Page MenuHomePhabricator
Feed Advanced Search

Today

bondhugula added a comment to D96397: Add C++ helpers to manage Unranked Memref in native code.

I haven't worked around the specific requirement in the CRunnerUtils file and I suspect it requires some refactoring. But I don't know also the exact use-case and how to test it in order to be able to propose a path forward here, @nicolasvasilache can you help figure out who's using this and how?

I don't either which is why the best proxy I found is mlir/lib/ExecutionEngine/CMakeLists.txt:set_property(TARGET mlir_c_runner_utils PROPERTY CXX_STANDARD 11).
I don't think there are active users though so this could be your window of opportunity to move everything into RunnerUtils as I suggested above.

I'm not sure how I could "move" things without knowing who depends on the things I'd move :)

That said my best shot at a plan for a future layering would be:

  • A low-level layer to manipulate memref descriptor written purely in C. That would make it a C runtime very suitable for embedding on micro environment as well I think.
  • A higher-level layer written in modern C++ with all the possible niceties one want. This layer could target the low-level C API to manipulate the descriptors, that said it is possible that template+inline would give better perf for C++ users.

WDYT?
@bondhugula as well?

Thu, Apr 15, 3:44 AM · Restricted Project

Yesterday

bondhugula added a comment to D95330: [MLIR][GPU][NVVM] Add warp synchronous matrix-multiply accumulate ops.

Hi @ThomasRaoux,
Sorry for the late reply. Great to hear that these ops can be reused in the IREE pipeline too. I was actually busy in some parallel work using these ops and getting it ready for an upcoming submission. The comments regarding the types are still to be addressed. I will surely be working on this, But I will get started on any major changes only by next week. As you mention, It would be great to know what your plans are and how you wish to proceed.

Thanks!

Hi @navdeepkk, great to hear you were able to use those ops. Next week sounds good, if you think it you won't have bandwidth to progress in the next couple weeks please let me know and hopefully offload some of the work to me.
To give more details on what I'm doing on IREE side is to try to match what cutlass does using MLIR transformations. The flow will look like: linalg on tensor -> tiling along M,N,K and distribute to blocks -> promote operands to shared memory -> (double buffering and pipelining) -> Tiling to wrap -> tiling shared memory copy -> linalg vectorization -> vector.contract unrolling to mma size -> vector to GPU to create MMA GPU ops -> lowering to llvm/nvvm.
A lot of this infrastructure is already there in IREE.

If you are able to share, it would be great to hear about the flow you used those ops with.

@ThomasRaoux Just to clarify: @navdeepkk has all of the MLIR-based GPU code generation implemented on top of these ops (the ones in this revision) already working using the affine dialect infrastructure: we don't use IREE, linalg, or the vector dialects, but our pipeline effectively includes all of the transformations you list above and several more in between those lowerings.

Getting to this PR itself, are there specific changes to make before this can be landed? @ftynse and others (Stephan and Mehdi) appear to be fine with having these ops in the GPU dialect. Could you please list those changes out if any or LGTM this? This will allow further development and progress upstream. If the remaining changes are minor, these could be made quickly and landed.

Thanks for explaining, @bondhugula. It is great to know that you have all those transformations. Is it publicly available? (or will it be?).
I 100% support having these ops in the GPU dialect. The only concern I had with the PR was about the type including vector instead of being completely opaque because as I mentioned this will make it harder to share with Vulkan path. However I don't think it had to be a blocker and it is something that we can iterate on so I LGTM the patch.

Wed, Apr 14, 12:20 AM · Restricted Project

Tue, Apr 13

bondhugula added a comment to D100077: Add support for numpy arrays to memref conversions..

@bondhugula @nicolasvasilache : can you confirm the recent changes address your previous comments?

Tue, Apr 13, 10:45 PM · Restricted Project
bondhugula added a comment to D95330: [MLIR][GPU][NVVM] Add warp synchronous matrix-multiply accumulate ops.

Hi @ThomasRaoux,
Sorry for the late reply. Great to hear that these ops can be reused in the IREE pipeline too. I was actually busy in some parallel work using these ops and getting it ready for an upcoming submission. The comments regarding the types are still to be addressed. I will surely be working on this, But I will get started on any major changes only by next week. As you mention, It would be great to know what your plans are and how you wish to proceed.

