[mlir][vector] Replace vector.splat with vector.broadcast in some tests (#152230)

Splat is deprecated, and being prepared for removal in a future release.

https://discourse.llvm.org/t/rfc-mlir-vector-deprecate-then-remove-vector-splat/87143/5

The command I used, catches almost every splat op: 
 ```
perl -i -pe
's/vector\.splat\s+(\S+)\s*:\s*vector<((?:\[?\d+\]?x)*)\s*([^>]+)>/vector.broadcast
$1 : $3 to vector<$2$3>/g' filename
 ```

Author: newling

mlir/test/Dialect/Arith/canonicalize.mlir

@@ -654,7 +654,7 @@ func.func @signExtendConstant() -> i16 {
 //       CHECK:   return %[[cres]]
 func.func @signExtendConstantSplat() -> vector<4xi16> {
   %c-2 = arith.constant -2 : i8
-  %splat = vector.splat %c-2 : vector<4xi8>
+  %splat = vector.broadcast %c-2 : i8 to vector<4xi8>
   %ext = arith.extsi %splat : vector<4xi8> to vector<4xi16>
   return %ext : vector<4xi16>
 }
@@ -682,7 +682,7 @@ func.func @unsignedExtendConstant() -> i16 {
 //       CHECK:   return %[[cres]]
 func.func @unsignedExtendConstantSplat() -> vector<4xi16> {
   %c2 = arith.constant 2 : i8
-  %splat = vector.splat %c2 : vector<4xi8>
+  %splat = vector.broadcast %c2 : i8 to vector<4xi8>
   %ext = arith.extui %splat : vector<4xi8> to vector<4xi16>
   return %ext : vector<4xi16>
 }
@@ -866,7 +866,7 @@ func.func @truncExtsiVector(%arg0: vector<2xi32>) -> vector<2xi16> {
 //       CHECK:   return %[[cres]]
 func.func @truncConstantSplat() -> vector<4xi8> {
   %c-2 = arith.constant -2 : i16
-  %splat = vector.splat %c-2 : vector<4xi16>
+  %splat = vector.broadcast %c-2 : i16 to vector<4xi16>
   %trunc = arith.trunci %splat : vector<4xi16> to vector<4xi8>
   return %trunc : vector<4xi8>
 }
@@ -2334,7 +2334,7 @@ func.func @constant_FPtoUI_splat() -> vector<4xi32> {
   // CHECK: %[[C0:.+]] = arith.constant dense<2> : vector<4xi32>
   // CHECK: return %[[C0]]
   %c0 = arith.constant 2.0 : f32
-  %splat = vector.splat %c0 : vector<4xf32>
+  %splat = vector.broadcast %c0 : f32 to vector<4xf32>
   %res = arith.fptoui %splat : vector<4xf32> to vector<4xi32>
   return %res : vector<4xi32>
 }
@@ -2374,7 +2374,7 @@ func.func @constant_FPtoSI_splat() -> vector<4xi32> {
   // CHECK: %[[C0:.+]] = arith.constant dense<-2> : vector<4xi32>
   // CHECK: return %[[C0]]
   %c0 = arith.constant -2.0 : f32
-  %splat = vector.splat %c0 : vector<4xf32>
+  %splat = vector.broadcast %c0 : f32 to vector<4xf32>
   %res = arith.fptosi %splat : vector<4xf32> to vector<4xi32>
   return %res : vector<4xi32>
 }
@@ -2413,7 +2413,7 @@ func.func @constant_SItoFP_splat() -> vector<4xf32> {
   // CHECK: %[[C0:.+]] = arith.constant dense<2.000000e+00> : vector<4xf32>
   // CHECK: return %[[C0]]
   %c0 = arith.constant 2 : i32
-  %splat = vector.splat %c0 : vector<4xi32>
+  %splat = vector.broadcast %c0 : i32 to vector<4xi32>
   %res = arith.sitofp %splat : vector<4xi32> to vector<4xf32>
   return %res : vector<4xf32>
 }
@@ -2442,7 +2442,7 @@ func.func @constant_UItoFP_splat() -> vector<4xf32> {
   // CHECK: %[[C0:.+]] = arith.constant dense<2.000000e+00> : vector<4xf32>
   // CHECK: return %[[C0]]
   %c0 = arith.constant 2 : i32
-  %splat = vector.splat %c0 : vector<4xi32>
+  %splat = vector.broadcast %c0 : i32 to vector<4xi32>
   %res = arith.uitofp %splat : vector<4xi32> to vector<4xf32>
   return %res : vector<4xf32>
 }

