Skip to content

[mlir][spirv] Fix constant.mlir test #152015

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Aug 4, 2025
Merged

[mlir][spirv] Fix constant.mlir test #152015

merged 1 commit into from
Aug 4, 2025
+192 −176

Conversation

kuhar
Copy link
Member

@kuhar kuhar commented Aug 4, 2025

Fix forward a few issues instead of reverting recent PRs....

  1. Remove split so that the output is properly serialized. This got added in Reland "[mlir][spirv] Fix int type declaration duplication when serializing" #145687.
  2. Add missing extensions and capabilities so that the spirv-val passes (tensors_arm, linkage).
  3. Disable spirv-val test for arm tensor constants. These fail to verify. Added in [mlir][spirv] Enable (de)serialization of TensorARM to/from OpConstan… #151485.

Issue: #152012

@kuhar
[mlir][spirv] Fix constant.mlir test
d0902bd 
1. Remove split so that the output is properly serialized. This got
   added in llvm#145687.
2. Disable spir-val test for arm tensor constants. These fail to verify.
   Added in llvm#151485.
@kuhar kuhar requested a review from antiagainst as a code owner August 4, 2025 19:04
@llvmbot
Copy link
Member

llvmbot commented Aug 4, 2025 •
edited
Loading

@llvm/pr-subscribers-mlir-spirv

@llvm/pr-subscribers-mlir

Author: Jakub Kuderski (kuhar)

Changes

Fix forward a few issues instead of reverting recent PRs....

  1. Remove split so that the output is properly serialized. This got added in #145687.
  2. Add missing extensions and capabilities so that the spirv-val passes (tensors_arm, linkage).
  3. Disable spirv-val test for arm tensor constants. These fail to verify. Added in #151485.

Issue: #152012

Patch is 22.50 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/152015.diff

3 Files Affected:

  • (added) mlir/test/Target/SPIRV/arm-tensor-constant.mlir (+49)
  • (modified) mlir/test/Target/SPIRV/constant.mlir (+8-176)
  • (added) mlir/test/Target/SPIRV/replicated-composites-constant.mlir (+135)
diff --git a/mlir/test/Target/SPIRV/arm-tensor-constant.mlir b/mlir/test/Target/SPIRV/arm-tensor-constant.mlir
new file mode 100644
index 0000000000000..275e586f70634
--- /dev/null
+++ b/mlir/test/Target/SPIRV/arm-tensor-constant.mlir
@@ -0,0 +1,49 @@
+// RUN: mlir-translate --no-implicit-module --test-spirv-roundtrip %s | FileCheck %s
+// DISABLED: %if spirv-tools %{ mlir-translate --no-implicit-module --serialize-spirv %s | spirv-val %}
+
+// FIXME(#152012): Fix arm tensor constant validation errors and reenable spirv-val tests.
+
+spirv.module Logical Vulkan requires #spirv.vce<v1.3,
+             [VulkanMemoryModel, Shader, TensorsARM, Linkage], [SPV_KHR_vulkan_memory_model, SPV_ARM_tensors]> {
+  // CHECK-LABEL: @arm_tensor_of_i32
+  spirv.func @arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
+    // CHECK: {{%.*}} = spirv.Constant dense<{{\[}}[1, 2, 3], [4, 5, 6]]> : !spirv.arm.tensor<2x3xi32>
+    %0 = spirv.Constant dense<[[1, 2, 3], [4, 5, 6]]> : !spirv.arm.tensor<2x3xi32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
+  }
+
+  // CHECK-LABEL: @splat_arm_tensor_of_i32
+  spirv.func @splat_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
+    // CHECK: {{%.*}} = spirv.Constant dense<2> : !spirv.arm.tensor<2x3xi32>
+    %0 = spirv.Constant dense<2> : !spirv.arm.tensor<2x3xi32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
+  }
+
+  // CHECK-LABEL: @arm_tensor_of_f32
+  spirv.func @arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
+    // CHECK: {{%.*}} = spirv.Constant dense<{{\[}}[1.000000e+00, 2.000000e+00, 3.000000e+00], [4.000000e+00, 5.000000e+00, 6.000000e+00]]> : !spirv.arm.tensor<2x3xf32>
+    %0 = spirv.Constant dense<[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]>: !spirv.arm.tensor<2x3xf32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
+  }
+
+  // CHECK-LABEL: @splat_arm_tensor_of_f32
+  spirv.func @splat_arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
+    // CHECK: {{%.*}} = spirv.Constant dense<2.000000e+00> : !spirv.arm.tensor<2x3xf32>
+    %0 = spirv.Constant dense<2.0> : !spirv.arm.tensor<2x3xf32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
+  }
+
+  // CHECK-LABEL: @null_arm_tensor_of_i32
+  spirv.func @null_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
+    // CHECK: spirv.Constant dense<0> : !spirv.arm.tensor<2x3xi32>
+    %0 = spirv.Constant dense<0> : !spirv.arm.tensor<2x3xi32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
+  }
+
+  // CHECK-LABEL: @null_arm_tensor_of_f32
+  spirv.func @null_arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
+    // CHECK: spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<2x3xf32>
+    %0 = spirv.Constant dense<0.0> : !spirv.arm.tensor<2x3xf32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
+  }
+}
diff --git a/mlir/test/Target/SPIRV/constant.mlir b/mlir/test/Target/SPIRV/constant.mlir
index c81ceac072bd0..4838d3c510757 100644
--- a/mlir/test/Target/SPIRV/constant.mlir
+++ b/mlir/test/Target/SPIRV/constant.mlir
@@ -1,7 +1,12 @@
-// RUN: mlir-translate --no-implicit-module --split-input-file --test-spirv-roundtrip %s | FileCheck %s
-// RUN: %if spirv-tools %{ mlir-translate -no-implicit-module --split-input-file -serialize-spirv %s | spirv-val %}
+// RUN: mlir-translate --no-implicit-module --test-spirv-roundtrip %s | FileCheck %s
+// RUN: %if spirv-tools %{ mlir-translate --no-implicit-module --serialize-spirv %s | spirv-val %}
 
-spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader, Int64, Int16, Int8, Float64, Float16, CooperativeMatrixKHR], [SPV_KHR_vulkan_memory_model, SPV_KHR_cooperative_matrix]> {
+// Note: Since the output of this test (optionally) gets validated by spirv-val,
+// we cannot use splits.
+
+spirv.module Logical Vulkan requires #spirv.vce<v1.3,
+             [VulkanMemoryModel, Shader, Int64, Int16, Int8, Float64, Float16, CooperativeMatrixKHR, TensorsARM, Linkage],
+             [SPV_KHR_vulkan_memory_model, SPV_KHR_cooperative_matrix, SPV_ARM_tensors]> {
   // CHECK-LABEL: @bool_const
   spirv.func @bool_const() -> () "None" {
     // CHECK: spirv.Constant true
@@ -307,178 +312,5 @@ spirv.module Logical Vulkan requires #spirv.vce<v1.3, [VulkanMemoryModel, Shader
     spirv.ReturnValue %coop : !spirv.coopmatrix<16x16xi8, Subgroup, MatrixAcc>
   }
 
-  // CHECK-LABEL: @arm_tensor_of_i32
-  spirv.func @arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
-    // CHECK: {{%.*}} = spirv.Constant dense<{{\[}}[1, 2, 3], [4, 5, 6]]> : !spirv.arm.tensor<2x3xi32>
-    %0 = spirv.Constant dense<[[1, 2, 3], [4, 5, 6]]> : !spirv.arm.tensor<2x3xi32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
-  }
-
-  // CHECK-LABEL: @splat_arm_tensor_of_i32
-  spirv.func @splat_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
-    // CHECK: {{%.*}} = spirv.Constant dense<2> : !spirv.arm.tensor<2x3xi32>
-    %0 = spirv.Constant dense<2> : !spirv.arm.tensor<2x3xi32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
-  }
-
-  // CHECK-LABEL: @arm_tensor_of_f32
-  spirv.func @arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
-    // CHECK: {{%.*}} = spirv.Constant dense<{{\[}}[1.000000e+00, 2.000000e+00, 3.000000e+00], [4.000000e+00, 5.000000e+00, 6.000000e+00]]> : !spirv.arm.tensor<2x3xf32>
-    %0 = spirv.Constant dense<[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]>: !spirv.arm.tensor<2x3xf32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
-  }
-
-  // CHECK-LABEL: @splat_arm_tensor_of_f32
-  spirv.func @splat_arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
-    // CHECK: {{%.*}} = spirv.Constant dense<2.000000e+00> : !spirv.arm.tensor<2x3xf32>
-    %0 = spirv.Constant dense<2.0> : !spirv.arm.tensor<2x3xf32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
-  }
-
-  // CHECK-LABEL: @null_arm_tensor_of_i32
-  spirv.func @null_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
-    // CHECK: spirv.Constant dense<0> : !spirv.arm.tensor<2x3xi32>
-    %0 = spirv.Constant dense<0> : !spirv.arm.tensor<2x3xi32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
-  }
-
-  // CHECK-LABEL: @null_arm_tensor_of_f32
-  spirv.func @null_arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
-    // CHECK: spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<2x3xf32>
-    %0 = spirv.Constant dense<0.0> : !spirv.arm.tensor<2x3xf32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
-  }
-
   spirv.EntryPoint "GLCompute" @bool_const
 }
-
-// -----
-
-spirv.module Logical GLSL450 requires #spirv.vce<v1.0, [Shader, ReplicatedCompositesEXT], [SPV_EXT_replicated_composites]> {
-
-  // CHECK-LABEL: @splat_vector_i32
-  spirv.func @splat_vector_i32() -> (vector<3xi32>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [1 : i32] : vector<3xi32>
-    %1 = spirv.EXT.ConstantCompositeReplicate [1 : i32] : vector<3xi32>
-    spirv.ReturnValue %1 : vector<3xi32>
-  }
-
-  // CHECK-LABEL: @splat_array_of_i32
