diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp
index 8c62290b5e0cf..cd777416ba730 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp
@@ -450,26 +450,37 @@ static void updateReduc(Merger &merger, CodeGen &codegen, Value reduc) {
 /// the output buffer is already initialized to all zeroes and only nonzeroes
 /// values are computed and written out. For updates (viz. x(i) += y(i) * z(i)),
 /// only nonzeroes values are used for the updates and no assumption on the
-/// original contents of the output buffer is necessary..
+/// original contents of the output buffer is necessary.
 static Value genOutputBuffer(CodeGen &codegen, OpBuilder &builder,
                              linalg::GenericOp op, MemRefType denseTp,
                              ArrayRef<Value> args) {
   Location loc = op.getLoc();
-  Value tensor = op.getOutputOperand(0)->get();
-  // The output tensor simply could materialize from the buffer that will
-  // be generated for the tensor present in the outs() clause. This has
-  // the major advantage that the sparse kernel only updates the nonzero
-  // positions for the output tensor.
-  if (isInPlace(tensor))
-    return builder.create<bufferization::ToMemrefOp>(loc, denseTp, tensor);
-  // By default, a new buffer is allocated which is initialized to the
-  // tensor defined in the outs() clause. This is always correct but
-  // introduces a dense initialization component that may negatively
-  // impact the running complexity of the sparse kernel. If the tensor
-  // materializes into the computation, we need to preserve the zero
-  // initialization assumption of all sparse output buffers.
+  OpOperand *lhs = op.getOutputOperand(0);
+  Value tensor = lhs->get();
+  bool isInit = op.isInitTensor(lhs);
+  // An output tensor that is in-place can simply materialize from the buffer
+  // of the tensor that appears in the outs() clause. For updates, this has
+  // the advantage that only the nonzero value are involved in the computation,
+  // keeping the operation O(nnz). In all other cases, we are forced to zero
+  // out the buffer to enforce the assumption above, which may negatively
+  // impact running complexity (viz. O(n^2 + nnz) vs. O(nnz) for matrices).
+  // TODO: use better analysis to avoid zeroing out the buffer?
+  if (isInPlace(tensor)) {
+    Value init =
+        builder.create<bufferization::ToMemrefOp>(loc, denseTp, tensor);
+    if (!isInit) {
+      Value zero = constantZero(builder, loc, denseTp.getElementType());
+      builder.create<linalg::FillOp>(loc, ValueRange{zero}, ValueRange{init});
+    }
+    return init;
+  }
+  // By default, a new buffer is allocated which is either set to zero (when
+  // no updates occur or the tensor materializes into this computation) or
+  // initialized to the value of the tensor defined in the outs() clause.
+  // This is always correct (since it enforces all assumptions above) but
+  // may negatively impact running complexity as explained above.
   Value alloc = builder.create<memref::AllocOp>(loc, denseTp, args);
-  if (isMaterializing(tensor)) {
+  if (!isInit || isMaterializing(tensor)) {
     Value zero = constantZero(builder, loc, denseTp.getElementType());
     builder.create<linalg::FillOp>(loc, ValueRange{zero}, ValueRange{alloc});
   } else {
diff --git a/mlir/test/Dialect/SparseTensor/dense.mlir b/mlir/test/Dialect/SparseTensor/dense.mlir
index 247a4b810fa72..dff9b20f16fbc 100644
--- a/mlir/test/Dialect/SparseTensor/dense.mlir
+++ b/mlir/test/Dialect/SparseTensor/dense.mlir
@@ -35,15 +35,15 @@
 // CHECK-LABEL:   func @dense1(
 // CHECK-SAME:                 %[[VAL_0:.*]]: tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>,
 // CHECK-SAME:                 %[[VAL_1:.*]]: tensor<32x16xf32> {linalg.inplaceable = false}) -> tensor<32x16xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_8]], %[[VAL_9]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_9]] : memref<32x16xf32>)
 // CHECK:           scf.for %[[VAL_10:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_6]] {
 // CHECK:             scf.for %[[VAL_11:.*]] = %[[VAL_5]] to %[[VAL_4]] step %[[VAL_6]] {
 // CHECK:               %[[VAL_12:.*]] = arith.muli %[[VAL_10]], %[[VAL_4]] : index
diff --git a/mlir/test/Dialect/SparseTensor/sparse_1d.mlir b/mlir/test/Dialect/SparseTensor/sparse_1d.mlir
index 4dec12c56ec13..6a9d26c65b957 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_1d.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_1d.mlir
@@ -21,9 +21,8 @@
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_7]], %[[VAL_8]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_8]] : memref<32xf32>)
 // CHECK:           scf.for %[[VAL_9:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_10:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_9]]] : memref<?xf32>
 // CHECK:             %[[VAL_11:.*]] = arith.addf %[[VAL_10]], %[[VAL_1]] : f32
@@ -81,9 +80,8 @@ func.func @add_d_init(%arga: tensor<32xf32, #DV>, %argb: f32) -> tensor<32xf32>
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_7]], %[[VAL_8]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_8]] : memref<32xf32>)
 // CHECK:           scf.for %[[VAL_9:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_10:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_9]]] : memref<?xf32>
 // CHECK:             %[[VAL_11:.*]] = arith.mulf %[[VAL_10]], %[[VAL_1]] : f32
@@ -114,9 +112,8 @@ func.func @mul_d(%arga: tensor<32xf32, #DV>, %argb: f32, %argx: tensor<32xf32>)
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_4]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_11]] : memref<32xf32>)
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]]:2 = scf.while (%[[VAL_15:.*]] = %[[VAL_12]], %[[VAL_16:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) {
@@ -167,9 +164,8 @@ func.func @add_s(%arga: tensor<32xf32, #SV>, %argb: f32, %argx: tensor<32xf32>)
 // CHECK-DAG:       %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_7]], %[[VAL_8]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_8]] : memref<32xf32>)