mlir/test/Dialect/Vector/vector-sink.mlir

@@ -40,7 +40,7 @@ func.func @broadcast_scalar_with_bcast_scalable(%arg1: index, %arg2: index) -> v
 // CHECK:           %[[BCAST:.*]] = vector.broadcast %[[ADD]] : index to vector<1x4xindex>
 // CHECK:           return %[[BCAST]] : vector<1x4xindex>
 func.func @broadcast_scalar_with_bcast_and_splat(%arg1: index, %arg2: index) -> vector<1x4xindex> {
-  %0 = vector.splat %arg1 : vector<1x4xindex>
+  %0 = vector.broadcast %arg1 : index to vector<1x4xindex>
   %1 = vector.broadcast %arg2 : index to vector<1x4xindex>
   %2 = arith.addi %0, %1 : vector<1x4xindex>
   return %2 : vector<1x4xindex>
@@ -53,7 +53,7 @@ func.func @broadcast_scalar_with_bcast_and_splat(%arg1: index, %arg2: index) ->
 // CHECK:           %[[BCAST:.*]] = vector.broadcast %[[ADD]] : index to vector<1x[4]xindex>
 // CHECK:           return %[[BCAST]] : vector<1x[4]xindex>
 func.func @broadcast_scalar_with_bcast_and_splat_scalable(%arg1: index, %arg2: index) -> vector<1x[4]xindex> {
-  %0 = vector.splat %arg1 : vector<1x[4]xindex>
+  %0 = vector.broadcast %arg1 : index to vector<1x[4]xindex>
   %1 = vector.broadcast %arg2 : index to vector<1x[4]xindex>
   %2 = arith.addi %0, %1 : vector<1x[4]xindex>
   return %2 : vector<1x[4]xindex>
@@ -94,12 +94,12 @@ func.func @broadcast_vector_scalable(%arg1: vector<[4]xf32>, %arg2: vector<[4]xf
 // CHECK-LABEL:   func.func @broadcast_scalar_and_vec(
 // CHECK-SAME:       %[[ARG1:.*]]: index,
 // CHECK-SAME:       %[[ARG2:.*]]: vector<4xindex>) -> vector<1x4xindex> {
-// CHECK:            %[[SPLAT:.*]] = vector.splat %[[ARG1]] : vector<1x4xindex>
+// CHECK:            %[[SPLAT:.*]] = vector.broadcast %[[ARG1]] : index to vector<1x4xindex>
 // CHECK:            %[[BCAST:.*]] = vector.broadcast %[[ARG2]] : vector<4xindex> to vector<1x4xindex>
 // CHECK:            %[[ADD:.*]] = arith.addi %[[SPLAT]], %[[BCAST]] : vector<1x4xindex>
 // CHECK:            return %[[ADD]] : vector<1x4xindex>
 func.func @broadcast_scalar_and_vec(%arg1: index, %arg2: vector<4xindex>) -> vector<1x4xindex> {
-  %0 = vector.splat %arg1 : vector<1x4xindex>
+  %0 = vector.broadcast %arg1 : index to vector<1x4xindex>
   %1 = vector.broadcast %arg2 : vector<4xindex> to vector<1x4xindex>
   %2 = arith.addi %0, %1 : vector<1x4xindex>
   return %2 : vector<1x4xindex>