-  spirv.func @splat_array_of_i32() -> (!spirv.array<3 x i32>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [1 : i32] : !spirv.array<3 x i32>
-    %1 = spirv.EXT.ConstantCompositeReplicate [1 : i32] : !spirv.array<3 x i32>
-    spirv.ReturnValue %1 : !spirv.array<3 x i32>
-  }
-
-  // CHECK-LABEL: @splat_array_of_vectors_of_i32
-  spirv.func @splat_array_of_vectors_of_i32() -> (!spirv.array<3 x vector<2xi32>>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [dense<[1, 2]> : vector<2xi32>] : !spirv.array<3 x vector<2xi32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [dense<[1, 2]> : vector<2xi32>] : !spirv.array<3 x vector<2xi32>>
-    spirv.ReturnValue %0 : !spirv.array<3 x vector<2xi32>>
-  }
-
-  // CHECK-LABEL: @splat_array_of_splat_array_of_i32
-  spirv.func @splat_array_of_splat_array_of_i32() -> (!spirv.array<2 x !spirv.array<3 x i32>>) "None" {
-    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate [3 : i32] : !spirv.array<2 x !spirv.array<3 x i32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [3 : i32] : !spirv.array<2 x !spirv.array<3 x i32>>
-    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x i32>>
-  }
-
-  // CHECK-LABEL: @splat_array_of_non_splat_array_of_i32
-  spirv.func @splat_array_of_non_splat_array_of_i32() -> (!spirv.array<2 x !spirv.array<3 x i32>>) "None" {
-    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate {{\[}}[1 : i32, 2 : i32, 3 : i32]] : !spirv.array<2 x !spirv.array<3 x i32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [[1 : i32, 2 : i32, 3 : i32]] : !spirv.array<2 x !spirv.array<3 x i32>>
-    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x i32>>
-  }
-
-  // CHECK-LABEL: @splat_array_of_splat_vectors_of_i32
-  spirv.func @splat_array_of_splat_vectors_of_i32() -> (!spirv.array<2 x vector<2xi32>>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.array<2 x vector<2xi32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.array<2 x vector<2xi32>>
-    spirv.ReturnValue %0 : !spirv.array<2 x vector<2xi32>>
-  }
-
-  // CHECK-LABEL: @splat_arm_tensor_of_i32
-  spirv.func @splat_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.arm.tensor<2x3xi32>
-    %0 = spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.arm.tensor<2x3xi32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
-  }
-
-  // CHECK-LABEL: @splat_array_of_non_splat_array_of_arrays_of_i32
-  spirv.func @splat_array_of_non_splat_array_of_arrays_of_i32() -> !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>> "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate {{\[}}{{\[}}[1 : i32, 2 : i32, 3 : i32], [4 : i32, 5 : i32, 6 : i32]]] : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [[[1 : i32, 2 : i32, 3 : i32], [4 : i32, 5 : i32, 6 : i32]]] : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>>
-    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>>
-  }
-
-  // CHECK-LABEL: @null_cc_arm_tensor_of_i32
-  spirv.func @null_cc_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
-    // CHECK: spirv.Constant dense<0> : !spirv.arm.tensor<2x3xi32>
-    %0 = spirv.EXT.ConstantCompositeReplicate [0 : i32] : !spirv.arm.tensor<2x3xi32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
-  }
-
-  // CHECK-LABEL: @splat_vector_f32
-  spirv.func @splat_vector_f32() -> (vector<3xf32>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [1.000000e+00 : f32] : vector<3xf32>
-    %1 = spirv.EXT.ConstantCompositeReplicate [1.0 : f32] : vector<3xf32>
