<feed xmlns='http://www.w3.org/2005/Atom'>
<title>llvm-project.git/mlir/lib/Dialect, branch main</title>
<subtitle>Unnamed repository; edit this file 'description' to name the repository.
</subtitle>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/'/>
<entry>
<title>[MLIR] [XeGPU] Fix dropSgLayoutAndData &amp; dropInstData in SliceAttr (#168618)</title>
<updated>2025-11-21T20:40:16+00:00</updated>
<author>
<name>Nishant Patel</name>
<email>nishant.b.patel@intel.com</email>
</author>
<published>2025-11-21T20:40:16+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=778e104dee8186c8a8c8ed781692825fb3a8bec8'/>
<id>778e104dee8186c8a8c8ed781692825fb3a8bec8</id>
<content type='text'>
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
</pre>
</div>
</content>
</entry>
<entry>
<title>Revert "[MLIR][GPU] subgroup_mma fp64 extension" (#169049)</title>
<updated>2025-11-21T15:02:59+00:00</updated>
<author>
<name>Fabian Mora</name>
<email>fmora.dev@gmail.com</email>
</author>
<published>2025-11-21T15:02:59+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=8c3f59f1b297ec65b5870ebfd727a40f632de966'/>
<id>8c3f59f1b297ec65b5870ebfd727a40f632de966</id>
<content type='text'>
Reverts llvm/llvm-project#165873

The revert is triggered by a failing integration test on a couple of
buildbots.</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Reverts llvm/llvm-project#165873

