Page MenuHomePhabricator

ThomasRaoux (Thomas Raoux)
User

Projects

User does not belong to any projects.

User Details

User Since
Jul 29 2019, 2:32 PM (151 w, 5 d)

Recent Activity

Fri, Jun 24

ThomasRaoux committed rGd343cdd50992: [mlir][vector] Fix bug when swapping scf.for and vector warp op (authored by ThomasRaoux).
[mlir][vector] Fix bug when swapping scf.for and vector warp op
Fri, Jun 24, 12:14 PM · Restricted Project, Restricted Project
ThomasRaoux closed D128405: [mlir][vector] Fix bug when swapping scf.for and vector warp op.
Fri, Jun 24, 12:14 PM · Restricted Project, Restricted Project
ThomasRaoux committed rG7eba5cdf9ce4: [mlir][vector] Relax transfer_write vector distribution pattern (authored by ThomasRaoux).
[mlir][vector] Relax transfer_write vector distribution pattern
Fri, Jun 24, 12:04 PM · Restricted Project, Restricted Project
ThomasRaoux closed D128545: [mlir][vector] Relax transfer_write vector distribution pattern.
Fri, Jun 24, 12:03 PM · Restricted Project, Restricted Project
ThomasRaoux requested review of D128545: [mlir][vector] Relax transfer_write vector distribution pattern.
Fri, Jun 24, 11:42 AM · Restricted Project, Restricted Project

Thu, Jun 23

ThomasRaoux requested review of D128481: Add expensive/extensive pass to set Load/Store alignment.
Thu, Jun 23, 5:31 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D128456: [mlir][Transform] Fix applyToOne corner case when no op is matched..
Thu, Jun 23, 12:17 PM · Restricted Project, Restricted Project

Wed, Jun 22

ThomasRaoux requested review of D128405: [mlir][vector] Fix bug when swapping scf.for and vector warp op.
Wed, Jun 22, 8:01 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D128266: [mlir][Linalg] SplitReduction implementation without tensor::ExpandShapeOp.
Wed, Jun 22, 9:21 AM · Restricted Project, Restricted Project

Fri, Jun 17

ThomasRaoux accepted D128079: [mlir][ROCDL] Define MLIR wrappers around new MFMA intrinsics.

LGTM

Fri, Jun 17, 11:37 AM · Restricted Project, Restricted Project
ThomasRaoux accepted D128074: [mlir][VectorToGPU] Fix support for i4, col-major operand support.
Fri, Jun 17, 10:59 AM · Restricted Project, Restricted Project

Thu, Jun 16

ThomasRaoux committed rGf011d32c3a62: [mlir][vector] Fix contraction op lowering with mixed types (authored by ThomasRaoux).
[mlir][vector] Fix contraction op lowering with mixed types
Thu, Jun 16, 9:41 AM · Restricted Project, Restricted Project
ThomasRaoux closed D127926: [mlir][vector] Fix contraction op lowering with mixed types.
Thu, Jun 16, 9:41 AM · Restricted Project, Restricted Project
ThomasRaoux committed rG046ebeb60504: [mlir][linalg] Relax convolution vectorization to support mixed types (authored by ThomasRaoux).
[mlir][linalg] Relax convolution vectorization to support mixed types
Thu, Jun 16, 9:36 AM · Restricted Project, Restricted Project
ThomasRaoux closed D127925: [mlir][linalg] Relax convolution vectorization to support mixed types.
Thu, Jun 16, 9:36 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127925: [mlir][linalg] Relax convolution vectorization to support mixed types.
Thu, Jun 16, 9:29 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127925: [mlir][linalg] Relax convolution vectorization to support mixed types.

clarify comment

Thu, Jun 16, 9:29 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127925: [mlir][linalg] Relax convolution vectorization to support mixed types.
Thu, Jun 16, 7:59 AM · Restricted Project, Restricted Project
ThomasRaoux accepted D127930: [mlir][spirv] Workaround driver bug in math.ctlz conversion again.
Thu, Jun 16, 7:42 AM · Restricted Project, Restricted Project