Thanks!

Hi @navdeepkk, great to hear you were able to use those ops. Next week sounds good, if you think it you won't have bandwidth to progress in the next couple weeks please let me know and hopefully offload some of the work to me.
To give more details on what I'm doing on IREE side is to try to match what cutlass does using MLIR transformations. The flow will look like: linalg on tensor -> tiling along M,N,K and distribute to blocks -> promote operands to shared memory -> (double buffering and pipelining) -> Tiling to wrap -> tiling shared memory copy -> linalg vectorization -> vector.contract unrolling to mma size -> vector to GPU to create MMA GPU ops -> lowering to llvm/nvvm.
A lot of this infrastructure is already there in IREE.

If you are able to share, it would be great to hear about the flow you used those ops with.

Tue, Apr 13, 8:50 PM · Restricted Project
bondhugula accepted D99761: [mlir] Prevent operations with users from being hoisted.

Looks great.

Tue, Apr 13, 11:12 AM · Restricted Project
bondhugula added inline comments to D99761: [mlir] Prevent operations with users from being hoisted.
Tue, Apr 13, 10:27 AM · Restricted Project

Mon, Apr 12

bondhugula added a comment to D98041: [MLIR] Create memref dialect and move dialect-specific ops from std..

I think just a patch splitting the op into tensor.dim and memref.dim would be fine. There isn't much to discuss.

So there was a discussion that this could be done via interfaces, for example. We could have a generic dim and rank operation that can be applied to any ShapedType type. That would also make it easier to rewrite between these types, as the dim and rank operation would no longer need to change.

Just to be clear, I am not saying we should not do this split, I am just trying to avoid the extra work if soon later we decide that there is a different and better way to model this. This is more work than "just a patch", especially for downstream users, so we should avoid unnecessary churn. As far as I understand it, this is about an extra link time dependency that is not worse than the monolithic dialect before.

ShapedType only covers tensor and memref. I'm not currently aware of any effort to generalize that, so splitting it now seems fairly safe.

Mon, Apr 12, 8:52 PM · Restricted Project
bondhugula committed rG9cc417cbca1c: [mlir][affine] Fix unfolded bounding maps for affine.for (authored by eopXD).
[mlir][affine] Fix unfolded bounding maps for affine.for
Mon, Apr 12, 11:43 AM
bondhugula closed D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.
Mon, Apr 12, 11:43 AM · Restricted Project
bondhugula requested changes to D100268: [mlir] Canonicalize AllocOp's with only store and dealloc uses.
Mon, Apr 12, 9:00 AM · Restricted Project
bondhugula added a comment to D100119: [mlir] Rename AVX512 dialect to X86Vector.

Is there a proposal for the first non AVX op to be added to this and which one it would be?

There is a patch, D99818, which triggered the discussion about renaming. That patch adds an AVX op. I am highly supportive of renaming the dialect in a separate patch.

Mon, Apr 12, 8:30 AM · Restricted Project, Restricted Project
bondhugula requested changes to D100077: Add support for numpy arrays to memref conversions..
Mon, Apr 12, 5:21 AM · Restricted Project

Fri, Apr 9

bondhugula added a comment to D100119: [mlir] Rename AVX512 dialect to X86Vector.

We will soon be adding non-AVX512 operations to MLIR, such as AVX's rsqrt. In https://reviews.llvm.org/D99818 several possibilities were discussed, namely to (1) add non-AVX512 ops to
the AVX512 dialect, (2) add more dialects (e.g. AVX dialect for AVX rsqrt), and (3) expand the scope of the AVX512 to include these SIMD x86 ops, thereby renaming the dialect to
something more accurate such as X86Vector.
Consensus was reached on option (3), which this patch implements.

Fri, Apr 9, 11:55 PM · Restricted Project, Restricted Project
bondhugula accepted D99548: [mlir] introduce "encoding" attribute to tensor type.

This revision is missing an update to the tensor type documentation in mlir/doc/LangRef.md.