The revert is triggered by a failing integration test on a couple of
buildbots.</pre>
</div>
</content>
</entry>
<entry>
<title>[MLIR][GPU] subgroup_mma fp64 extension (#165873)</title>
<updated>2025-11-21T14:07:43+00:00</updated>
<author>
<name>Giacomo Castiglioni</name>
<email>giacastiglioni@gmail.com</email>
</author>
<published>2025-11-21T14:07:43+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=49995b2af0abbec1095031dfe9eb049945b6d4f7'/>
<id>49995b2af0abbec1095031dfe9eb049945b6d4f7</id>
<content type='text'>
This PR extends the `gpu.subgroup_mma_*` ops to support fp64 type.
The extension requires special handling during the lowering to `nvvm`
due to the return type for load ops for fragment a and b (they return a
scalar instead of a struct).</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
This PR extends the `gpu.subgroup_mma_*` ops to support fp64 type.
The extension requires special handling during the lowering to `nvvm`
due to the return type for load ops for fragment a and b (they return a
scalar instead of a struct).</pre>
</div>
</content>
</entry>
<entry>
<title>[MLIR] Apply clang-tidy fixes for readability-container-size-empty in AffineOps.cpp (NFC)</title>
<updated>2025-11-21T12:15:31+00:00</updated>
<author>
<name>Mehdi Amini</name>
<email>joker.eph@gmail.com</email>
</author>
<published>2025-08-21T13:07:21+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=764c1d40ce89a2ece97d32240e5e943afdb2c9c9'/>
<id>764c1d40ce89a2ece97d32240e5e943afdb2c9c9</id>
<content type='text'>
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
</pre>
</div>
</content>
</entry>
<entry>
<title>[MLIR] Apply clang-tidy fixes for readability-container-size-empty in ShardOps.cpp (NFC)</title>
<updated>2025-11-21T11:57:31+00:00</updated>
<author>
<name>Mehdi Amini</name>
<email>joker.eph@gmail.com</email>
</author>
<published>2025-08-21T15:41:13+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=18d3db4bcd42e21e45b499a2999834904a925af0'/>
<id>18d3db4bcd42e21e45b499a2999834904a925af0</id>
<content type='text'>
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
</pre>
</div>
</content>
</entry>
<entry>
<title>[MLIR][XeGPU][TransformOps] Add slice_dims argument to set_op_layout_attr and set_desc_layout (#168929)</title>
<updated>2025-11-21T08:08:12+00:00</updated>
<author>
<name>Tuomas Kärnä</name>
<email>tuomas.karna@intel.com</email>
</author>
<published>2025-11-21T08:08:12+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=e9fc393a9e431d1a0aebc3fe448f3cf1668fbb34'/>
<id>e9fc393a9e431d1a0aebc3fe448f3cf1668fbb34</id>
<content type='text'>
`set_op_layout_attr` and `set_desc_layout` transform ops wrap
`xegpu.layout` in an `xegpu.slice` attribute if `slice_dims` argument is
set.</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
`set_op_layout_attr` and `set_desc_layout` transform ops wrap
`xegpu.layout` in an `xegpu.slice` attribute if `slice_dims` argument is
set.</pre>
</div>
</content>
</entry>
<entry>
<title>[mlir][linalg] Clean up op verifiers without custom checks(NFC) (#168712)</title>
<updated>2025-11-21T02:05:20+00:00</updated>
<author>
<name>Longsheng Mou</name>
<email>longshengmou@gmail.com</email>
</author>
<published>2025-11-21T02:05:20+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=bf4dc96bc15464fc4af9dde3a5a686a129abfdbd'/>
<id>bf4dc96bc15464fc4af9dde3a5a686a129abfdbd</id>
<content type='text'>
This PR removes op verifiers that do not implement any custom
verification logic.</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
This PR removes op verifiers that do not implement any custom
verification logic.</pre>
</div>
</content>
</entry>
<entry>
<title>[MLIR] [XeGPU] Add distribution pattern for vector.constant_mask from Wg To Sg (#168118)</title>
<updated>2025-11-20T23:00:57+00:00</updated>
<author>
<name>Nishant Patel</name>
<email>nishant.b.patel@intel.com</email>
</author>
<published>2025-11-20T23:00:57+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=310abe0e4b4ebb57976928cc0b520f9e878b54a7'/>
<id>310abe0e4b4ebb57976928cc0b520f9e878b54a7</id>
<content type='text'>
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
</pre>
</div>
</content>
</entry>
<entry>
<title>[mlir][SCF] Add `scf::tileAndFuseConsumer` that tiles a consumer into a given tiled loop nest. (#167634)</title>
<updated>2025-11-20T22:13:57+00:00</updated>
<author>
<name>MaheshRavishankar</name>
<email>1663364+MaheshRavishankar@users.noreply.github.com</email>
</author>
<published>2025-11-20T22:13:57+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=dbeda4f4199de87c7ecd8dabcc1c58944a7a37ad'/>
<id>dbeda4f4199de87c7ecd8dabcc1c58944a7a37ad</id>
<content type='text'>
The existing `scf::tileAndFuseConsumerOfSlices` takes a list of slices
(and loops they are part of), tries to find the consumer of these slices
(all slices are expected to be the same consumer), and then tiles the
consumer into the loop nest using the `TilingInterface`. A more natural
way of doing consumer fusion is to just start from the consumer, look
for operands that are produced by the loop nest passed in as `loops`
(presumably these loops are generated by tiling, but that is not a
requirement for consumer fusion). Using the consumer you can find the
slices of the operands that are accessed within the loop which you can
then use to tile and fuse the consumer (using `TilingInterface`). This
handles more naturally the case where multiple operands of the consumer
come from the loop nest.

The `scf::tileAndFuseConsumerOfSlices` was implemented as a mirror of
`scf::tileAndFuseProducerOfSlice`. For the latter, the slice has a
single producer for the source of the slice, which makes it a natural
way of specifying producer fusion. But for consumers, the result might
have multiple users, resulting in multiple candidates for fusion, as
well as a fusion candidate using multiple results from the tiled loop
nest. This means using slices
(`tensor.insert_slice`/`tensor.parallel_insert_slice`) as a hook for
consumer fusion turns out to be quite hard to navigate. The use of the
consumer directly avoids all those pain points. In time the
`scf::tileAndFuseConsumerOfSlices` should be deprecated in favor of
`scf::tileAndFuseConsumer`. There is a lot of tech-debt that has
accumulated in `scf::tileAndFuseConsumerOfSlices` that needs to be
cleanedup. So while that gets cleaned up, and required functionality is
moved to `scf::tileAndFuseConsumer`, the old path is still maintained.