Wed, Jun 15

ThomasRaoux requested review of D127926: [mlir][vector] Fix contraction op lowering with mixed types.
Wed, Jun 15, 5:56 PM · Restricted Project, Restricted Project
ThomasRaoux requested review of D127925: [mlir][linalg] Relax convolution vectorization to support mixed types.
Wed, Jun 15, 5:54 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D127909: [mlir][spirv] Define spv.ISubBorrowOp.
Wed, Jun 15, 4:04 PM · Restricted Project, Restricted Project
ThomasRaoux committed rGa6f2c2291ede: [mlir][GPUToNVVM] Fix bug in mma elementwise lowering (authored by ThomasRaoux).
[mlir][GPUToNVVM] Fix bug in mma elementwise lowering
Wed, Jun 15, 10:23 AM · Restricted Project, Restricted Project
ThomasRaoux closed D127879: [mlir][GPUToNVVM] Fix bug in mma elementwise lowering.
Wed, Jun 15, 10:23 AM · Restricted Project, Restricted Project
ThomasRaoux requested review of D127879: [mlir][GPUToNVVM] Fix bug in mma elementwise lowering.
Wed, Jun 15, 10:19 AM · Restricted Project, Restricted Project
ThomasRaoux committed rG6834803c3dbc: [mlir][vector] NFC remove dependency of VectorTransform to GPU dialect (authored by ThomasRaoux).
[mlir][vector] NFC remove dependency of VectorTransform to GPU dialect
Wed, Jun 15, 9:09 AM · Restricted Project, Restricted Project
ThomasRaoux closed D127867: [mlir][vector] NFC remove dependency of VectorTransform to GPU dialect.
Wed, Jun 15, 9:09 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127867: [mlir][vector] NFC remove dependency of VectorTransform to GPU dialect.

Address review comments

Wed, Jun 15, 9:05 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.

address review comment

Wed, Jun 15, 9:05 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.
Wed, Jun 15, 8:20 AM · Restricted Project, Restricted Project
ThomasRaoux requested review of D127867: [mlir][vector] NFC remove dependency of VectorTransform to GPU dialect.
Wed, Jun 15, 8:19 AM · Restricted Project, Restricted Project

Tue, Jun 14

ThomasRaoux accepted D127816: [mlir][spirv] Handle corner cases for math.powf conversion.
Tue, Jun 14, 6:27 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D127687: [mlir][linalg] Add named ops for depthwise 3d convolution.
Tue, Jun 14, 9:21 AM · Restricted Project, Restricted Project

Mon, Jun 13

ThomasRaoux committed rG087aba4f0f96: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles (authored by ThomasRaoux).
[mlir][vector] Add pattern to distribute vector reduction to GPU shuffles
Mon, Jun 13, 10:54 PM · Restricted Project, Restricted Project
ThomasRaoux closed D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.
Mon, Jun 13, 10:54 PM · Restricted Project, Restricted Project
ThomasRaoux committed rG76cf33dab2d8: [mlir][vector] Add patterns to ppropagate vector distribution (authored by ThomasRaoux).
[mlir][vector] Add patterns to ppropagate vector distribution
Mon, Jun 13, 10:28 PM · Restricted Project, Restricted Project
ThomasRaoux added a reverting change for rG1c84800c42d2: [mlir][vector] Add patterns to ppropagate vector distribution: rG2d32dac8bb10: Revert "[mlir][vector] Add patterns to ppropagate vector distribution".
Mon, Jun 13, 10:56 AM · Restricted Project, Restricted Project
ThomasRaoux committed rG2d32dac8bb10: Revert "[mlir][vector] Add patterns to ppropagate vector distribution" (authored by ThomasRaoux).
Revert "[mlir][vector] Add patterns to ppropagate vector distribution"
Mon, Jun 13, 10:56 AM · Restricted Project, Restricted Project
ThomasRaoux added a reverting change for D127167: [mlir][vector] Add patterns to propagate vector distribution: rG2d32dac8bb10: Revert "[mlir][vector] Add patterns to ppropagate vector distribution".
Mon, Jun 13, 10:56 AM · Restricted Project, Restricted Project
ThomasRaoux accepted D127582: [mlir][spirv] Convert math.ctlz to spv.GLSL.FindUMsb.
Mon, Jun 13, 9:58 AM · Restricted Project, Restricted Project
ThomasRaoux accepted D127574: [mlir][spirv] Fix result type for arith.cmpi/cmpf conversion.
Mon, Jun 13, 9:52 AM · Restricted Project, Restricted Project
ThomasRaoux accepted D127573: [mlir][spirv] Use UnrealizedConversionCast in ArithmeticToSPIRV.
Mon, Jun 13, 9:43 AM · Restricted Project, Restricted Project
ThomasRaoux committed rG1c84800c42d2: [mlir][vector] Add patterns to ppropagate vector distribution (authored by ThomasRaoux).
[mlir][vector] Add patterns to ppropagate vector distribution
Mon, Jun 13, 9:39 AM · Restricted Project, Restricted Project
ThomasRaoux closed D127167: [mlir][vector] Add patterns to propagate vector distribution.
Mon, Jun 13, 9:39 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127167: [mlir][vector] Add patterns to propagate vector distribution.
Mon, Jun 13, 9:27 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127167: [mlir][vector] Add patterns to propagate vector distribution.