What part in LangRef.md do you want changed? Note that the tensor type doc is generated from the doc in "include/mlir/IR/BuiltinTypes.td", which is part of the CL already
(in the sense that I have a TODO, but without attribute interface I am not sure how much I can say)

Fri, Apr 9, 8:17 PM · Restricted Project

Thu, Apr 8

bondhugula accepted D96397: Add C++ helpers to manage Unranked Memref in native code.

Looks good. Could you add an additional line to the commit summary for better context? The commit title is a bit cryptic and the connection to runtime utils and runners is missing.

Thu, Apr 8, 9:27 PM · Restricted Project
bondhugula requested changes to D100119: [mlir] Rename AVX512 dialect to X86Vector.

There isn't a commit summary at all. It'd be good to explain the rationale for the renaming.

Thu, Apr 8, 9:11 PM · Restricted Project, Restricted Project

Wed, Apr 7

bondhugula requested changes to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..
Wed, Apr 7, 10:16 PM · Restricted Project
bondhugula requested changes to D99548: [mlir] introduce "encoding" attribute to tensor type.

This revision is missing an update to the tensor type documentation in mlir/doc/LangRef.md.

Wed, Apr 7, 9:00 PM · Restricted Project
bondhugula added a comment to D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

Hi @bondhugula,
I am trying to rebase with $ git rebase origin/main with the current HEAD on my patching commit.
Then I encountered merge conflicts of file:

CONFLICT (content): Merge conflict in libunwind/CMakeLists.txt
CONFLICT (content): Merge conflict in libcxx/CMakeLists.txt

Is this normal or how shall I do something different to rebase to main?

Wed, Apr 7, 8:47 PM · Restricted Project
bondhugula requested changes to D99761: [mlir] Prevent operations with users from being hoisted.
Wed, Apr 7, 8:43 PM · Restricted Project
bondhugula added a reviewer for D99761: [mlir] Prevent operations with users from being hoisted: vinayaka-polymage.
Wed, Apr 7, 8:37 PM · Restricted Project
bondhugula added inline comments to D100077: Add support for numpy arrays to memref conversions..
Wed, Apr 7, 8:36 PM · Restricted Project
bondhugula requested changes to D100077: Add support for numpy arrays to memref conversions..

I'd really like to see this support! Thanks for implementing this.

Wed, Apr 7, 8:35 PM · Restricted Project
bondhugula added reviewers for D100077: Add support for numpy arrays to memref conversions.: mehdi_amini, stellaraccident, bondhugula.
Wed, Apr 7, 8:27 PM · Restricted Project
bondhugula added a comment to D99172: [mlir] Introduce CloneOp and adapt test cases in BufferDeallocation..

@dfki-jugr But why does this revision remove the copy removal pass when there are things that generate copies? The extension to canonicalization done eliminates clone ops, but IR with copy ops will no longer have a way to remove copies? For eg. -linalg-bufferize would still generate copy operations, right? @silvas

I believe all uses of copy in linalg-bufferize can be replaced with memref.clone easily -- it's a much more natural representation and we should do that cleanup.

Wed, Apr 7, 10:46 AM · Restricted Project
bondhugula added a comment to D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

@bondhugula I am not able to commit this.
Please help me to commit, thank you.

Wed, Apr 7, 10:39 AM · Restricted Project
bondhugula accepted D99925: [MLIR] Add a switch operation to the standard dialect.
Wed, Apr 7, 10:30 AM · Restricted Project

Tue, Apr 6

bondhugula added a comment to D99172: [mlir] Introduce CloneOp and adapt test cases in BufferDeallocation..

@dfki-jugr But why does this revision remove the copy removal pass when there are things that generate copies? The extension to canonicalization done eliminates clone ops, but IR with copy ops will no longer have a way to remove copies? For eg. -linalg-bufferize would still generate copy operations, right? @silvas

Tue, Apr 6, 3:27 AM · Restricted Project
bondhugula added a comment to D99922: [MLIR][docs] Fixes to operation syntax in Lang Ref.

Nice catch! I think I caught another redundancy right below.

Tue, Apr 6, 2:55 AM · Restricted Project
bondhugula requested changes to D99925: [MLIR] Add a switch operation to the standard dialect.

