Page MenuHomePhabricator

Please use GitHub pull requests for new patches. Phabricator shutdown timeline

rsuderman (Rob Suderman)
User

Projects

User does not belong to any projects.

User Details

User Since
Jan 2 2020, 10:50 AM (195 w, 4 d)

Recent Activity

Mon, Sep 18

rsuderman committed rGcbd475040f89: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass (authored by rsuderman).
[mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass
Mon, Sep 18, 5:11 PM · Restricted Project, Restricted Project
rsuderman closed D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass.
Mon, Sep 18, 5:11 PM · Restricted Project, Restricted Project

Tue, Sep 12

rsuderman added inline comments to D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass.
Tue, Sep 12, 11:04 AM · Restricted Project, Restricted Project
rsuderman updated the diff for D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass.

Updated for jpienaar@ comments

Tue, Sep 12, 11:02 AM · Restricted Project, Restricted Project

Thu, Sep 7

rsuderman committed rGcf29d0a73764: [mlir][tosa]Create a check for i64 input in apply_scale lowering in TosaToArith (authored by NatashaKnk).
[mlir][tosa]Create a check for i64 input in apply_scale lowering in TosaToArith
Thu, Sep 7, 11:10 AM · Restricted Project, Restricted Project
rsuderman closed D159473: [mlir][tosa]Create a check for i64 input in apply_scale lowering in TosaToArith.
Thu, Sep 7, 11:10 AM · Restricted Project, Restricted Project
rsuderman accepted D159473: [mlir][tosa]Create a check for i64 input in apply_scale lowering in TosaToArith.
Thu, Sep 7, 10:05 AM · Restricted Project, Restricted Project

Aug 31 2023

rsuderman requested changes to D156362: [mlir][Arith] Change F32 to BF16 truncation to match __truncsfbf2.
Aug 31 2023, 10:09 AM · Restricted Project, Restricted Project
rsuderman added inline comments to D156362: [mlir][Arith] Change F32 to BF16 truncation to match __truncsfbf2.
Aug 31 2023, 10:09 AM · Restricted Project, Restricted Project
rsuderman added a comment to D156362: [mlir][Arith] Change F32 to BF16 truncation to match __truncsfbf2.

@rsuderman do you have some info on the original implementation (it was yours right?) and the change in numerics here?

Aug 31 2023, 10:08 AM · Restricted Project, Restricted Project
rsuderman accepted D154832: [mlir][tosa] Constant optimizations for reduce operations.
Aug 31 2023, 10:02 AM · Restricted Project, Restricted Project
rsuderman abandoned D155645: [mlir][vector] Materialize out non-matching vector.contract operands.
Aug 31 2023, 10:01 AM · Restricted Project, Restricted Project
rsuderman updated the diff for D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass.

Rebased.

Aug 31 2023, 9:54 AM · Restricted Project, Restricted Project

Aug 30 2023

rsuderman added a reviewer for D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass: jpienaar.
Aug 30 2023, 2:24 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass: NatashaKnk.
Aug 30 2023, 2:22 PM · Restricted Project, Restricted Project
rsuderman requested review of D159228: [mlir][mlprogram] Add `mlprogram-pipeline-globals` optimization pass.
Aug 30 2023, 2:22 PM · Restricted Project, Restricted Project

Aug 25 2023

rsuderman committed rGf66e4bd67adf: [mlir][math] Modify math.powf to handle negative bases. (authored by bviyer).
[mlir][math] Modify math.powf to handle negative bases.
Aug 25 2023, 3:36 PM · Restricted Project, Restricted Project
rsuderman closed D158797: [mlir][math] Modify math.powf to handle negative bases..
Aug 25 2023, 3:35 PM · Restricted Project, Restricted Project
rsuderman accepted D158797: [mlir][math] Modify math.powf to handle negative bases..
Aug 25 2023, 12:09 PM · Restricted Project, Restricted Project

Aug 17 2023

rsuderman added a comment to D154832: [mlir][tosa] Constant optimizations for reduce operations.

Overall I would move these next to the other constant folders. Looking standing back they are all the same as the other layerwise folders and should be kept together. There is not much value is separating the reduction folders into a separate file. Once thats done I can approve and land.

