banach-space wrote:
**UPDATE: 31/7/25**
* Implemented https://github.com/llvm/llvm-project/pull/151503 (effectively
NFC) and rebased this PR on top of it to help keep the noise level contained.
* The latest
[commit](https://github.com/llvm/llvm-project/pull/149293/commits/97fbccdc118f1f3da28b9
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/149293
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
banach-space wrote:
**UPDATE: 30/7/25**
* This
[commit](https://github.com/llvm/llvm-project/pull/149293/commits/56108b1df69e150c475adc58880ca7dce5355b21)
addresses the remaining comments from @hanhanW .
* I have rebased this PR on top of
https://github.com/llvm/llvm-project/pull/151334. Thi
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/149293
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
banach-space wrote:
> I'm not very familiar with the linalg vectorization, so I might've left more
> questions than comments, sorry about that!
No need to apologize! You took the time to review my PR, and it’s totally
expected that you'd have questions - I'm more than happy to clarify things
@@ -1860,25 +1866,54 @@ vectorizeAsTensorUnpackOp(RewriterBase &rewriter,
linalg::UnPackOp unpackOp,
auto destSize = unpackOp.getDestRank();
- if (!inputVectorSizes.empty())
-assert(inputVectorSizes.size() == destSize &&
+ if (!inputVectorSizes.empty()) {
+assert
@@ -940,34 +940,113 @@ module attributes {transform.with_named_sequence} {
///
// CHECK-LABEL: func @test_vectorize_dynamic_shapes_unpack
-// CHECK-SAME: %[[ARG_0:.*]]: tensor,
-func.
@@ -1860,25 +1866,54 @@ vectorizeAsTensorUnpackOp(RewriterBase &rewriter,
linalg::UnPackOp unpackOp,
auto destSize = unpackOp.getDestRank();
- if (!inputVectorSizes.empty())
-assert(inputVectorSizes.size() == destSize &&
+ if (!inputVectorSizes.empty()) {
+assert
@@ -1902,17 +1937,20 @@ vectorizeAsTensorUnpackOp(RewriterBase &rewriter,
linalg::UnPackOp unpackOp,
// After applying outer_dims_perm: [8, 16]
// After appending the rest of the sourceShape: [8, 16, 32, 16]
- SmallVector readVectorSizes(vectorSizes.begin(), vectorSi
https://github.com/banach-space updated
https://github.com/llvm/llvm-project/pull/149293
From 159b519fbf48c962507191c4e3e5622e40595da6 Mon Sep 17 00:00:00 2001
From: Andrzej Warzynski
Date: Wed, 16 Jul 2025 17:08:55 +
Subject: [PATCH 1/7] [mlir][linalg] Enable scalable vectorization of
lin
@@ -2487,12 +2531,14 @@ vectorizeScalableVectorPrecondition(Operation *op,
if (numOfScalableDims == 0)
return success();
+ // TODO: Check the following!
banach-space wrote:
Let me remove this. It was meant as a reminder for myself to double check this.
https://github.com/banach-space updated
https://github.com/llvm/llvm-project/pull/149293
From 159b519fbf48c962507191c4e3e5622e40595da6 Mon Sep 17 00:00:00 2001
From: Andrzej Warzynski
Date: Wed, 16 Jul 2025 17:08:55 +
Subject: [PATCH 1/6] [mlir][linalg] Enable scalable vectorization of
lin
@@ -1173,7 +1252,7 @@ module attributes {transform.with_named_sequence} {
func.func @test_vectorize_padded_pack(%arg0: tensor<32x7x15xf32>, %arg1:
tensor<32x4x1x16x2xf32>) -> tensor<32x4x1x16x2xf32> {
%pad = arith.constant 0.00e+00 : f32
- %pack = linalg.pack %arg0 pad
https://github.com/banach-space updated
https://github.com/llvm/llvm-project/pull/149293
From 159b519fbf48c962507191c4e3e5622e40595da6 Mon Sep 17 00:00:00 2001
From: Andrzej Warzynski
Date: Wed, 16 Jul 2025 17:08:55 +
Subject: [PATCH 1/5] [mlir][linalg] Enable scalable vectorization of
lin
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/149293
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
banach-space wrote:
Should we also update:
```
- The following relationship for the tiled dimensions holds:
`shape(result)[inner_dims_pos[i]] = shape(source)[inner_dims_pos[i]] /
inner_tiles[i]`.
```
as
```
- The following relationship for the ti
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/149624
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -150,9 +150,15 @@ def Linalg_PackOp : Linalg_RelayoutOp<"pack", [
`padding_value` specifies a padding value at the boundary on non-perfectly
divisible dimensions. Padding is optional:
-- If absent, it is UB if the tile does not perfectly divide the dimension.