Address review comments.

Mon, Jun 13, 9:22 AM · Restricted Project, Restricted Project

Sun, Jun 12

ThomasRaoux added inline comments to D127573: [mlir][spirv] Use UnrealizedConversionCast in ArithmeticToSPIRV.
Sun, Jun 12, 9:07 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D127572: [mlir][spirv] Convert single element vector.splat/fma.
Sun, Jun 12, 8:37 PM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127244: [mlir][AMDGPU] Add `mfma` operation to wrap mfma intrinsics..
Sun, Jun 12, 8:35 PM · Restricted Project, Restricted Project

Fri, Jun 10

ThomasRaoux accepted D127457: [mlir][nvgpu] shared memory access optimization pass.

LGTM!

Fri, Jun 10, 3:46 PM · Restricted Project, Restricted Project
ThomasRaoux committed rGed0288f7c4a5: [mlir][vector] Add patterns for vector distribution (authored by ThomasRaoux).
[mlir][vector] Add patterns for vector distribution
Fri, Jun 10, 10:47 AM · Restricted Project, Restricted Project
ThomasRaoux closed D127152: [mlir][vector] Add patterns for vector distribution.
Fri, Jun 10, 10:47 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127152: [mlir][vector] Add patterns for vector distribution.
Fri, Jun 10, 10:33 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127152: [mlir][vector] Add patterns for vector distribution.

Address review comments

Fri, Jun 10, 10:28 AM · Restricted Project, Restricted Project
ThomasRaoux added a comment to D127457: [mlir][nvgpu] shared memory access optimization pass.

This looks great! Few minor comments.

Fri, Jun 10, 8:58 AM · Restricted Project, Restricted Project

Wed, Jun 8

ThomasRaoux accepted D127004: [mlir][vector] Allow unroll of vector.contract, vector.transfer_read/write in arbitrary order.

Thanks, LGTM

Wed, Jun 8, 3:03 PM · Restricted Project, Restricted Project

Tue, Jun 7

ThomasRaoux accepted D127265: [mlir][spirv] NFC: fix typo in UnifyAliasedResourcePass pass.
Tue, Jun 7, 9:25 PM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127167: [mlir][vector] Add patterns to propagate vector distribution.
Tue, Jun 7, 8:22 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127167: [mlir][vector] Add patterns to propagate vector distribution.

Update comment

Tue, Jun 7, 8:22 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127167: [mlir][vector] Add patterns to propagate vector distribution.
Tue, Jun 7, 8:16 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.
Tue, Jun 7, 8:09 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.
Tue, Jun 7, 8:08 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.