Aug 17 2023, 5:11 PM · Restricted Project, Restricted Project
rsuderman abandoned D158224: [mlir] Unneeded exports were added to the float utils.
Aug 17 2023, 4:50 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D158224: [mlir] Unneeded exports were added to the float utils: NatashaKnk.
Aug 17 2023, 3:13 PM · Restricted Project, Restricted Project
rsuderman requested review of D158224: [mlir] Unneeded exports were added to the float utils.
Aug 17 2023, 3:12 PM · Restricted Project, Restricted Project
rsuderman accepted D156369: [mlir] Add trait SameOperandsAndResultRank.
Aug 17 2023, 2:44 PM · Restricted Project, Restricted Project

Aug 16 2023

rsuderman accepted D156361: [mlir][AMDGPU] Improve BF16 handling through AMDGPU compilation.
Aug 16 2023, 2:48 PM · Restricted Project, Restricted Project
rsuderman accepted D156362: [mlir][Arith] Change F32 to BF16 truncation to match __truncsfbf2.

I agree that this is the right decomposition. The original implementation matched this behavior as well, however it was reverted due to other parties requesting truncation of fp should include rounding. Approving so you can land, just be aware this may bring wider discussions.

Aug 16 2023, 2:48 PM · Restricted Project, Restricted Project
rsuderman added inline comments to D157439: [mlir][tosa]Fix Rescale shift attr data type.
Aug 16 2023, 2:45 PM · Restricted Project, Restricted Project

Aug 15 2023

rsuderman accepted D157424: [mlir][tosa] Change axis to I32Attr.
Aug 15 2023, 11:32 AM · Restricted Project, Restricted Project

Jul 31 2023

rsuderman added inline comments to D156369: [mlir] Add trait SameOperandsAndResultRank.
Jul 31 2023, 11:23 AM · Restricted Project, Restricted Project
rsuderman requested changes to D156369: [mlir] Add trait SameOperandsAndResultRank.
Jul 31 2023, 10:35 AM · Restricted Project, Restricted Project

Jul 20 2023

rsuderman added a comment to D154832: [mlir][tosa] Constant optimizations for reduce operations.

Sorry for the delay

Jul 20 2023, 10:34 AM · Restricted Project, Restricted Project

Jul 19 2023

rsuderman added a comment to D155645: [mlir][vector] Materialize out non-matching vector.contract operands.

Thanks, Rob! I would like to know what @nicolasvasilache thinks, in case I'm missing something, but I think both linalg.matmul and vector.contract shouldn't allow inputs with different types at all. That should be something to check in the op verification and the proper conversion should happen up in the pipeline, when we generate the linalg.matmul op in first place. Does that make sense to you?

Jul 19 2023, 12:59 PM · Restricted Project, Restricted Project
rsuderman retitled D155645: [mlir][vector] Materialize out non-matching vector.contract operands from [mlir][vector] Drop matching element type requirement on `vector.contract` to [mlir][vector] Materialize out non-matching vector.contract operands.
Jul 19 2023, 12:37 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D155645: [mlir][vector] Materialize out non-matching vector.contract operands.

Updated to remove requirement change and instead update vectorization code

Jul 19 2023, 12:37 PM · Restricted Project, Restricted Project
rsuderman added a comment to D155645: [mlir][vector] Materialize out non-matching vector.contract operands.

Hey Rob! Thanks for sending this patch.

E.g. i8 and i16 may accumulate into an i32 legally, despite not matching in bitwidth.

I think if that is the case, we should extend i8 to i16 first and then use an i16, i16 -> i32 contract. We have to keep the number of variants under control at this level of abstraction. Otherwise, the transformations and contract lowerings would have to deal with too many cases.
Would that work for your case?

Thanks,
Diego

So there is actually a slight problem with that. I used the i8 and i16 case for a simple example. The actual case I am encountering is f16 and bf16 which have lossy casting both ways. Happy to look into an alternate solution but none are coming to my mind.

Ha, that's interesting. I guess it's up to the framework to define what happens when we operate an f16 and bf16 (what is the direction of the conversion). It sounds like we want to disambiguate that as early as possible instead of rolling down the ball. That call would have to be made by specific spec in charge.

Jul 19 2023, 10:57 AM · Restricted Project, Restricted Project

Jul 18 2023

rsuderman added a comment to D155645: [mlir][vector] Materialize out non-matching vector.contract operands.