Looks great - some minor comments.

Tue, Apr 6, 12:22 AM · Restricted Project

Sat, Apr 3

bondhugula added a comment to D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

@eopXD Are you able to commit this? If not, I can commit it for you.

Sat, Apr 3, 11:08 PM · Restricted Project

Thu, Apr 1

bondhugula requested changes to D99761: [mlir] Prevent operations with users from being hoisted.
Thu, Apr 1, 8:40 PM · Restricted Project
bondhugula accepted D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

Looks good.

Thu, Apr 1, 8:34 PM · Restricted Project
bondhugula added a comment to D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

Loop bounds of affine.load

Thu, Apr 1, 8:33 PM · Restricted Project
bondhugula requested changes to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..
Thu, Apr 1, 8:32 PM · Restricted Project
bondhugula removed a reviewer for D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout.: avarmapml.
Thu, Apr 1, 8:24 PM · Restricted Project
bondhugula committed rGdc537158d537: [MLIR][Affine] Add utility to check if the slice is valid (authored by vinayaka-polymage).
[MLIR][Affine] Add utility to check if the slice is valid
Thu, Apr 1, 2:23 AM
bondhugula closed D98239: [MLIR][Affine] Add utility to check if the slice is valid.
Thu, Apr 1, 2:22 AM · Restricted Project
bondhugula accepted D98239: [MLIR][Affine] Add utility to check if the slice is valid.
Thu, Apr 1, 2:05 AM · Restricted Project

Tue, Mar 30

bondhugula accepted D98239: [MLIR][Affine] Add utility to check if the slice is valid.
Tue, Mar 30, 7:04 AM · Restricted Project
bondhugula added a comment to D98239: [MLIR][Affine] Add utility to check if the slice is valid.

Looks great with the fast check. Relatively minor comments.

Tue, Mar 30, 7:02 AM · Restricted Project

Fri, Mar 26

bondhugula added a comment to D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

Adjust testcase to include lower bound composition,
also adjusted testcase to avoid numbered SSA.

Fri, Mar 26, 2:37 AM · Restricted Project

Thu, Mar 25

bondhugula added a reverting change for rG361b7d125b43: [Canonicalizer] Process regions top-down instead of bottom up & reuse existing…: rG0b20413ef671: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse….
Thu, Mar 25, 9:51 AM
bondhugula committed rG0b20413ef671: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse… (authored by bondhugula).
Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse…
Thu, Mar 25, 9:51 AM
bondhugula added a reverting change for D99006: [Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.: rG0b20413ef671: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse….
Thu, Mar 25, 9:51 AM · Restricted Project
bondhugula closed D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 9:51 AM · Restricted Project
bondhugula requested changes to D99323: [mlir][affine] Fix unfolded bounding maps for affine.for.

Thanks for improving this. A few comments.

Thu, Mar 25, 7:03 AM · Restricted Project
bondhugula updated the summary of D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 5:00 AM · Restricted Project
bondhugula updated the summary of D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 5:00 AM · Restricted Project
bondhugula updated the summary of D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 4:59 AM · Restricted Project
bondhugula updated the summary of D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 4:59 AM · Restricted Project
bondhugula updated the summary of D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 4:59 AM · Restricted Project
bondhugula updated the diff for D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".

Updated commit summary.

Thu, Mar 25, 4:45 AM · Restricted Project
bondhugula added a reverting change for rG361b7d125b43: [Canonicalizer] Process regions top-down instead of bottom up & reuse existing…: D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 4:37 AM
bondhugula requested review of D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 4:37 AM · Restricted Project
bondhugula added a reverting change for D99006: [Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.: D99329: Revert "[Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants.".
Thu, Mar 25, 4:37 AM · Restricted Project