-    spirv.ReturnValue %1 : vector<3xf32>
-  }
-
-  // CHECK-LABEL: @splat_array_of_f32
-  spirv.func @splat_array_of_f32() -> (!spirv.array<3 x f32>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [1.000000e+00 : f32] : !spirv.array<3 x f32>
-    %1 = spirv.EXT.ConstantCompositeReplicate [1.0 : f32] : !spirv.array<3 x f32>
-    spirv.ReturnValue %1 : !spirv.array<3 x f32>
-  }
- 
-  // CHECK-LABEL: @splat_array_of_splat_array_of_f32
-  spirv.func @splat_array_of_splat_array_of_f32() -> (!spirv.array<2 x !spirv.array<3 x f32>>) "None" {
-    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate [3.000000e+00 : f32] : !spirv.array<2 x !spirv.array<3 x f32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [3.0 : f32] : !spirv.array<2 x !spirv.array<3 x f32>>
-    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x f32>>
-  }
-
-  // CHECK-LABEL: @splat_array_of_non_splat_array_of_f32
-  spirv.func @splat_array_of_non_splat_array_of_f32() -> (!spirv.array<2 x !spirv.array<3 x f32>>) "None" {
-    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate {{\[}}[1.000000e+00 : f32, 2.000000e+00 : f32, 3.000000e+00 : f32]] : !spirv.array<2 x !spirv.array<3 x f32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [[1.0 : f32, 2.0 : f32, 3.0 : f32]] : !spirv.array<2 x !spirv.array<3 x f32>>
-    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x f32>>
-  }
-
-  // CHECK-LABEL: @splat_array_of_vectors_of_f32
-  spirv.func @splat_array_of_vectors_of_f32() -> (!spirv.array<3 x vector<2xf32>>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [dense<[1.000000e+00, 2.000000e+00]> : vector<2xf32>] : !spirv.array<3 x vector<2xf32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [dense<[1.0, 2.0]> : vector<2xf32>] : !spirv.array<3 x vector<2xf32>>
-    spirv.ReturnValue %0 : !spirv.array<3 x vector<2xf32>>
-  }
-
-  // CHECK-LABEL: @splat_array_of_splat_vectors_of_f32
-  spirv.func @splat_array_of_splat_vectors_of_f32() -> (!spirv.array<2 x vector<2xf32>>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [2.000000e+00 : f32] : !spirv.array<2 x vector<2xf32>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [2.0 : f32] : !spirv.array<2 x vector<2xf32>>
-    spirv.ReturnValue %0 : !spirv.array<2 x vector<2xf32>>
-  }
-
-  // CHECK-LABEL: @splat_arm_tensor_of_f32
-  spirv.func @splat_arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate [2.000000e+00 : f32] : !spirv.arm.tensor<2x3xf32>
-    %0 = spirv.EXT.ConstantCompositeReplicate [2.0 : f32] : !spirv.arm.tensor<2x3xf32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
-  }
-
-  // CHECK-LABEL: @splat_array_of_non_splat_array_of_arrays_of_f32
-  spirv.func @splat_array_of_non_splat_array_of_arrays_of_f32() -> !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x f32>>> "None" {
-    // CHECK: spirv.EXT.ConstantCompositeReplicate {{\[}}{{\[}}[1.000000e+00 : f32, 2.000000e+00 : f32, 3.000000e+00 : f32], [4.000000e+00 : f32, 5.000000e+00 : f32, 6.000000e+00 : f32]]] : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x f32>>>
-    %0 = spirv.EXT.ConstantCompositeReplicate [[[1.0 : f32, 2.0 : f32, 3.0 : f32], [4.0 : f32, 5.0 : f32, 6.0 : f32]]] : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x f32>>>
-    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x f32>>>
-  }
-
-  // CHECK-LABEL: @null_cc_arm_tensor_of_f32
-  spirv.func @null_cc_arm_tensor_of_f32() -> (!spirv.arm.tensor<2x3xf32>) "None" {
