Page MenuHomePhabricator

aartbik (Aart Bik)
PhD

Projects

User does not belong to any projects.

User Details

User Since
Jan 10 2020, 10:13 AM (6 w, 6 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 created D75320: [mlir] [VectorOps] Add vector.broadcast to EDSC.
Thu, Feb 27, 4:56 PM · Restricted Project
aartbik added a reviewer for D75320: [mlir] [VectorOps] Add vector.broadcast to EDSC: andydavis1.
Thu, Feb 27, 4:56 PM · Restricted Project

Yesterday

aartbik accepted D75217: [mlir] NFC - Move runner utils from mlir-cpu-runner to ExecutionEngine.
Wed, Feb 26, 3:42 PM · Restricted Project

Tue, Feb 25

aartbik committed rG3cefebc3fe1e: [mlir] [VectorOps] Add vector.print to EDSC (authored by aartbik).
[mlir] [VectorOps] Add vector.print to EDSC
Tue, Feb 25, 3:36 PM
aartbik closed D75147: [mlir] [VectorOps] Add vector.print to EDSC.
Tue, Feb 25, 3:35 PM · Restricted Project
aartbik retitled D75147: [mlir] [VectorOps] Add vector.print to EDSC from [VectorOps] Add vector.print to EDSC to [mlir] [VectorOps] Add vector.print to EDSC.
Tue, Feb 25, 3:08 PM · Restricted Project
aartbik created D75147: [mlir] [VectorOps] Add vector.print to EDSC.
Tue, Feb 25, 3:05 PM · Restricted Project
aartbik added a reviewer for D75147: [mlir] [VectorOps] Add vector.print to EDSC: andydavis1.
Tue, Feb 25, 3:05 PM · Restricted Project

Mon, Feb 24

aartbik accepted D74889: [mlir] Generalize intrinsic builders in the LLVM dialect definition.
Mon, Feb 24, 9:44 AM · Restricted Project

Thu, Feb 20

aartbik committed rGee260c1a0f1c: [mlir] [VectorOps] Multi-dim reductions for lowering vector.contract (authored by aartbik).
[mlir] [VectorOps] Multi-dim reductions for lowering vector.contract
Thu, Feb 20, 2:21 PM
aartbik closed D74880: [mlir] [VectorOps] Multi-dim reductions for lowering vector.contract.
Thu, Feb 20, 2:21 PM · Restricted Project
aartbik updated the diff for D74880: [mlir] [VectorOps] Multi-dim reductions for lowering vector.contract.

rebase

Thu, Feb 20, 2:02 PM · Restricted Project
aartbik added a comment to D74880: [mlir] [VectorOps] Multi-dim reductions for lowering vector.contract.

The order is always the same of a given mapping and op, but the accumulation may indeed introduce instabilities if an exact order is assumed (i.e. acc feeds into the chain rather than being just added at the end). We can make semantics more strict if needed.

Thu, Feb 20, 2:02 PM · Restricted Project
aartbik added a comment to D74889: [mlir] Generalize intrinsic builders in the LLVM dialect definition.

I like where this is going a lot. Some suggestions on the doc though

Thu, Feb 20, 9:15 AM · Restricted Project

Wed, Feb 19

aartbik added a reviewer for D74880: [mlir] [VectorOps] Multi-dim reductions for lowering vector.contract: andydavis1.
Wed, Feb 19, 5:19 PM · Restricted Project
aartbik created D74880: [mlir] [VectorOps] Multi-dim reductions for lowering vector.contract.
Wed, Feb 19, 4:23 PM · Restricted Project
aartbik committed rG0ba9ee9f0e55: [mlir] [VectorOps] Framework for progressive lowering of vector.contract (authored by aartbik).
[mlir] [VectorOps] Framework for progressive lowering of vector.contract
Wed, Feb 19, 11:37 AM
aartbik closed D74797: [mlir] [VectorOps] Framework for progressive lowering of vector.contract.
Wed, Feb 19, 11:37 AM · Restricted Project
aartbik updated the diff for D74797: [mlir] [VectorOps] Framework for progressive lowering of vector.contract.

added TODOs as requested

Wed, Feb 19, 11:36 AM · Restricted Project

Tue, Feb 18

aartbik updated the diff for D74797: [mlir] [VectorOps] Framework for progressive lowering of vector.contract.

addressed comments

Tue, Feb 18, 4:50 PM · Restricted Project
aartbik updated the summary of D74797: [mlir] [VectorOps] Framework for progressive lowering of vector.contract.
Tue, Feb 18, 2:46 PM · Restricted Project
aartbik created D74797: [mlir] [VectorOps] Framework for progressive lowering of vector.contract.
Tue, Feb 18, 2:44 PM · Restricted Project

