
aartbik accepted D89850: [mlir][vector] Add folder for ExtractStridedSliceOp.
Oct 23 2020, 11:20 AM · Restricted Project
aartbik accepted D89853: [mlir][vector] Add folding for ExtractOp with ShapeCastOp source.
Oct 23 2020, 11:18 AM · Restricted Project

Oct 22 2020

aartbik added a comment to D89744: [mlir][Vector] Introduce UnrollVectorOptions to control vector unrolling..

Overall looks good, two more comments. Please find out what causes windows to fail (I am curious too what makes the CHECK different)

Oct 22 2020, 11:16 AM · Restricted Project

Oct 21 2020

aartbik added inline comments to D89726: MLIR: add ceil/floor divide in Standard Dialect.
Oct 21 2020, 11:00 AM · Restricted Project
aartbik added inline comments to D89850: [mlir][vector] Add folder for ExtractStridedSliceOp.
Oct 21 2020, 10:50 AM · Restricted Project
aartbik added inline comments to D89853: [mlir][vector] Add folding for ExtractOp with ShapeCastOp source.
Oct 21 2020, 10:45 AM · Restricted Project
aartbik added a comment to D89744: [mlir][Vector] Introduce UnrollVectorOptions to control vector unrolling..

Did you upload the changes? Revision seems the same?
Also, please see the failure in the report.

Oct 21 2020, 10:41 AM · Restricted Project

Oct 20 2020

aartbik requested changes to D89744: [mlir][Vector] Introduce UnrollVectorOptions to control vector unrolling..
Oct 20 2020, 9:57 AM · Restricted Project

Oct 16 2020

aartbik added inline comments to D89563: [mlir][vector] Update doc strings for insert_map/extract_map and fix insert_map semantic.
Oct 16 2020, 2:25 PM · Restricted Project

Oct 13 2020

aartbik accepted D89261: [mlir] Linalg refactor for using "bufferize" terminology..

bufferize it is!

Oct 13 2020, 4:18 PM · Restricted Project
aartbik added inline comments to D89291: [mlir][vector] Improve vector distribute integration test and fix block distribution.
Oct 13 2020, 3:16 PM · Restricted Project

Oct 12 2020

aartbik accepted D89265: [mlir][Linalg] Temporarily circumvent TensorConstant bufferize bug.
Oct 12 2020, 1:23 PM · Restricted Project

Oct 9 2020

aartbik committed rG3c366740ca24: [mlir] [standard] fixed typo in comment (authored by aartbik).
[mlir] [standard] fixed typo in comment
Oct 9 2020, 5:04 PM
aartbik closed D89172: [mlir] [standard] fixed typo in comment.
Oct 9 2020, 5:04 PM · Restricted Project
aartbik added a comment to D89172: [mlir] [standard] fixed typo in comment.

Thank you River!

Oct 9 2020, 4:59 PM · Restricted Project
aartbik added reviewers for D89172: [mlir] [standard] fixed typo in comment: mehdi_amini, grosul1.
Oct 9 2020, 4:57 PM · Restricted Project
aartbik requested review of D89172: [mlir] [standard] fixed typo in comment.
Oct 9 2020, 4:56 PM · Restricted Project

Oct 8 2020

aartbik added a comment to D89062: [mlir][vector] Add integration test for vector distribute transformation.

and thank you for contributing to the integration tests!

Oct 8 2020, 2:17 PM · Restricted Project
aartbik accepted D89061: [mlir][vector] Add unit test for vector distribute by block.
Oct 8 2020, 2:17 PM · Restricted Project
aartbik added inline comments to D89061: [mlir][vector] Add unit test for vector distribute by block.
Oct 8 2020, 2:00 PM · Restricted Project
aartbik accepted D89062: [mlir][vector] Add integration test for vector distribute transformation.
Oct 8 2020, 1:59 PM · Restricted Project
aartbik added inline comments to D89061: [mlir][vector] Add unit test for vector distribute by block.
Oct 8 2020, 12:38 PM · Restricted Project
aartbik added inline comments to D89062: [mlir][vector] Add integration test for vector distribute transformation.
Oct 8 2020, 12:36 PM · Restricted Project