-    // CHECK: spirv.Constant dense<0.000000e+00> : !spirv.arm.tensor<2x3xf32>
-    %0 = spirv.EXT.ConstantCompositeReplicate [0.0 : f32] : !spirv.arm.tensor<2x3xf32>
-    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xf32>
-  }
-}
diff --git a/mlir/test/Target/SPIRV/replicated-composites-constant.mlir b/mlir/test/Target/SPIRV/replicated-composites-constant.mlir
new file mode 100644
index 0000000000000..02a252f346fae
--- /dev/null
+++ b/mlir/test/Target/SPIRV/replicated-composites-constant.mlir
@@ -0,0 +1,135 @@
+
+// RUN: mlir-translate --no-implicit-module --test-spirv-roundtrip %s | FileCheck %s
+// RUN: %if spirv-tools %{ mlir-translate --no-implicit-module --serialize-spirv %s | spirv-val %}
+
+// Note: Since the output of this test (optionally) gets validated by spirv-val,
+// we cannot use splits.
+
+spirv.module Logical GLSL450 requires #spirv.vce<v1.0, [Shader, ReplicatedCompositesEXT, TensorsARM, Linkage],
+                                                       [SPV_EXT_replicated_composites, SPV_ARM_tensors]> {
+  // CHECK-LABEL: @splat_vector_i32
+  spirv.func @splat_vector_i32() -> (vector<3xi32>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [1 : i32] : vector<3xi32>
+    %1 = spirv.EXT.ConstantCompositeReplicate [1 : i32] : vector<3xi32>
+    spirv.ReturnValue %1 : vector<3xi32>
+  }
+
+  // CHECK-LABEL: @splat_array_of_i32
+  spirv.func @splat_array_of_i32() -> (!spirv.array<3 x i32>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [1 : i32] : !spirv.array<3 x i32>
+    %1 = spirv.EXT.ConstantCompositeReplicate [1 : i32] : !spirv.array<3 x i32>
+    spirv.ReturnValue %1 : !spirv.array<3 x i32>
+  }
+
+  // CHECK-LABEL: @splat_array_of_vectors_of_i32
+  spirv.func @splat_array_of_vectors_of_i32() -> (!spirv.array<3 x vector<2xi32>>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [dense<[1, 2]> : vector<2xi32>] : !spirv.array<3 x vector<2xi32>>
+    %0 = spirv.EXT.ConstantCompositeReplicate [dense<[1, 2]> : vector<2xi32>] : !spirv.array<3 x vector<2xi32>>
+    spirv.ReturnValue %0 : !spirv.array<3 x vector<2xi32>>
+  }
+
+  // CHECK-LABEL: @splat_array_of_splat_array_of_i32
+  spirv.func @splat_array_of_splat_array_of_i32() -> (!spirv.array<2 x !spirv.array<3 x i32>>) "None" {
+    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate [3 : i32] : !spirv.array<2 x !spirv.array<3 x i32>>
+    %0 = spirv.EXT.ConstantCompositeReplicate [3 : i32] : !spirv.array<2 x !spirv.array<3 x i32>>
+    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x i32>>
+  }
+
+  // CHECK-LABEL: @splat_array_of_non_splat_array_of_i32
+  spirv.func @splat_array_of_non_splat_array_of_i32() -> (!spirv.array<2 x !spirv.array<3 x i32>>) "None" {
+    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate {{\[}}[1 : i32, 2 : i32, 3 : i32]] : !spirv.array<2 x !spirv.array<3 x i32>>
+    %0 = spirv.EXT.ConstantCompositeReplicate [[1 : i32, 2 : i32, 3 : i32]] : !spirv.array<2 x !spirv.array<3 x i32>>
+    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x i32>>
+  }
+
+  // CHECK-LABEL: @splat_array_of_splat_vectors_of_i32
+  spirv.func @splat_array_of_splat_vectors_of_i32() -> (!spirv.array<2 x vector<2xi32>>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.array<2 x vector<2xi32>>
+    %0 = spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.array<2 x vector<2xi32>>
+    spirv.ReturnValue %0 : !spirv.array<2 x vector<2xi32>>
+  }
+
+  // CHECK-LABEL: @splat_arm_tensor_of_i32