Hey Rob! Thanks for sending this patch.

E.g. i8 and i16 may accumulate into an i32 legally, despite not matching in bitwidth.

I think if that is the case, we should extend i8 to i16 first and then use an i16, i16 -> i32 contract. We have to keep the number of variants under control at this level of abstraction. Otherwise, the transformations and contract lowerings would have to deal with too many cases.
Would that work for your case?

Thanks,
Diego

Jul 18 2023, 4:20 PM · Restricted Project, Restricted Project
rsuderman requested review of D155645: [mlir][vector] Materialize out non-matching vector.contract operands.
Jul 18 2023, 2:28 PM · Restricted Project, Restricted Project

Jul 17 2023

rsuderman committed rGa5cee3e386bd: [mlir][linalg] Add a padding case for `ComplexType` (authored by Robert Suderman <suderman@google.com>).
[mlir][linalg] Add a padding case for `ComplexType`
Jul 17 2023, 5:23 PM · Restricted Project, Restricted Project
rsuderman closed D154908: [mlir][linalg] Add a padding case for `ComplexType`.
Jul 17 2023, 5:22 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D154908: [mlir][linalg] Add a padding case for `ComplexType`: kuhar.
Jul 17 2023, 4:00 PM · Restricted Project, Restricted Project
rsuderman committed rGf513b70d43c3: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype` (authored by rsuderman).
[mlir] Add `ComplexType` conversion support for `convertScalarToDtype`
Jul 17 2023, 2:06 PM · Restricted Project, Restricted Project
rsuderman closed D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.
Jul 17 2023, 2:05 PM · Restricted Project, Restricted Project

Jul 14 2023

rsuderman added a comment to D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.

Do you plan to add some tests, or is this covered by some existing one?

Happy to add some. I am less familiar with this area in the codebase so thoughts on how / where to test is appreciate.

I checked and the code from this patch is not exercised by any lit test. The helper is used with the other types in these:

Dialect/Linalg/IR/LinalgOps.cpp
Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
Dialect/Linalg/Transforms/ConvertConv2DToImg2Col.cpp
Dialect/SparseTensor/Transforms/CodegenUtils.cpp

So I'd think we either add some linalg/sparse tensor tests, or a new test pass.

Jul 14 2023, 11:52 AM · Restricted Project, Restricted Project
rsuderman updated the diff for D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.

Added testing and added complex type support for ConvertConv2DtoImg2Col

Jul 14 2023, 11:50 AM · Restricted Project, Restricted Project
rsuderman added inline comments to D154908: [mlir][linalg] Add a padding case for `ComplexType`.
Jul 14 2023, 11:02 AM · Restricted Project, Restricted Project
rsuderman updated the diff for D154908: [mlir][linalg] Add a padding case for `ComplexType`.

git-clang-format

Jul 14 2023, 10:07 AM · Restricted Project, Restricted Project
rsuderman updated the diff for D154908: [mlir][linalg] Add a padding case for `ComplexType`.

Updated to check ComplexType directly.

Jul 14 2023, 10:04 AM · Restricted Project, Restricted Project

Jul 12 2023

rsuderman added a comment to D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.

Do you plan to add some tests, or is this covered by some existing one?

Jul 12 2023, 3:38 PM · Restricted Project, Restricted Project

Jul 11 2023

rsuderman added inline comments to D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.
Jul 11 2023, 2:59 PM · Restricted Project, Restricted Project

Jul 10 2023

rsuderman added a reviewer for D154908: [mlir][linalg] Add a padding case for `ComplexType`: mravishankar.
Jul 10 2023, 5:09 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D154908: [mlir][linalg] Add a padding case for `ComplexType`.

Rebased on head

Jul 10 2023, 5:04 PM · Restricted Project, Restricted Project
rsuderman requested review of D154908: [mlir][linalg] Add a padding case for `ComplexType`.
Jul 10 2023, 4:56 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`: jpienaar.
Jul 10 2023, 3:25 PM · Restricted Project, Restricted Project
rsuderman accepted D154539: [mlir][Arith] Add pass for emulating unsupported float ops (#1079).
Jul 10 2023, 3:04 PM · Restricted Project, Restricted Project
rsuderman requested changes to D154832: [mlir][tosa] Constant optimizations for reduce operations.
Jul 10 2023, 10:39 AM · Restricted Project, Restricted Project

Jul 7 2023

rsuderman added inline comments to D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.
Jul 7 2023, 3:57 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.