Wed, Mar 17

bondhugula added inline comments to D98041: [MLIR] Create memref dialect and move dialect-specific ops from std..
Wed, Mar 17, 11:10 PM · Restricted Project
bondhugula committed rG8e3075c2b07e: [MLIR] Fix lowering of Affine IfOp in the presence of yield values. (authored by Shukla-Gaurav).
[MLIR] Fix lowering of Affine IfOp in the presence of yield values.
Wed, Mar 17, 4:05 AM
bondhugula closed D98760: [MLIR] Fix lowering of Affine IfOp in the presence of yield values..
Wed, Mar 17, 4:04 AM · Restricted Project
bondhugula accepted D98760: [MLIR] Fix lowering of Affine IfOp in the presence of yield values..
Wed, Mar 17, 3:18 AM · Restricted Project

Tue, Mar 16

bondhugula retitled D98760: [MLIR] Fix lowering of Affine IfOp in the presence of yield values. from [MLIR] Add lowering of Affine IfOp in the presence of yield values. to [MLIR] Fix lowering of Affine IfOp in the presence of yield values..
Tue, Mar 16, 11:22 PM · Restricted Project
bondhugula accepted D98760: [MLIR] Fix lowering of Affine IfOp in the presence of yield values..

Thanks for fixing this!
Nit: in the commit summary: introduces the lowering -> fixes the lowering
Similarly, in the commit title: Add -> Fix

Tue, Mar 16, 11:22 PM · Restricted Project

Mar 14 2021

bondhugula added inline comments to D98609: [Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants..
Mar 14 2021, 1:40 PM · Restricted Project
bondhugula added inline comments to D98609: [Canonicalizer] Process regions top-down instead of bottom up & reuse existing constants..
Mar 14 2021, 1:36 PM · Restricted Project

Mar 13 2021

bondhugula added inline comments to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..
Mar 13 2021, 3:50 AM · Restricted Project
bondhugula requested changes to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..

Some style related comments.

Mar 13 2021, 3:44 AM · Restricted Project
bondhugula resigned from D97622: [mlir] Add OpAdaptorVerifier trait.

Thanks. Please go with @rriddle's comments for the rest.

Mar 13 2021, 2:42 AM · Restricted Project
bondhugula added a comment to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..

Ping @avarmapml

Mar 13 2021, 2:41 AM · Restricted Project
bondhugula accepted D97542: [mlir]: Add folding for DimOp with AllocaOp operand.

LGTM.

Mar 13 2021, 2:40 AM · Restricted Project
bondhugula added inline comments to D98470: [mlir][amx] Add Intel AMX dialect (architectural-specific vector dialect).
Mar 13 2021, 2:34 AM · Restricted Project

Mar 12 2021

bondhugula added a comment to D98239: [MLIR][Affine] Add utility to check if the slice is valid.

Some minor comments.

Mar 12 2021, 9:01 PM · Restricted Project
bondhugula requested changes to D98239: [MLIR][Affine] Add utility to check if the slice is valid.
Mar 12 2021, 8:56 PM · Restricted Project

Mar 9 2021

bondhugula added a comment to D98203: [mlir] Add NVVM to CUBIN conversion to mlir-opt.

switch CUDA integration tests from mlir-cuda-runner to mlir-opt + mlir-cpu-runner

Mar 9 2021, 9:34 AM · Restricted Project
bondhugula added a comment to D98168: [mlir] Default for gpu-binary-annotation option..

Remove triple, target and feature options.

Mar 9 2021, 9:27 AM · Restricted Project
bondhugula accepted D98168: [mlir] Default for gpu-binary-annotation option..

A couple of minor comments.

Mar 9 2021, 9:26 AM · Restricted Project

Mar 8 2021

bondhugula added inline comments to D98171: [mlir] Use affine.apply when distributing to processors.
Mar 8 2021, 10:23 AM · Restricted Project

Mar 7 2021

bondhugula resigned from D97442: [mlir][Vector][Affine] Improve affine vectorizer algorithm.
Mar 7 2021, 10:47 PM · Restricted Project
bondhugula resigned from D97892: [mlir][Affine][Vector] Add initial support for 'iter_args' to Affine vectorizer..
Mar 7 2021, 10:47 PM · Restricted Project

Mar 4 2021

bondhugula added inline comments to D97893: [MLIR][TOSA] Added lowerings for Reduce operations to Linalg.
Mar 4 2021, 5:51 AM · Restricted Project
bondhugula accepted D97879: [mlir][Affine][NFC] Return BlockArgument in AffineForOp::getInductionVar.