+
https://github.com/banach-space commented:
Thanks for the updates!
I think that having two separate PRs makes more sense. Code-wise, these are
relatively small patches. But the underlying logic that's updated is a bit
fiddly and every change deserve a dedicated commit summary :)
I've left so
@@ -150,9 +150,13 @@ def Linalg_PackOp : Linalg_RelayoutOp<"pack", [
`padding_value` specifies a padding value at the boundary on non-perfectly
divisible dimensions. Padding is optional:
-- If absent, it is UB if the tile does not perfectly divide the dimension.
+
@@ -1824,27 +1825,47 @@ func.func @unpack_invalid_outer_dims_perm(%source:
tensor<128x256xf32>, %dest: t
// -
+func.func @pack_with_artificial_padding(%input: tensor<9xf32>, %output:
tensor<3x8xf32>) -> tensor<3x8xf32> {
+ %cst = arith.constant 0.0 : f32
+ // expected
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/149624
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -150,9 +150,13 @@ def Linalg_PackOp : Linalg_RelayoutOp<"pack", [
`padding_value` specifies a padding value at the boundary on non-perfectly
divisible dimensions. Padding is optional:
-- If absent, it is UB if the tile does not perfectly divide the dimension.
+
https://github.com/banach-space approved this pull request.
LGTM, thanks for splitting this into two!
https://github.com/llvm/llvm-project/pull/148198
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bi
https://github.com/banach-space updated
https://github.com/llvm/llvm-project/pull/149293
From 5deba69e57d2e13927987de919fecd30a87ac33d Mon Sep 17 00:00:00 2001
From: Andrzej Warzynski
Date: Wed, 16 Jul 2025 17:08:55 +
Subject: [PATCH] [mlir][linalg] Enable scalable vectorization of linalg.u
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/149293
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space approved this pull request.
https://github.com/llvm/llvm-project/pull/149488
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space approved this pull request.
https://github.com/llvm/llvm-project/pull/149487
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/143146
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -298,16 +298,139 @@ struct LegalizeSVEMaskLoadConversion : public
OpRewritePattern {
}
};
+/// Transforms a `transfer_read` operation so it reads vector of a type that
+/// can be mapped to an LLVM type. This is done by collapsing trailing
+/// dimensions so we obtain a
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
+// CHECK: %[[COLLAPSE:.+]]
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
+// CHECK: %[[COLLAPSE:.+]]
@@ -298,16 +298,139 @@ struct LegalizeSVEMaskLoadConversion : public
OpRewritePattern {
}
};
+/// Transforms a `transfer_read` operation so it reads vector of a type that
+/// can be mapped to an LLVM type. This is done by collapsing trailing
+/// dimensions so we obtain a
@@ -298,16 +298,139 @@ struct LegalizeSVEMaskLoadConversion : public
OpRewritePattern {
}
};
+/// Transforms a `transfer_read` operation so it reads vector of a type that
+/// can be mapped to an LLVM type. This is done by collapsing trailing
+/// dimensions so we obtain a
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
banach-space wrote:
Is i
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
+// CHECK: %[[COLLAPSE:.+]]
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
+// CHECK: %[[COLLAPSE:.+]]
@@ -298,16 +298,139 @@ struct LegalizeSVEMaskLoadConversion : public
OpRewritePattern {
}
};
+/// Transforms a `transfer_read` operation so it reads vector of a type that
+/// can be mapped to an LLVM type. This is done by collapsing trailing
+/// dimensions so we obtain a
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
+// CHECK: %[[COLLAPSE:.+]]
@@ -298,16 +298,139 @@ struct LegalizeSVEMaskLoadConversion : public
OpRewritePattern {
}
};
+/// Transforms a `transfer_read` operation so it reads vector of a type that
+/// can be mapped to an LLVM type. This is done by collapsing trailing
+/// dimensions so we obtain a
@@ -0,0 +1,262 @@
+// RUN: mlir-opt --arm-sve-legalize-vector-storage --split-input-file %s |
FileCheck %s
+
+// -
+
+// CHECK-LABEL: @test_base_case
+// CHECK-SAME: %[[I:arg0]]: index, %[[J:arg1]]: index, %[[M:arg2]]:
+// CHECK: %[[COLLAPSE:.+]]
banach-space wrote:
[nit] Avoid using the word `test` in test function names. It's just noise that
doesn't add any new info. Instead, try to convey what makes a particular test
case unique. See here for MLIR guidelines:
https://mlir.llvm.org/getting_started/T
https://github.com/banach-space commented:
Great work, Momchil - thank you!