Thu, Feb 13

aartbik committed rGb21c7999520a: [mlir] [VectorOps] Initial framework for progressively lowering vector.contract (authored by aartbik).
[mlir] [VectorOps] Initial framework for progressively lowering vector.contract
Thu, Feb 13, 3:10 PM
aartbik closed D74520: [mlir] [VectorOps] Initial framework for progressively lowering vector.contract.
Thu, Feb 13, 3:09 PM · Restricted Project
aartbik updated the diff for D74520: [mlir] [VectorOps] Initial framework for progressively lowering vector.contract.

addressed comments, rebased

Thu, Feb 13, 3:00 PM · Restricted Project

Wed, Feb 12

aartbik created D74520: [mlir] [VectorOps] Initial framework for progressively lowering vector.contract.
Wed, Feb 12, 7:13 PM · Restricted Project
aartbik accepted D74457: [mlir][EDSC][Linalg] Compose linalg_matmul and vector.contract.

trusting that you will fix both typos, good to land

Wed, Feb 12, 9:56 AM · Restricted Project

Tue, Feb 11

aartbik added inline comments to D74457: [mlir][EDSC][Linalg] Compose linalg_matmul and vector.contract.
Tue, Feb 11, 10:00 PM · Restricted Project
aartbik committed rGe83b7b99da2e: [mlir] [VectorOps] Implement vector.reduce operation (authored by aartbik).
[mlir] [VectorOps] Implement vector.reduce operation
Tue, Feb 11, 11:37 AM
aartbik closed D74370: [mlir] [VectorOps] Implement vector.reduce operation.
Tue, Feb 11, 11:37 AM · Restricted Project
aartbik updated the diff for D74370: [mlir] [VectorOps] Implement vector.reduce operation.

rebased

Tue, Feb 11, 11:19 AM · Restricted Project
aartbik added inline comments to D74370: [mlir] [VectorOps] Implement vector.reduce operation.
Tue, Feb 11, 11:01 AM · Restricted Project

Mon, Feb 10

aartbik added inline comments to D74370: [mlir] [VectorOps] Implement vector.reduce operation.
Mon, Feb 10, 8:25 PM · Restricted Project
aartbik updated the diff for D74370: [mlir] [VectorOps] Implement vector.reduce operation.

updated description, one more verify test

Mon, Feb 10, 8:25 PM · Restricted Project
aartbik added reviewers for D74370: [mlir] [VectorOps] Implement vector.reduce operation: andydavis1, ftynse.
Mon, Feb 10, 5:06 PM · Restricted Project
aartbik created D74370: [mlir] [VectorOps] Implement vector.reduce operation.
Mon, Feb 10, 5:06 PM · Restricted Project
aartbik committed rG83003db430b9: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect (authored by aartbik).
[mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect
Mon, Feb 10, 9:24 AM
aartbik closed D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect.
Mon, Feb 10, 9:23 AM · Restricted Project

Fri, Feb 7

aartbik updated the diff for D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect.

less code is better

Fri, Feb 7, 2:34 PM · Restricted Project
aartbik added a comment to D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect.

Okay, friends of concise code, PTAL

Fri, Feb 7, 2:34 PM · Restricted Project
aartbik accepted D74219: [mlir][VectorOps] Expose and use llvm.intrin.fma*.
Fri, Feb 7, 12:17 PM · Restricted Project
aartbik accepted D74075: [mlir][VectorOps] Introduce a `vector.fma` op that works on n-D vectors and lowers to `llvm.intrin.fmuladd`.
Fri, Feb 7, 11:39 AM · Restricted Project
aartbik added inline comments to D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect.
Fri, Feb 7, 10:35 AM · Restricted Project
aartbik added inline comments to D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect.
Fri, Feb 7, 10:26 AM · Restricted Project
aartbik committed rGe52414b1ae46: [mlir][VectorOps] Generalized vector.print to i32/i64 (authored by aartbik).
[mlir][VectorOps] Generalized vector.print to i32/i64
Fri, Feb 7, 9:31 AM
aartbik closed D74179: [mlir][VectorOps] Generalized vector.print to i32/i64.
Fri, Feb 7, 9:31 AM · Restricted Project
aartbik added inline comments to D74179: [mlir][VectorOps] Generalized vector.print to i32/i64.
Fri, Feb 7, 9:22 AM · Restricted Project
aartbik updated the diff for D74179: [mlir][VectorOps] Generalized vector.print to i32/i64.

addressed comments

Fri, Feb 7, 9:13 AM · Restricted Project

Thu, Feb 6

aartbik added a comment to D74075: [mlir][VectorOps] Introduce a `vector.fma` op that works on n-D vectors and lowers to `llvm.intrin.fmuladd`.

Is there a way to remove your blocking review but without approving yet? I don't see that option here.
Anyway, please look at the other comments that are still open.