Good catch!

Mar 4 2021, 5:41 AM · Restricted Project

Mar 2 2021

bondhugula added inline comments to D97542: [mlir]: Add folding for DimOp with AllocaOp operand.
Mar 2 2021, 2:27 AM · Restricted Project
bondhugula requested changes to D97542: [mlir]: Add folding for DimOp with AllocaOp operand.
Mar 2 2021, 2:14 AM · Restricted Project
bondhugula added a comment to D97542: [mlir]: Add folding for DimOp with AllocaOp operand.

A couple more comments @akuegel

Mar 2 2021, 2:11 AM · Restricted Project

Mar 1 2021

bondhugula added inline comments to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..
Mar 1 2021, 8:36 PM · Restricted Project
bondhugula requested changes to D97655: [mlir] Normalize dynamic memrefs with a map of tiled-layout..
Mar 1 2021, 12:58 AM · Restricted Project
bondhugula requested changes to D97542: [mlir]: Add folding for DimOp with AllocaOp operand.
Mar 1 2021, 12:56 AM · Restricted Project

Feb 27 2021

bondhugula added inline comments to D97590: [mlir][AttrDefGen] Add support for specifying the value type of an attribute.
Feb 27 2021, 9:02 PM · Restricted Project
bondhugula requested changes to D97622: [mlir] Add OpAdaptorVerifier trait.
Feb 27 2021, 8:51 PM · Restricted Project

Feb 26 2021

bondhugula committed rGce0f10a1d19b: [MLIR][affine] Certain Call Ops to prevent fusion (authored by vinayaka-polymage).
[MLIR][affine] Certain Call Ops to prevent fusion
Feb 26 2021, 1:59 AM
bondhugula closed D97252: [MLIR][affine] Certain Call Ops to prevent fusion.
Feb 26 2021, 1:59 AM · Restricted Project
bondhugula accepted D97252: [MLIR][affine] Certain Call Ops to prevent fusion.
Feb 26 2021, 1:49 AM · Restricted Project

Feb 25 2021

bondhugula accepted D97252: [MLIR][affine] Certain Call Ops to prevent fusion.

One minor comment.

Feb 25 2021, 9:55 PM · Restricted Project

Feb 24 2021

bondhugula added inline comments to D97252: [MLIR][affine] Certain Call Ops to prevent fusion.
Feb 24 2021, 8:14 PM · Restricted Project
bondhugula added inline comments to D97252: [MLIR][affine] Certain Call Ops to prevent fusion.
Feb 24 2021, 11:07 AM · Restricted Project

Feb 23 2021

bondhugula accepted D97217: [mlir] Extend Operation visitor with pre-order traversal.

This is nice! Thanks. Just a typo I caught. I think we should fix the iteration invalidation semantics in a subsequent revision (I can do that if you want.)

Feb 23 2021, 5:56 PM · Restricted Project
bondhugula requested changes to D97217: [mlir] Extend Operation visitor with pre-order traversal.
Feb 23 2021, 5:53 PM · Restricted Project
bondhugula added a comment to D97030: [MLIR][affine-loop-fusion] Handle defining ops between the source and dest loops.

Hi Uday and Diego,

Thank you so much for your comments! I just updated the patch.

I don't have permission to land this, please help me to do it. My "name (email)" is "Tung D. Le (tung@jp.ibm.com)".

Thanks,
Tung.

Feb 23 2021, 9:22 AM · Restricted Project

Feb 22 2021

bondhugula added inline comments to D97006: [MLIR] Add Linalg support for integer (generalized) matmuls.
Feb 22 2021, 1:01 PM · Restricted Project
bondhugula committed rG817d343fb048: [MLIR] Fix tilePerfectlyNested utility for handling non-unit step size (authored by vivekkhandelwal1).
[MLIR] Fix tilePerfectlyNested utility for handling non-unit step size
Feb 22 2021, 11:21 AM
bondhugula added inline comments to D97006: [MLIR] Add Linalg support for integer (generalized) matmuls.
Feb 22 2021, 11:13 AM · Restricted Project