Address review comments

Tue, Jun 7, 8:08 AM · Restricted Project, Restricted Project

Mon, Jun 6

ThomasRaoux requested review of D127176: [mlir][vector] Add pattern to distribute vector reduction to GPU shuffles.
Mon, Jun 6, 7:40 PM · Restricted Project, Restricted Project
ThomasRaoux retitled D127167: [mlir][vector] Add patterns to propagate vector distribution from [mlir][vector] Add patterns to ppropagate vector distribution to [mlir][vector] Add patterns to propagate vector distribution.
Mon, Jun 6, 4:53 PM · Restricted Project, Restricted Project
ThomasRaoux requested review of D127167: [mlir][vector] Add patterns to propagate vector distribution.
Mon, Jun 6, 4:42 PM · Restricted Project, Restricted Project
ThomasRaoux requested review of D127152: [mlir][vector] Add patterns for vector distribution.
Mon, Jun 6, 2:04 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D126386: [ModuloSchedule] Fix terminator update when peeling..
Mon, Jun 6, 12:48 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D126995: [mlir][linalg] add conv_2d_nhwc_fhwc named op.

LGTM

Mon, Jun 6, 10:03 AM · Restricted Project, Restricted Project

Sun, Jun 5

ThomasRaoux accepted D127086: [mlir][linalg] fix crash when promoting rank-reducing memref.subviews.
Sun, Jun 5, 10:17 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D127004: [mlir][vector] Allow unroll of vector.contract, vector.transfer_read/write in arbitrary order.

Thanks! Looks good to me, I added few style related comments.

Sun, Jun 5, 9:56 PM · Restricted Project, Restricted Project

Fri, Jun 3

ThomasRaoux accepted D127000: [mlir][linalg] fix crash in vectorization of elementwise operations.

Good catch!

Fri, Jun 3, 2:04 PM · Restricted Project, Restricted Project
ThomasRaoux added a comment to D126995: [mlir][linalg] add conv_2d_nhwc_fhwc named op.

Can you also add a test here:
https://github.com/llvm/llvm-project/blob/main/mlir/test/Dialect/Linalg/named-ops.mlir

Fri, Jun 3, 12:57 PM · Restricted Project, Restricted Project

Thu, Jun 2

ThomasRaoux committed rG271a48e02917: [mlir][VectorToGPU] Fix bug generating incorrect ldmatrix ops (authored by ThomasRaoux).
[mlir][VectorToGPU] Fix bug generating incorrect ldmatrix ops
Thu, Jun 2, 9:30 PM · Restricted Project, Restricted Project
ThomasRaoux closed D126846: [mlir][VectorToGPU] Fix bug generating incorrect ldmatrix ops.
Thu, Jun 2, 9:30 PM · Restricted Project, Restricted Project
ThomasRaoux committed rG205c08b54d70: [mlir][scf] Add option to loop pipelining to not peel the epilogue (authored by ThomasRaoux).
[mlir][scf] Add option to loop pipelining to not peel the epilogue
Thu, Jun 2, 9:21 PM · Restricted Project, Restricted Project
ThomasRaoux closed D126753: [mlir][scf] Add option to loop pipelining to not peel the epilogue.
Thu, Jun 2, 9:20 PM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D126753: [mlir][scf] Add option to loop pipelining to not peel the epilogue.
Thu, Jun 2, 9:13 PM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D126753: [mlir][scf] Add option to loop pipelining to not peel the epilogue.
Thu, Jun 2, 9:06 PM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D126753: [mlir][scf] Add option to loop pipelining to not peel the epilogue.

Address review comments

Thu, Jun 2, 9:05 PM · Restricted Project, Restricted Project

Wed, Jun 1

ThomasRaoux added a comment to D126846: [mlir][VectorToGPU] Fix bug generating incorrect ldmatrix ops.

The intrinsic returns i32 values. So I thought as long as the data type is <= 32 bits and you are reading 8x128bit rows (8xf16, 4xf32, 16xi8, etc), then there is no problem.

In your test you are reading a 8x8xf32 B operand. So I was under the impression such an operand could be loaded with two ldmatrix calls, which loads two 8x128bit tiles. The distributed values (one per tile / thread) would be returned as two i32 values.