@@ -108,12 +108,12 @@ func.func @broadcast_scalar_and_vec(%arg1: index, %arg2: vector<4xindex>) -> vec
 // CHECK-LABEL:   func.func @broadcast_scalar_and_vec_scalable(
 // CHECK-SAME:       %[[ARG1:.*]]: index,
 // CHECK-SAME:       %[[ARG2:.*]]: vector<[4]xindex>) -> vector<1x[4]xindex> {
-// CHECK:            %[[SPLAT:.*]] = vector.splat %[[ARG1]] : vector<1x[4]xindex>
+// CHECK:            %[[SPLAT:.*]] = vector.broadcast %[[ARG1]] : index to vector<1x[4]xindex>
 // CHECK:            %[[BCAST:.*]] = vector.broadcast %[[ARG2]] : vector<[4]xindex> to vector<1x[4]xindex>
 // CHECK:            %[[ADD:.*]] = arith.addi %[[SPLAT]], %[[BCAST]] : vector<1x[4]xindex>
 // CHECK:            return %[[ADD]] : vector<1x[4]xindex>
 func.func @broadcast_scalar_and_vec_scalable(%arg1: index, %arg2: vector<[4]xindex>) -> vector<1x[4]xindex> {
-  %0 = vector.splat %arg1 : vector<1x[4]xindex>
+  %0 = vector.broadcast %arg1 : index to vector<1x[4]xindex>
   %1 = vector.broadcast %arg2 : vector<[4]xindex> to vector<1x[4]xindex>
   %2 = arith.addi %0, %1 : vector<1x[4]xindex>
   return %2 : vector<1x[4]xindex>
@@ -787,7 +787,7 @@ func.func @negative_extract_load_scalable(%arg0: memref<?xf32>, %arg1: index) ->
 //  CHECK-SAME:   (%[[ARG0:.*]]: memref<?xf32>, %[[ARG1:.*]]: index, %[[ARG2:.*]]: f32)
 func.func @store_splat(%arg0: memref<?xf32>, %arg1: index, %arg2: f32) {
 // CHECK:   memref.store %[[ARG2]], %[[ARG0]][%[[ARG1]]] : memref<?xf32>
-  %0 = vector.splat %arg2 : vector<1xf32>
+  %0 = vector.broadcast %arg2 : f32 to vector<1xf32>
   vector.store %0, %arg0[%arg1] : memref<?xf32>, vector<1xf32>
   return
 }
@@ -813,40 +813,40 @@ func.func @store_broadcast_1d_to_2d(%arg0: memref<?x?xf32>, %arg1: index, %arg2:
 // CHECK-LABEL: @negative_store_scalable
 //  CHECK-SAME:   (%[[ARG0:.*]]: memref<?xf32>, %[[ARG1:.*]]: index, %[[ARG2:.*]]: f32)
 func.func @negative_store_scalable(%arg0: memref<?xf32>, %arg1: index, %arg2: f32) {
-// CHECK:   %[[RES:.*]] = vector.splat %[[ARG2]] : vector<[1]xf32>
+// CHECK:   %[[RES:.*]] = vector.broadcast %[[ARG2]] : f32 to vector<[1]xf32>
 // CHECK:   vector.store %[[RES]], %[[ARG0]][%[[ARG1]]] : memref<?xf32>, vector<[1]xf32>
-  %0 = vector.splat %arg2 : vector<[1]xf32>
+  %0 = vector.broadcast %arg2 : f32 to vector<[1]xf32>
   vector.store %0, %arg0[%arg1] : memref<?xf32>, vector<[1]xf32>
   return
 }
 