I've left a number of comments, but nothing major. My main high-level
suggestion is to follow the guidance in [MLIR's Testing
Guide](https://mlir.llvm.org/getting_started/TestingGuide/#contributor-guidelines)
a bit mo
@@ -0,0 +1,336 @@
+// REQUIRES: arm-emulator
+
+// DEFINE: %{compile} = mlir-opt %s \
+// DEFINE: --convert-vector-to-scf --convert-scf-to-cf
--convert-vector-to-llvm='enable-arm-neon enable-arm-i8mm' \
+// DEFINE: --expand-strided-metadata --convert-to-llvm
--finalize-memr
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/144699
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space approved this pull request.
Thanks, it's great to see more tests for `i8mm`. The documentation makes it
relatively easy to follow (despite this being fairly complex!) - that's much
appreciated!
Overall LGTM, but I have one request. Could you unify the input data
@@ -83,16 +84,48 @@ func.func @transfer_read_dims_mismatch_contiguous(
return %res : vector<1x1x2x2xi8>
}
-// CHECK-LABEL: func.func @transfer_read_dims_mismatch_contiguous(
+// CHECK-LABEL: func.func @transfer_read_dims_mismatch_contiguous_unit_dims(
// CHECK-SAME:
https://github.com/banach-space approved this pull request.
LGTM, thanks!
https://github.com/llvm/llvm-project/pull/140572
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bran
banach-space wrote:
Thanks - great to finally be reaching this stage! I have a few high-level
questions and suggestions:
**1. Why is the scalable dimension always [4]?**
From the current tests, it looks like the scalable dim is always `[4]`. Could
you remind me why that value is chosen?
**2.
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/141567
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/141567
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/141567
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/139706
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/139706
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/139706
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
https://github.com/banach-space commented:
Thanks Momchil - this is great!
I skimmed through the pattern logic, and it's very neatly written. It's
actually quite easy to follow, despite the underlying logic being a bit
convoluted - well done! I've left a few minor suggestions, but nothing majo
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/135636
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
@@ -0,0 +1,304 @@
+//===- LowerContractionToSMMLAPattern.cpp - Contract to SMMLA ---*- C++
-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM
Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Ap
https://github.com/banach-space commented:
Thanks! This one is a bit longer, so I may need to wait till Thursday before I
can review.
One high-level question - would sharing some code between NEON and SVE be
possible?
https://github.com/llvm/llvm-project/pull/135636
__
@@ -273,6 +273,34 @@ def UmmlaOp : ArmSVE_Op<"ummla",
"$acc `,` $src1 `,` $src2 attr-dict `:` type($src1) `to` type($dst)";
}
+def UsmmlaOp : ArmSVE_Op<"usmmla", [Pure,
+AllTypesMatch<["src1", "src2"]>,
+
https://github.com/banach-space approved this pull request.
LGTM % nit
Thanks!
https://github.com/llvm/llvm-project/pull/135634
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llv
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/135634
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
https://github.com/banach-space edited
https://github.com/llvm/llvm-project/pull/133286
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -1033,7 +1033,7 @@ llvm.func @scalable_vector() -> i16 {
llvm.func @scalable_llvm_vector() -> i16 {
%0 = llvm.mlir.constant(1 : i32) : i32
// CHECK: llvm.alloca
- %1 = llvm.alloca %0 x !llvm.vec : (i32) -> !llvm.ptr
+ %1 = llvm.alloca %0 x vector<[4] x !llvm.ppc_fp128>
@@ -150,8 +150,7 @@ generatedTypePrinter(Type def, AsmPrinter &printer);
bool LLVMArrayType::isValidElementType(Type type) {
return !llvm::isa(
- type);
+LLVMFunctionType, LLVMTokenType>(type);
banach-space wrote:
That `LLVMScalable
https://github.com/banach-space approved this pull request.
Nice cleanup, thanks! LGTM
https://github.com/llvm/llvm-project/pull/133286
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listi
https://github.com/banach-space approved this pull request.
The non-GPU changes LGTM. The CHECK lines removed in tests were just dead code,
so thanks for the clean-up!
The GPU parts look reasonable, but it might be worth waiting a few days in case
someone more experienced wants to take a look.
https://github.com/banach-space approved this pull request.
https://github.com/llvm/llvm-project/pull/100155
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
@@ -894,6 +894,7 @@ static bool parseDiagArgs(CompilerInvocation &res,
llvm::opt::ArgList &args,
/// options accordingly. Returns false if new errors are generated.
static bool parseDialectArgs(CompilerInvocation &res, llvm::opt::ArgList &args,
banach-space wro
79 matches
Mail list logo