The test for `scf::tileAndFuseConsumerUsingSlices` is copied to
`tile-and-fuse-consumer.mlir` to
`tile-and-fuse-consumer-using-slices.mlir`. All the tests that were
there in this file are now using the `tileAndFuseConsumer` method. The
test op `test.tile_and_fuse_consumer` is modified to call
`scf::tileAndFuseConsumer`, while a new op
`test.tile_and_fuse_consumer_of_slice` is used to keep the old path
tested while it is deprecated.

---------

Signed-off-by: MaheshRavishankar &lt;mahesh.ravishankar@gmail.com&gt;</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
The existing `scf::tileAndFuseConsumerOfSlices` takes a list of slices
(and loops they are part of), tries to find the consumer of these slices
(all slices are expected to be the same consumer), and then tiles the
consumer into the loop nest using the `TilingInterface`. A more natural
way of doing consumer fusion is to just start from the consumer, look
for operands that are produced by the loop nest passed in as `loops`
(presumably these loops are generated by tiling, but that is not a
requirement for consumer fusion). Using the consumer you can find the
slices of the operands that are accessed within the loop which you can
then use to tile and fuse the consumer (using `TilingInterface`). This
handles more naturally the case where multiple operands of the consumer
come from the loop nest.

The `scf::tileAndFuseConsumerOfSlices` was implemented as a mirror of
`scf::tileAndFuseProducerOfSlice`. For the latter, the slice has a
single producer for the source of the slice, which makes it a natural
way of specifying producer fusion. But for consumers, the result might
have multiple users, resulting in multiple candidates for fusion, as
well as a fusion candidate using multiple results from the tiled loop
nest. This means using slices
(`tensor.insert_slice`/`tensor.parallel_insert_slice`) as a hook for
consumer fusion turns out to be quite hard to navigate. The use of the
consumer directly avoids all those pain points. In time the
`scf::tileAndFuseConsumerOfSlices` should be deprecated in favor of
`scf::tileAndFuseConsumer`. There is a lot of tech-debt that has
accumulated in `scf::tileAndFuseConsumerOfSlices` that needs to be
cleanedup. So while that gets cleaned up, and required functionality is
moved to `scf::tileAndFuseConsumer`, the old path is still maintained.

The test for `scf::tileAndFuseConsumerUsingSlices` is copied to
`tile-and-fuse-consumer.mlir` to
`tile-and-fuse-consumer-using-slices.mlir`. All the tests that were
there in this file are now using the `tileAndFuseConsumer` method. The
test op `test.tile_and_fuse_consumer` is modified to call
`scf::tileAndFuseConsumer`, while a new op
`test.tile_and_fuse_consumer_of_slice` is used to keep the old path
tested while it is deprecated.

---------

Signed-off-by: MaheshRavishankar &lt;mahesh.ravishankar@gmail.com&gt;</pre>
</div>
</content>
</entry>
<entry>
<title>[mlir][spirv] Add support for SwitchOp (#168713)</title>
<updated>2025-11-20T15:19:10+00:00</updated>
<author>
<name>Igor Wodiany</name>
<email>igor.wodiany@imgtec.com</email>
</author>
<published>2025-11-20T15:19:10+00:00</published>
<link rel='alternate' type='text/html' href='https://git.belthelziquor.com/llvm-project.git/commit/?id=891b3cf63e160c83309b90728034ab832184c964'/>
<id>891b3cf63e160c83309b90728034ab832184c964</id>
<content type='text'>
The dialect implementation mostly copies the one of `cf.switch`, but
aligns naming to the SPIR-V spec.</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
The dialect implementation mostly copies the one of `cf.switch`, but
aligns naming to the SPIR-V spec.</pre>
</div>
</content>
</entry>
</feed>