 // CHECK-LABEL: @negative_store_memref_of_vec
 //  CHECK-SAME:   (%[[ARG0:.*]]: memref<?xvector<1xf32>>, %[[ARG1:.*]]: index, %[[ARG2:.*]]: f32)
 func.func @negative_store_memref_of_vec(%arg0: memref<?xvector<1xf32>>, %arg1: index, %arg2: f32) {
-// CHECK:   %[[RES:.*]] = vector.splat %[[ARG2]] : vector<1xf32>
+// CHECK:   %[[RES:.*]] = vector.broadcast %[[ARG2]] : f32 to vector<1xf32>
 // CHECK:   vector.store %[[RES]], %[[ARG0]][%[[ARG1]]] : memref<?xvector<1xf32>>, vector<1xf32>
-  %0 = vector.splat %arg2 : vector<1xf32>
+  %0 = vector.broadcast %arg2 : f32 to vector<1xf32>
   vector.store %0, %arg0[%arg1] : memref<?xvector<1xf32>>, vector<1xf32>
   return
 }
 
 // CHECK-LABEL: @negative_store_more_than_one_element
 //  CHECK-SAME:   (%[[ARG0:.*]]: memref<?xf32>, %[[ARG1:.*]]: index, %[[ARG2:.*]]: f32)
 func.func @negative_store_more_than_one_element(%arg0: memref<?xf32>, %arg1: index, %arg2: f32) {
-// CHECK:   %[[RES:.*]] = vector.splat %[[ARG2]] : vector<4xf32>
+// CHECK:   %[[RES:.*]] = vector.broadcast %[[ARG2]] : f32 to vector<4xf32>
 // CHECK:   vector.store %[[RES]], %[[ARG0]][%[[ARG1]]] : memref<?xf32>, vector<4xf32>
-  %0 = vector.splat %arg2 : vector<4xf32>
+  %0 = vector.broadcast %arg2 : f32 to vector<4xf32>
   vector.store %0, %arg0[%arg1] : memref<?xf32>, vector<4xf32>
   return
 }
 
 // CHECK-LABEL: @negative_store_no_single_use
 //  CHECK-SAME:   (%[[ARG0:.*]]: memref<?xf32>, %[[ARG1:.*]]: index, %[[ARG2:.*]]: f32)
 func.func @negative_store_no_single_use(%arg0: memref<?xf32>, %arg1: index, %arg2: f32) -> vector<1xf32> {
-// CHECK:   %[[RES:.*]] = vector.splat %[[ARG2]] : vector<1xf32>
+// CHECK:   %[[RES:.*]] = vector.broadcast %[[ARG2]] : f32 to vector<1xf32>
 // CHECK:   vector.store %[[RES]], %[[ARG0]][%[[ARG1]]] : memref<?xf32>, vector<1xf32>
 // CHECK:   return %[[RES:.*]] : vector<1xf32>
-  %0 = vector.splat %arg2 : vector<1xf32>
+  %0 = vector.broadcast %arg2 : f32 to vector<1xf32>
   vector.store %0, %arg0[%arg1] : memref<?xf32>, vector<1xf32>
   return %0 : vector<1xf32>
 }

mlir/test/Dialect/Vector/vector-transfer-to-vector-load-store.mlir

@@ -285,19 +285,19 @@ func.func @transfer_read_permutations(%mem_0 : memref<?x?xf32>, %mem_1 : memref<
   %c0 = arith.constant 0 : index
 
 // CHECK: %[[MASK0:.*]] = vector.broadcast %{{.*}} : i1 to vector<14x7xi1>
-  %mask0 = vector.splat %m : vector<14x7xi1>
+  %mask0 = vector.broadcast %m : i1 to vector<14x7xi1>
   %0 = vector.transfer_read %mem_1[%c0, %c0, %c0, %c0], %cst, %mask0 {in_bounds = [true, false, true, true], permutation_map = #map0} : memref<?x?x?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: vector.transfer_read {{.*}} %[[MASK0]] {in_bounds = [false, true, true, true], permutation_map = #[[$MAP0]]} : memref<?x?x?x?xf32>, vector<14x7x8x16xf32>
 // CHECK: vector.transpose %{{.*}}, [1, 0, 2, 3] : vector<14x7x8x16xf32> to vector<7x14x8x16xf32>
 
 // CHECK: %[[MASK1:.*]] = vector.broadcast %{{.*}} : i1 to vector<16x14xi1>
-  %mask1 = vector.splat %m : vector<16x14xi1>
+  %mask1 = vector.broadcast %m : i1 to vector<16x14xi1>
   %1 = vector.transfer_read %mem_1[%c0, %c0, %c0, %c0], %cst, %mask1 {in_bounds = [true, false, true, false], permutation_map = #map1} : memref<?x?x?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: vector.transfer_read {{.*}} %[[MASK1]] {in_bounds = [false, false, true, true], permutation_map = #[[$MAP0]]} : memref<?x?x?x?xf32>, vector<16x14x7x8xf32>
 // CHECK: vector.transpose %{{.*}}, [2, 1, 3, 0] : vector<16x14x7x8xf32> to vector<7x14x8x16xf32>
 
 // CHECK: %[[MASK3:.*]] = vector.broadcast %{{.*}} : i1 to vector<14x7xi1>
-  %mask2 = vector.splat %m : vector<14x7xi1>
+  %mask2 = vector.broadcast %m : i1 to vector<14x7xi1>
   %2 = vector.transfer_read %mem_1[%c0, %c0, %c0, %c0], %cst, %mask2 {in_bounds = [true, false, true, true], permutation_map = #map2} : memref<?x?x?x?xf32>, vector<7x14x8x16xf32>
 // CHECK: vector.transfer_read {{.*}} %[[MASK3]] {in_bounds = [false, true, true], permutation_map = #[[$MAP1]]} : memref<?x?x?x?xf32>, vector<14x16x7xf32>
 // CHECK: vector.broadcast %{{.*}} : vector<14x16x7xf32> to vector<8x14x16x7xf32>
@@ -337,7 +337,7 @@ func.func @transfer_write_permutations_tensor_masked(
   %c0 = arith.constant 0 : index
 