Wed, Jun 1, 10:19 PM · Restricted Project, Restricted Project
ThomasRaoux requested review of D126846: [mlir][VectorToGPU] Fix bug generating incorrect ldmatrix ops.
Wed, Jun 1, 7:46 PM · Restricted Project, Restricted Project

Tue, May 31

ThomasRaoux requested review of D126753: [mlir][scf] Add option to loop pipelining to not peel the epilogue.
Tue, May 31, 10:47 PM · Restricted Project, Restricted Project

May 26 2022

ThomasRaoux accepted D126454: [mlir][Linalg] Relax vectorization condition to allow transposed output..
May 26 2022, 8:16 AM · Restricted Project, Restricted Project

May 24 2022

ThomasRaoux committed rG89aaa2d03327: [mlir][vector] Add new lowering mode to vector.contractionOp (authored by ThomasRaoux).
[mlir][vector] Add new lowering mode to vector.contractionOp
May 24 2022, 7:24 AM · Restricted Project, Restricted Project
ThomasRaoux closed D126120: [mlir][vector] Add new lowering mode to vector.contractionOp.
May 24 2022, 7:23 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D126120: [mlir][vector] Add new lowering mode to vector.contractionOp.
May 24 2022, 7:18 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D126120: [mlir][vector] Add new lowering mode to vector.contractionOp.

Address review comments and fix bug

May 24 2022, 7:17 AM · Restricted Project, Restricted Project

May 23 2022

ThomasRaoux added inline comments to D123597: [mlir][vector] Fix CombineContractBroadcast for invalid cases.
May 23 2022, 4:04 PM · Restricted Project, Restricted Project

May 22 2022

ThomasRaoux updated the diff for D126120: [mlir][vector] Add new lowering mode to vector.contractionOp.
May 22 2022, 1:34 PM · Restricted Project, Restricted Project

May 21 2022

ThomasRaoux requested review of D126120: [mlir][vector] Add new lowering mode to vector.contractionOp.
May 21 2022, 8:50 AM · Restricted Project, Restricted Project

May 20 2022

ThomasRaoux accepted D126092: [mlir][NvGpuToNVVM] Fix missing i4 support for nvgpu.mma.sync.
May 20 2022, 5:27 PM · Restricted Project, Restricted Project
ThomasRaoux accepted D125838: [mlir][NvGpuToNVVM] Fix byte size calculation in async copy lowering.
May 20 2022, 5:17 PM · Restricted Project, Restricted Project

May 19 2022

ThomasRaoux committed rG4c1b65e7bc92: [mlir][vector] Fix crash in DropInnerMostUnitDims pattern (authored by ThomasRaoux).
[mlir][vector] Fix crash in DropInnerMostUnitDims pattern
May 19 2022, 10:40 AM · Restricted Project, Restricted Project
ThomasRaoux closed D125984: [mlir][vector] Fix crash in DropInnerMostUnitDims pattern.
May 19 2022, 10:40 AM · Restricted Project, Restricted Project
ThomasRaoux committed rGf2676b151d6f: [mlir][tensor] Add canonicalization for tensor.cast from extract_slice (authored by ThomasRaoux).
[mlir][tensor] Add canonicalization for tensor.cast from extract_slice
May 19 2022, 10:36 AM · Restricted Project, Restricted Project
ThomasRaoux closed D125972: [mlir][tensor] Add canonicalization for tensor.cast from extract_slice.
May 19 2022, 10:35 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D125972: [mlir][tensor] Add canonicalization for tensor.cast from extract_slice.

update test

May 19 2022, 10:32 AM · Restricted Project, Restricted Project
ThomasRaoux updated the diff for D125972: [mlir][tensor] Add canonicalization for tensor.cast from extract_slice.

remove debug code

May 19 2022, 9:53 AM · Restricted Project, Restricted Project
ThomasRaoux added inline comments to D125972: [mlir][tensor] Add canonicalization for tensor.cast from extract_slice.
May 19 2022, 9:46 AM · Restricted Project, Restricted Project