Thu, Feb 6, 7:58 PM · Restricted Project
aartbik added reviewers for D74179: [mlir][VectorOps] Generalized vector.print to i32/i64: andydavis1, nicolasvasilache, ftynse.
Thu, Feb 6, 4:16 PM · Restricted Project
aartbik created D74179: [mlir][VectorOps] Generalized vector.print to i32/i64.
Thu, Feb 6, 4:14 PM · Restricted Project
aartbik added reviewers for D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect: ftynse, andydavis1, nicolasvasilache.
Thu, Feb 6, 2:25 PM · Restricted Project
aartbik created D74171: [mlir] [LLVMIR] add all vector reduction intrinsics to LLVM IR dialect.
Thu, Feb 6, 2:25 PM · Restricted Project
aartbik added inline comments to D74075: [mlir][VectorOps] Introduce a `vector.fma` op that works on n-D vectors and lowers to `llvm.intrin.fmuladd`.
Thu, Feb 6, 10:49 AM · Restricted Project
aartbik requested changes to D74075: [mlir][VectorOps] Introduce a `vector.fma` op that works on n-D vectors and lowers to `llvm.intrin.fmuladd`.
Thu, Feb 6, 10:20 AM · Restricted Project

Wed, Feb 5

aartbik committed rG6e2309d7fae2: [mlir] [VectorOps] generalized vector.contract semantics (authored by aartbik).
[mlir] [VectorOps] generalized vector.contract semantics
Wed, Feb 5, 5:23 PM
aartbik closed D74014: [mlir] [VectorOps] generalized vector.contract semantics.
Wed, Feb 5, 5:23 PM · Restricted Project
aartbik updated the diff for D74014: [mlir] [VectorOps] generalized vector.contract semantics.

rebased with head

Wed, Feb 5, 4:56 PM · Restricted Project
aartbik requested changes to D74075: [mlir][VectorOps] Introduce a `vector.fma` op that works on n-D vectors and lowers to `llvm.intrin.fmuladd`.
Wed, Feb 5, 2:45 PM · Restricted Project
aartbik added inline comments to D74056: [mlir][AVX512] Start an AVX512 dialect.
Wed, Feb 5, 1:30 PM · Restricted Project

Tue, Feb 4

aartbik added inline comments to D74014: [mlir] [VectorOps] generalized vector.contract semantics.
Tue, Feb 4, 9:14 PM · Restricted Project
aartbik updated the diff for D74014: [mlir] [VectorOps] generalized vector.contract semantics.

address comments

Tue, Feb 4, 9:14 PM · Restricted Project
aartbik updated the summary of D74014: [mlir] [VectorOps] generalized vector.contract semantics.
Tue, Feb 4, 6:21 PM · Restricted Project
aartbik added reviewers for D74014: [mlir] [VectorOps] generalized vector.contract semantics: andydavis1, rriddle.
Tue, Feb 4, 6:12 PM · Restricted Project
aartbik created D74014: [mlir] [VectorOps] generalized vector.contract semantics.
Tue, Feb 4, 6:12 PM · Restricted Project
aartbik committed rG3c7e9c34b38f: [mlir] [VectorOps] refined description of vector.contract (authored by aartbik).
[mlir] [VectorOps] refined description of vector.contract
Tue, Feb 4, 11:26 AM
aartbik closed D73923: [mlir] [VectorOps] refined description of vector.contract.
Tue, Feb 4, 11:26 AM · Restricted Project

Mon, Feb 3

aartbik added inline comments to rG34cd354ea967: [mlir][Linalg][doc] Add Design Document for the Linalg Dialect.
Mon, Feb 3, 9:43 PM
aartbik added a reviewer for D73923: [mlir] [VectorOps] refined description of vector.contract: andydavis1.
Mon, Feb 3, 1:33 PM · Restricted Project
aartbik created D73923: [mlir] [VectorOps] refined description of vector.contract.
Mon, Feb 3, 1:33 PM · Restricted Project

Fri, Jan 31

aartbik accepted D72965: [mlir][VectorOps] Support vector transfer_read/write unrolling for memrefs with vector element type..
Fri, Jan 31, 3:30 PM · Restricted Project
aartbik committed rGc8fc76a99b30: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering (authored by aartbik).
[mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering
Fri, Jan 31, 11:31 AM
aartbik closed D73733: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering.
Fri, Jan 31, 11:30 AM · Restricted Project
aartbik added inline comments to D73733: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering.
Fri, Jan 31, 11:03 AM · Restricted Project
aartbik updated the diff for D73733: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering.

args in patterns

Fri, Jan 31, 11:03 AM · Restricted Project

Thu, Jan 30

aartbik updated the diff for D73733: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering.