+  spirv.func @splat_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.arm.tensor<2x3xi32>
+    %0 = spirv.EXT.ConstantCompositeReplicate [2 : i32] : !spirv.arm.tensor<2x3xi32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
+  }
+
+  // CHECK-LABEL: @splat_array_of_non_splat_array_of_arrays_of_i32
+  spirv.func @splat_array_of_non_splat_array_of_arrays_of_i32() -> !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>> "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate {{\[}}{{\[}}[1 : i32, 2 : i32, 3 : i32], [4 : i32, 5 : i32, 6 : i32]]] : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>>
+    %0 = spirv.EXT.ConstantCompositeReplicate [[[1 : i32, 2 : i32, 3 : i32], [4 : i32, 5 : i32, 6 : i32]]] : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>>
+    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<2 x !spirv.array<3 x i32>>>
+  }
+
+  // CHECK-LABEL: @null_cc_arm_tensor_of_i32
+  spirv.func @null_cc_arm_tensor_of_i32() -> (!spirv.arm.tensor<2x3xi32>) "None" {
+    // CHECK: spirv.Constant dense<0> : !spirv.arm.tensor<2x3xi32>
+    %0 = spirv.EXT.ConstantCompositeReplicate [0 : i32] : !spirv.arm.tensor<2x3xi32>
+    spirv.ReturnValue %0 : !spirv.arm.tensor<2x3xi32>
+  }
+
+  // CHECK-LABEL: @splat_vector_f32
+  spirv.func @splat_vector_f32() -> (vector<3xf32>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [1.000000e+00 : f32] : vector<3xf32>
+    %1 = spirv.EXT.ConstantCompositeReplicate [1.0 : f32] : vector<3xf32>
+    spirv.ReturnValue %1 : vector<3xf32>
+  }
+
+  // CHECK-LABEL: @splat_array_of_f32
+  spirv.func @splat_array_of_f32() -> (!spirv.array<3 x f32>) "None" {
+    // CHECK: spirv.EXT.ConstantCompositeReplicate [1.000000e+00 : f32] : !spirv.array<3 x f32>
+    %1 = spirv.EXT.ConstantCompositeReplicate [1.0 : f32] : !spirv.array<3 x f32>
+    spirv.ReturnValue %1 : !spirv.array<3 x f32>
+  }
+
+  // CHECK-LABEL: @splat_array_of_splat_array_of_f32
+  spirv.func @splat_array_of_splat_array_of_f32() -> (!spirv.array<2 x !spirv.array<3 x f32>>) "None" {
+    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate [3.000000e+00 : f32] : !spirv.array<2 x !spirv.array<3 x f32>>
+    %0 = spirv.EXT.ConstantCompositeReplicate [3.0 : f32] : !spirv.array<2 x !spirv.array<3 x f32>>
+    spirv.ReturnValue %0 : !spirv.array<2 x !spirv.array<3 x f32>>
+  }
+
+  // CHECK-LABEL: @splat_array_of_non_splat_array_of_f32
+  spirv.func @splat_array_of_non_splat_array_of_f32() -> (!spirv.array<2 x !spirv.array<3 x f32>>) "None" {
+    // CHECK: %0 = spirv.EXT.ConstantCompositeReplicate {{\[}}[1.000000e+00 : f32, 2.000000e+00 : f32, 3.000000e+00 : f32]] : !spirv.array<2 x !spirv.array<3 ...
[truncated]

@kuhar
Copy link
Member Author

kuhar commented Aug 4, 2025

cc: @mahabadm @davidegrohmann

@kuhar kuhar merged commit 5514e5e into llvm:main Aug 4, 2025
12 checks passed
@davidegrohmann
Copy link
Contributor

@kuhar Thanks for taking care of this and sorry for the caused issues. I will look into fixing the disabled tests.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@krzysz00 krzysz00 krzysz00 approved these changes

@antiagainst antiagainst Awaiting requested review from antiagainst antiagainst is a code owner

@Hardcode84 Hardcode84 Awaiting requested review from Hardcode84

@IgWod-IMG IgWod-IMG Awaiting requested review from IgWod-IMG

Assignees
No one assigned
Labels
mlir:spirv mlir
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@kuhar @llvmbot @davidegrohmann @krzysz00