   // CHECK: %[[MASK:.*]] = vector.broadcast %[[M]] : i1 to vector<16x14x7x8xi1>
-  %mask0 = vector.splat %m : vector<16x14x7x8xi1>
+  %mask0 = vector.broadcast %m : i1 to vector<16x14x7x8xi1>
   %res = vector.transfer_write %vec, %dst[%c0, %c0, %c0, %c0], %mask0 {in_bounds = [true, false, false, true], permutation_map = affine_map<(d0, d1, d2, d3) -> (d2, d1, d3, d0)>} : vector<7x14x8x16xf32>, tensor<?x?x?x?xf32>
   // CHECK: %[[NEW_VEC0:.*]] = vector.transpose %{{.*}} [3, 1, 0, 2] : vector<7x14x8x16xf32> to vector<16x14x7x8xf32>
   // CHECK: %[[NEW_RES0:.*]] = vector.transfer_write %[[NEW_VEC0]], %[[DST]][%c0, %c0, %c0, %c0], %[[MASK]] {in_bounds = [true, false, true, false]} : vector<16x14x7x8xf32>, tensor<?x?x?x?xf32>

mlir/test/Integration/Dialect/Vector/CPU/ArmSME/outerproduct-f32.mlir

@@ -46,7 +46,7 @@ func.func @test_outerproduct_with_accumulator_4x4xf32() {
   %c0 = arith.constant 0 : index
   %f10 = arith.constant 10.0 : f32
 
-  %acc = vector.splat %f10 : vector<[4]x[4]xf32>
+  %acc = vector.broadcast %f10 : f32 to vector<[4]x[4]xf32>
   %vector_i32 = llvm.intr.stepvector : vector<[4]xi32>
   %vector = arith.sitofp %vector_i32 : vector<[4]xi32> to vector<[4]xf32>
   %tile = vector.outerproduct %vector, %vector, %acc : vector<[4]xf32>, vector<[4]xf32>
@@ -103,7 +103,7 @@ func.func @test_masked_outerproduct_with_accumulator_4x4xf32() {
   %ones = arith.constant dense<1> : vector<[4]xi32>
   %f10 = arith.constant 10.0 : f32
 
-  %acc = vector.splat %f10 : vector<[4]x[4]xf32>
+  %acc = vector.broadcast %f10 : f32 to vector<[4]x[4]xf32>
   %step_vector = llvm.intr.stepvector : vector<[4]xi32>
   %vector_i32 = arith.addi %step_vector, %ones : vector<[4]xi32>
   %vector = arith.sitofp %vector_i32 : vector<[4]xi32> to vector<[4]xf32>

mlir/test/Integration/Dialect/Vector/CPU/ArmSME/outerproduct-f64.mlir

@@ -52,7 +52,7 @@ func.func @test_outerproduct_with_accumulator_2x2xf64() {
   %ones = arith.constant dense<1> : vector<[2]xi32>
   %f10 = arith.constant 10.0 : f64
 
-  %acc = vector.splat %f10 : vector<[2]x[2]xf64>
+  %acc = vector.broadcast %f10 : f64 to vector<[2]x[2]xf64>
   %step_vector = llvm.intr.stepvector : vector<[2]xi32>
   %vector_i32 = arith.addi %step_vector, %ones : vector<[2]xi32>
   %vector = arith.sitofp %vector_i32 : vector<[2]xi32> to vector<[2]xf64>
@@ -108,7 +108,7 @@ func.func @test_masked_outerproduct_with_accumulator_2x2xf64() {
   %ones = arith.constant dense<1> : vector<[2]xi32>
   %f10 = arith.constant 10.0 : f64
 
