[Mlir-commits] [mlir] [mlir][tensor] Relax input type requirement on `tensor.splat` (PR #145893)

[Markus Böck]

https://github.com/zero9178 created https://github.com/llvm/llvm-project/pull/145893

`tensor.splat` is currently restricted to only accepting input values that are of integer, index or float type.

This is much more restrictive than the tensor type itself as well as any lowerings of it.

This PR therefore removes this restriction by using `AnyType` for the input value. Whether the type is actually valid or not for a tensor remains verified through the type equality of the result tensor element type and the input type.

>From 5a1cd6d4260fb7741ad08d8137adceb833d8df1e Mon Sep 17 00:00:00 2001
From: mboeck <mboeck at nvidia.com>
Date: Thu, 26 Jun 2025 15:24:11 +0200
Subject: [PATCH] [mlir][tensor] Relax input type requirement on `tensor.splat`

`tensor.splat` is currently restricted to only accepting input values that are of integer, index or float type.

This is much more restrictive than the tensor type itself as well as any lowerings of it.

This PR therefore removes this restriction by using `AnyType` for the input value. Whether the type is actually valid or not for a tensor remains verified through the type equality of the result tensor element type and the input type.
---
 mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td |  6 ++----
 mlir/test/Dialect/Tensor/bufferize.mlir          | 15 +++++++++++++++
 mlir/test/Dialect/Tensor/invalid.mlir            |  7 ++++---
 mlir/test/Dialect/Tensor/ops.mlir                |  6 +++++-
 4 files changed, 26 insertions(+), 8 deletions(-)

diff --git a/mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td b/mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td
index 47962f75558ea..7d396e5c64c28 100644
--- a/mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td
+++ b/mlir/include/mlir/Dialect/Tensor/IR/TensorOps.td
@@ -1771,8 +1771,7 @@ def Tensor_SplatOp : Tensor_Op<"splat", [
   ]> {
   let summary = "tensor splat or broadcast operation";
   let description = [{
-    Broadcast the operand to all elements of the result tensor. The operand is
-    required to be of integer/index/float type.
+    Broadcast the operand to all elements of the result tensor.
 
     An additional argument of type `index` must be provided for each dynamic
     dimension present in the result type.
@@ -1795,8 +1794,7 @@ def Tensor_SplatOp : Tensor_Op<"splat", [
     ```
   }];
 
-  let arguments = (ins AnyTypeOf<[AnySignlessInteger, Index, AnyFloat],
-                                 "integer/index/float type">:$input,
+  let arguments = (ins AnyType:$input,
                        Variadic<Index>:$dynamicSizes);
   let results = (outs AnyRankedTensor:$aggregate);
 
diff --git a/mlir/test/Dialect/Tensor/bufferize.mlir b/mlir/test/Dialect/Tensor/bufferize.mlir
index c0adc8a49bf70..e202a6b3f3e7a 100644
--- a/mlir/test/Dialect/Tensor/bufferize.mlir
+++ b/mlir/test/Dialect/Tensor/bufferize.mlir
@@ -615,6 +615,21 @@ func.func @tensor.splat(%f: f32) -> tensor<10x2x4xf32> {
 
 // -----
 
+// CHECK-LABEL:   func @tensor.splat_other(
+// CHECK-SAME:        %[[F:.*]]: !llvm.ptr)
+// CHECK-DAG:       %[[ALLOC:.*]] = memref.alloc() {{.*}} : memref<10x2x4x!llvm.ptr>
+// CHECK:           %[[ALLOC_T:.*]] = bufferization.to_tensor %[[ALLOC]]
+// CHECK:           %[[MAPPED:.*]] = linalg.map
+// CHECK:                 outs(%[[ALLOC_T]] : tensor<10x2x4x!llvm.ptr>)
+// CHECK:             linalg.yield %[[F]]
+// CHECK:           return %[[MAPPED]] : tensor<10x2x4x!llvm.ptr>
+func.func @tensor.splat_other(%f: !llvm.ptr) -> tensor<10x2x4x!llvm.ptr> {
+  %t = tensor.splat %f : tensor<10x2x4x!llvm.ptr>
+  return %t : tensor<10x2x4x!llvm.ptr>
+}
+
+// -----
+
 // CHECK-LABEL:   func @tensor.concat(
 // CHECK-SAME:        %[[F:.*]]: tensor<8xf32>)
 // CHECK:           %[[F_MEMREF:.*]] = bufferization.to_buffer %[[F]]
diff --git a/mlir/test/Dialect/Tensor/invalid.mlir b/mlir/test/Dialect/Tensor/invalid.mlir
index f35d52e700084..665657a67dc61 100644
--- a/mlir/test/Dialect/Tensor/invalid.mlir
+++ b/mlir/test/Dialect/Tensor/invalid.mlir
@@ -466,9 +466,10 @@ func.func @invalid_splat(%v : f32) {
 
 // -----
 
-func.func @invalid_splat(%v : vector<8xf32>) {
-  // expected-error at +1 {{must be integer/index/float type}}
-  %w = tensor.splat %v : tensor<8xvector<8xf32>>
+// expected-note at +1 {{prior use here}}
+func.func @invalid_splat(%v : f32) {
+  // expected-error at +1 {{expects different type than prior uses: 'i32' vs 'f32'}}
+  %w = tensor.splat %v : tensor<1xi32>
   return
 }
 
diff --git a/mlir/test/Dialect/Tensor/ops.mlir b/mlir/test/Dialect/Tensor/ops.mlir
index 930986211cb6d..0fd4b87508a79 100644
--- a/mlir/test/Dialect/Tensor/ops.mlir
+++ b/mlir/test/Dialect/Tensor/ops.mlir
@@ -314,12 +314,16 @@ func.func @pad_to_static_size(%arg0: tensor<?x?xf32>, %ub0: index, %ub1: index,
 
 // CHECK-LABEL: func @test_splat_op
 // CHECK-SAME: [[S:%arg[0-9]+]]: f32
-func.func @test_splat_op(%s : f32) {
+// CHECK-SAME: [[P:%arg[0-9]+]]: !llvm.ptr
+func.func @test_splat_op(%s : f32, %p : !llvm.ptr) {
   // CHECK: tensor.splat [[S]] : tensor<8xf32>
   %v = tensor.splat %s : tensor<8xf32>
 
   // CHECK: tensor.splat [[S]] : tensor<4xf32>
   %u = "tensor.splat"(%s) : (f32) -> tensor<4xf32>
+
+  // CHECK: tensor.splat [[P]] : tensor<8x!llvm.ptr>
+  %w = tensor.splat %p : tensor<8x!llvm.ptr>
   return
 }