Add operand bailout

Jul 7 2023, 3:56 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.

Factored into helper functions and performed general var name cleanup.

Jul 7 2023, 3:14 PM · Restricted Project, Restricted Project
rsuderman added inline comments to D154539: [mlir][Arith] Add pass for emulating unsupported float ops (#1079).
Jul 7 2023, 2:37 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`: mravishankar.
Jul 7 2023, 12:43 PM · Restricted Project, Restricted Project
rsuderman requested review of D154740: [mlir] Add `ComplexType` conversion support for `convertScalarToDtype`.
Jul 7 2023, 12:43 PM · Restricted Project, Restricted Project

Jul 6 2023

rsuderman committed rGa8df21f433d2: [mlir][complex] Add a `complex.bitcast` operation (authored by rsuderman).
[mlir][complex] Add a `complex.bitcast` operation
Jul 6 2023, 4:23 PM · Restricted Project, Restricted Project
rsuderman closed D154663: [mlir][complex] Add a `complex.bitcast` operation.
Jul 6 2023, 4:23 PM · Restricted Project, Restricted Project
rsuderman added inline comments to D154663: [mlir][complex] Add a `complex.bitcast` operation.
Jul 6 2023, 4:19 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D154663: [mlir][complex] Add a `complex.bitcast` operation.

Updated for jpienaar@ comments. Included adding complex.bitcast -> arith.bitcast canonicalizer.

Jul 6 2023, 4:19 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D154663: [mlir][complex] Add a `complex.bitcast` operation.

Rebased on main

Jul 6 2023, 3:38 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D154663: [mlir][complex] Add a `complex.bitcast` operation: jpienaar.
Jul 6 2023, 3:33 PM · Restricted Project, Restricted Project
rsuderman requested review of D154663: [mlir][complex] Add a `complex.bitcast` operation.
Jul 6 2023, 3:32 PM · Restricted Project, Restricted Project

Jun 23 2023

rsuderman committed rG0bedb667af1d: [mlir][math] Improved math.atan approximation (authored by Robert Suderman <suderman@google.com>).
[mlir][math] Improved math.atan approximation
Jun 23 2023, 5:33 PM · Restricted Project, Restricted Project
rsuderman closed D153656: [mlir][math] Improved math.atan approximation.
Jun 23 2023, 5:32 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D153656: [mlir][math] Improved math.atan approximation.

Formatting changes and float -> double correction.

Jun 23 2023, 4:30 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D153656: [mlir][math] Improved math.atan approximation.

Expand f32Cst to take double

Jun 23 2023, 2:02 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D153656: [mlir][math] Improved math.atan approximation.

More windows compilation failures.

Jun 23 2023, 1:46 PM · Restricted Project, Restricted Project
rsuderman updated the diff for D153656: [mlir][math] Improved math.atan approximation.

Fixed windows build

Jun 23 2023, 12:42 PM · Restricted Project, Restricted Project
rsuderman committed rG710dc7282a01: [mlir][math] Modified the 'math.exp' lowering for higher precision (authored by Robert Suderman <suderman@google.com>).
[mlir][math] Modified the 'math.exp' lowering for higher precision
Jun 23 2023, 12:28 PM · Restricted Project, Restricted Project
rsuderman closed D153592: [mlir][math] Modified the 'math.exp' lowering for higher precision.
Jun 23 2023, 12:27 PM · Restricted Project, Restricted Project
rsuderman added a reviewer for D153656: [mlir][math] Improved math.atan approximation: NatashaKnk.
Jun 23 2023, 11:46 AM · Restricted Project, Restricted Project
rsuderman requested review of D153656: [mlir][math] Improved math.atan approximation.
Jun 23 2023, 11:44 AM · Restricted Project, Restricted Project

Jun 22 2023

rsuderman added a reviewer for D153592: [mlir][math] Modified the 'math.exp' lowering for higher precision: cota.
Jun 22 2023, 3:42 PM · Restricted Project, Restricted Project
rsuderman requested review of D153592: [mlir][math] Modified the 'math.exp' lowering for higher precision.
Jun 22 2023, 3:39 PM · Restricted Project, Restricted Project