sync back with latest pulled master

Thu, Jan 30, 3:16 PM · Restricted Project
aartbik added a comment to D72965: [mlir][VectorOps] Support vector transfer_read/write unrolling for memrefs with vector element type..

note that the title still needs "[mlir] [VectorOps]"

Thu, Jan 30, 12:10 PM · Restricted Project
aartbik added reviewers for D73733: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering: andydavis1, ftynse.
Thu, Jan 30, 11:35 AM · Restricted Project
aartbik created D73733: [mlir] [VectorOps] fixed bug in vector.insert_strided_slice lowering.
Thu, Jan 30, 11:35 AM · Restricted Project

Wed, Jan 29

aartbik added inline comments to D73656: [mlir][Linalg] Add a Linalg DRR test to go from matmul to vectors.
Wed, Jan 29, 5:27 PM · Restricted Project
aartbik committed rG228ea1a46cc8: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file (authored by aartbik).
[mlir] [VectorOps] consolidate all vector utilities to one header/cc file
Wed, Jan 29, 3:47 PM
aartbik closed D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.
Wed, Jan 29, 3:47 PM · Restricted Project
aartbik added inline comments to D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.
Wed, Jan 29, 3:47 PM · Restricted Project
aartbik updated the diff for D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.

Utils -> VectorUtils typo

Wed, Jan 29, 3:29 PM · Restricted Project
aartbik added inline comments to D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.
Wed, Jan 29, 3:20 PM · Restricted Project
aartbik added a comment to D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.

Thanks! I was not sure on the naming, and like your suggestions better indeed. All addressed.

Wed, Jan 29, 3:02 PM · Restricted Project
aartbik updated the diff for D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.

addressed comments

Wed, Jan 29, 2:53 PM · Restricted Project
aartbik updated the diff for D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.

rebased with master

Wed, Jan 29, 11:31 AM · Restricted Project
aartbik added inline comments to D73635: [mlir][VectorOps] Add ShapeCastOp to the vector ops dialect..
Wed, Jan 29, 10:35 AM · Restricted Project

Jan 28 2020

aartbik added a reviewer for D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file: andydavis1.
Jan 28 2020, 4:52 PM · Restricted Project
aartbik created D73593: [mlir] [VectorOps] consolidate all vector utilities to one header/cc file.
Jan 28 2020, 4:52 PM · Restricted Project

Jan 27 2020

aartbik committed rG459cf6e5006a: [mlir] [VectorOps] Lowering of vector.extract/insert_slices to LLVM IR (authored by aartbik).
[mlir] [VectorOps] Lowering of vector.extract/insert_slices to LLVM IR
Jan 27 2020, 10:40 AM
aartbik closed D72808: [mlir] [VectorOps] Lowering of vector.extract/insert_slices to LLVM IR.
Jan 27 2020, 10:39 AM · Restricted Project

Jan 24 2020

aartbik added inline comments to D73391: [mlir] Add primitive transform pattern to rewrite linalg.fill into vector.broadcast form..
Jan 24 2020, 5:12 PM · Restricted Project
aartbik added inline comments to D73295: [mlir] [VectorOps] Rewriting of vector.extract/insert_slices to other vector ops.
Jan 24 2020, 4:54 PM · Restricted Project
aartbik added a comment to D72808: [mlir] [VectorOps] Lowering of vector.extract/insert_slices to LLVM IR.

\o/ yeah, it has been a long road, but PTAL

Jan 24 2020, 4:36 PM · Restricted Project
aartbik updated the diff for D72808: [mlir] [VectorOps] Lowering of vector.extract/insert_slices to LLVM IR.

synced with latest update

Jan 24 2020, 4:36 PM · Restricted Project
aartbik committed rG303fddeeab10: [mlir] [VectorOps] Rewriting of vector.extract/insert_slices to other vector ops (authored by aartbik).
[mlir] [VectorOps] Rewriting of vector.extract/insert_slices to other vector ops
Jan 24 2020, 4:27 PM
aartbik closed D73295: [mlir] [VectorOps] Rewriting of vector.extract/insert_slices to other vector ops.
Jan 24 2020, 4:27 PM · Restricted Project
aartbik added a comment to D73295: [mlir] [VectorOps] Rewriting of vector.extract/insert_slices to other vector ops.

Hmm. I would argue that the tuple discussion makes less sense in the context of this particular rewriting (since it really deals with getting rid of extract_slices/insert_slices using other slice ops). Cleaning up tuple uses is a nice bonus, but not a requirement, since tuples are really a part of the vector dialect to start with, as the leaking example shows. If you want something that blocks any tuple uses, I would argue that we need another pattern population for that, and run that at places where we need to guarantee they are gone (such as the lowering to LLVM, although the "legality" part takes care of it there already).

Jan 24 2020, 2:19 PM · Restricted Project