-  %acc = vector.splat %f10 : vector<[2]x[2]xf64>
+  %acc = vector.broadcast %f10 : f64 to vector<[2]x[2]xf64>
   %step_vector = llvm.intr.stepvector : vector<[2]xi32>
   %vector_i32 = arith.addi %step_vector, %ones : vector<[2]xi32>
   %vector = arith.sitofp %vector_i32 : vector<[2]xi32> to vector<[2]xf64>

mlir/test/Integration/Dialect/Vector/CPU/ArmSME/transfer-write-2d.mlir

@@ -10,7 +10,7 @@
 // Vector store.
 func.func @transfer_write_2d(%A : memref<?x?xf32>, %base1: index, %base2: index) {
   %c0 = arith.constant 0.0 : f32
-  %zero = vector.splat %c0 : vector<[4]x[4]xf32>
+  %zero = vector.broadcast %c0 : f32 to vector<[4]x[4]xf32>
   vector.transfer_write %zero, %A[%base1, %base2] {in_bounds=[true, true]} :
     vector<[4]x[4]xf32>, memref<?x?xf32>
   return
@@ -22,7 +22,7 @@ func.func @transfer_write_2d_mask(%A : memref<?x?xf32>, %base1: index, %base2: i
   %c2 = arith.constant 2 : index
   %c3 = arith.constant 3 : index
   %mask = vector.create_mask %c2, %c3 : vector<[4]x[4]xi1>
-  %zero = vector.splat %c0 : vector<[4]x[4]xf32>
+  %zero = vector.broadcast %c0 : f32 to vector<[4]x[4]xf32>
   vector.transfer_write %zero, %A[%base1, %base2], %mask {in_bounds=[true, true]} :
     vector<[4]x[4]xf32>, memref<?x?xf32>
   return

mlir/test/Integration/Dialect/Vector/CPU/ArmSVE/contraction.mlir

@@ -106,7 +106,7 @@ func.func @matvec_i32() {
   //    val = (123 * 314) * 4 * vscale
   // so ...
   %vscale = vector.vscale
-  %vscale_v = vector.splat %vscale : vector<3xindex>
+  %vscale_v = vector.broadcast %vscale : index to vector<3xindex>
   %vscale_i32 = arith.index_cast %vscale_v : vector<3xindex> to vector<3xi32>
   %mv1_div = arith.divui %mv1, %vscale_i32 : vector<3xi32>
   // ... val / vscale = 123 * 314 * 4 = 154488

mlir/test/Integration/Dialect/Vector/CPU/ArmSVE/scalable-interleave.mlir

@@ -7,8 +7,8 @@
 func.func @entry() {
   %f1 = arith.constant 1.0 : f32
   %f2 = arith.constant 2.0 : f32
-  %v1 = vector.splat %f1 : vector<[4]xf32>
-  %v2 = vector.splat %f2 : vector<[4]xf32>
+  %v1 = vector.broadcast %f1 : f32 to vector<[4]xf32>
+  %v2 = vector.broadcast %f2 : f32 to vector<[4]xf32>
   vector.print %v1 : vector<[4]xf32>
   vector.print %v2 : vector<[4]xf32>
   //

mlir/test/Integration/Dialect/Vector/CPU/interleave.mlir

@@ -6,8 +6,8 @@
 func.func @entry() {
   %f1 = arith.constant 1.0 : f32
   %f2 = arith.constant 2.0 : f32
-  %v1 = vector.splat %f1 : vector<2x4xf32>
-  %v2 = vector.splat %f2 : vector<2x4xf32>
+  %v1 = vector.broadcast %f1 : f32 to vector<2x4xf32>
+  %v2 = vector.broadcast %f2 : f32 to vector<2x4xf32>
   vector.print %v1 : vector<2x4xf32>
   vector.print %v2 : vector<2x4xf32>
   //

mlir/test/Integration/Dialect/Vector/CPU/outerproduct-f32.mlir

@@ -14,9 +14,9 @@
 !vector_type_R = vector<7xf32>
 
 func.func @vector_outerproduct_splat_8x8(%fa: f32, %fb: f32, %fc: f32) -> !vector_type_C {
-  %a = vector.splat %fa: !vector_type_A
-  %b = vector.splat %fb: !vector_type_B
-  %c = vector.splat %fc: !vector_type_C
+  %a = vector.broadcast %fa: f32 to !vector_type_A
+  %b = vector.broadcast %fb: f32 to !vector_type_B
+  %c = vector.broadcast %fc: f32 to !vector_type_C
   %d = vector.outerproduct %a, %b, %c : !vector_type_A, !vector_type_B
   return %d: !vector_type_C
 }