Jun 15 2023

rsuderman committed rG86c4972f5f67: [mlir][tosa] Improve lowering support for tosa.concat (authored by sabauma).
[mlir][tosa] Improve lowering support for tosa.concat
Jun 15 2023, 11:41 AM · Restricted Project, Restricted Project
rsuderman closed D151707: [mlir][tosa] Improve lowering support for tosa.concat.
Jun 15 2023, 11:41 AM · Restricted Project, Restricted Project
rsuderman accepted D151707: [mlir][tosa] Improve lowering support for tosa.concat.
Jun 15 2023, 11:22 AM · Restricted Project, Restricted Project
rsuderman requested changes to D150578: [mlir][tosa] Constant folding for reciprocal.

One change before landing. Everything is broken up between 3-4 files of <100 lines in length. It doesn't seem worth fracturing this up to this degree. Could you have TosaFolders.cpp and merge them all together, including the transpose pass)

Jun 15 2023, 11:21 AM · Restricted Project, Restricted Project
rsuderman added a comment to D150578: [mlir][tosa] Constant folding for reciprocal.

I can land now

Jun 15 2023, 11:17 AM · Restricted Project, Restricted Project
rsuderman committed rG2a00891107bc: [mlir][linalg] Fix linalg.conv vectorization for mixed int-fp types (authored by rsuderman).
[mlir][linalg] Fix linalg.conv vectorization for mixed int-fp types
Jun 15 2023, 11:16 AM · Restricted Project, Restricted Project
rsuderman closed D152982: [mlir][linalg] Fix linalg.conv vectorization for mixed int-fp types.
Jun 15 2023, 11:15 AM · Restricted Project, Restricted Project
rsuderman updated the diff for D152982: [mlir][linalg] Fix linalg.conv vectorization for mixed int-fp types.

Add matching element type

Jun 15 2023, 10:52 AM · Restricted Project, Restricted Project

Jun 14 2023

rsuderman requested review of D152982: [mlir][linalg] Fix linalg.conv vectorization for mixed int-fp types.
Jun 14 2023, 4:59 PM · Restricted Project, Restricted Project

Jun 12 2023

rsuderman added inline comments to D151707: [mlir][tosa] Improve lowering support for tosa.concat.
Jun 12 2023, 10:15 AM · Restricted Project, Restricted Project
rsuderman added inline comments to D151707: [mlir][tosa] Improve lowering support for tosa.concat.
Jun 12 2023, 10:00 AM · Restricted Project, Restricted Project

Jun 2 2023

rsuderman requested changes to D151707: [mlir][tosa] Improve lowering support for tosa.concat.
Jun 2 2023, 12:17 PM · Restricted Project, Restricted Project

May 2 2023

rsuderman committed rG486a2ca57a27: Add tensor.bitcast op to Tensor dialect (authored by trevor-m).
Add tensor.bitcast op to Tensor dialect
May 2 2023, 1:28 PM · Restricted Project, Restricted Project
rsuderman closed D149608: Add tensor.bitcast op to Tensor dialect.
May 2 2023, 1:28 PM · Restricted Project, Restricted Project
rsuderman accepted D149608: Add tensor.bitcast op to Tensor dialect.
May 2 2023, 11:20 AM · Restricted Project, Restricted Project

May 1 2023

rsuderman added a reviewer for D149608: Add tensor.bitcast op to Tensor dialect: ftynse.
May 1 2023, 2:50 PM · Restricted Project, Restricted Project
rsuderman added inline comments to D149608: Add tensor.bitcast op to Tensor dialect.
May 1 2023, 2:50 PM · Restricted Project, Restricted Project

Apr 24 2023

rsuderman added reviewers for D149109: [mlir][tosa] Fix `tosa.reshape` folder for quantized constants: NatashaKnk, jpienaar.
Apr 24 2023, 4:44 PM · Restricted Project, Restricted Project
rsuderman requested review of D149109: [mlir][tosa] Fix `tosa.reshape` folder for quantized constants.
Apr 24 2023, 4:43 PM · Restricted Project, Restricted Project

Apr 20 2023

rsuderman added a reviewer for D148820: [mlir] Expose Arith::ExpandOps options in constructor: NatashaKnk.
Apr 20 2023, 11:05 AM · Restricted Project, Restricted Project