 // CHECK:           %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
 // CHECK:           %[[VAL_10:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_3]] {
@@ -208,9 +204,8 @@ func.func @repeated_add_s(%arga: tensor<32xf32, #SV>, %argx: tensor<32xf32>) ->
 // CHECK-DAG:       %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_8]], %[[VAL_9]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_9]] : memref<32xf32>)
 // CHECK:           %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_10]] to %[[VAL_11]] step %[[VAL_4]] {
@@ -252,9 +247,8 @@ func.func @mul_s(%arga: tensor<32xf32, #SV>, %argb: f32, %argx: tensor<32xf32>)
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf32>
-// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_8]], %[[VAL_9]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_9]] : memref<32xf32>)
 // CHECK:           scf.for %[[VAL_10:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf32>
 // CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_10]]] : memref<32xf32>
@@ -284,9 +278,8 @@ func.func @add_dd(%arga: tensor<32xf32, #DV>, %argb: tensor<32xf32>, %argx: tens
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf32>
-// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_8]], %[[VAL_9]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_9]] : memref<32xf32>)
 // CHECK:           scf.for %[[VAL_10:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf32>
 // CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_10]]] : memref<32xf32>
@@ -319,9 +312,8 @@ func.func @mul_dd(%arga: tensor<32xf32, #DV>, %argb: tensor<32xf32>, %argx: tens
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_12]] : memref<32xf32>)
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]]:2 = scf.while (%[[VAL_16:.*]] = %[[VAL_13]], %[[VAL_17:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) {
@@ -377,9 +369,8 @@ func.func @add_ds(%arga: tensor<32xf32>, %argb: tensor<32xf32, #SV>, %argx: tens
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_9]], %[[VAL_10]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_10]] : memref<32xf32>)
 // CHECK:           %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_13:.*]] = %[[VAL_11]] to %[[VAL_12]] step %[[VAL_4]] {
@@ -415,9 +406,8 @@ func.func @mul_ds(%arga: tensor<32xf32>, %argb: tensor<32xf32, #SV>, %argx: tens
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_12]] : memref<32xf32>)
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]]:2 = scf.while (%[[VAL_16:.*]] = %[[VAL_13]], %[[VAL_17:.*]] = %[[VAL_4]]) : (index, index) -> (index, index) {
@@ -473,9 +463,8 @@ func.func @add_sd(%arga: tensor<32xf32, #SV>, %argb: tensor<32xf32>, %argx: tens
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32xf32>
-// CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_9]], %[[VAL_10]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_10]] : memref<32xf32>)
 // CHECK:           %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_13:.*]] = %[[VAL_11]] to %[[VAL_12]] step %[[VAL_4]] {
@@ -511,9 +500,8 @@ func.func @mul_sd(%arga: tensor<32xf32, #SV>, %argb: tensor<32xf32>, %argx: tens
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_12]] : memref<32xf32>)
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_3]]] : memref<?xindex>
@@ -595,9 +583,8 @@ func.func @add_ss(%arga: tensor<32xf32, #SV>, %argb: tensor<32xf32, #SV>, %argx:
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_3]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32xf32> to memref<32xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_12]] : memref<32xf32>)
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_3]]] : memref<?xindex>
@@ -658,9 +645,8 @@ func.func @mul_ss(%arga: tensor<32xf32, #SV>, %argb: tensor<32xf32, #SV>, %argx:
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_3]] : memref<16xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<16xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<16xf32> to memref<16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_13]] : memref<16xf32>)
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_4]]] : memref<?xindex>
@@ -752,9 +738,8 @@ func.func @two_way_inv(%arga: tensor<16xf32, #SV>, %argb: tensor<16xf32, #SV>, %
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_3]] : memref<16xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<16xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<16xf32> to memref<16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_13]] : memref<16xf32>)
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_4]]] : memref<?xindex>
@@ -1127,9 +1112,8 @@ func.func @sum_reduction_inv(%arga: tensor<16xf32, #SV>,
 // CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.indices %[[VAL_3]], %[[VAL_5]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_15:.*]] = sparse_tensor.values %[[VAL_3]] : tensor<?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf64>
 // CHECK-DAG:       %[[VAL_16:.*]] = tensor.dim %[[VAL_4]], %[[VAL_5]] : tensor<?xf64>
-// CHECK-DAG:       %[[VAL_17:.*]] = bufferization.to_memref %[[VAL_4]] : memref<?xf64>
 // CHECK-DAG:       %[[VAL_18:.*]] = memref.alloc(%[[VAL_16]]) : memref<?xf64>
-// CHECK:           memref.copy %[[VAL_17]], %[[VAL_18]] : memref<?xf64> to memref<?xf64>
+// CHECK:           linalg.fill ins(%{{.*}} : f64) outs(%[[VAL_18]] : memref<?xf64>)
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_20:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_21:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_5]]] : memref<?xindex>
diff --git a/mlir/test/Dialect/SparseTensor/sparse_2d.mlir b/mlir/test/Dialect/SparseTensor/sparse_2d.mlir
index d052489d4ff74..d6ed917517a32 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_2d.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_2d.mlir
@@ -26,9 +26,8 @@
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_9]], %[[VAL_10]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_10]] : memref<32x16xf32>)
 // CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_6]] {
 // CHECK:             scf.for %[[VAL_12:.*]] = %[[VAL_5]] to %[[VAL_4]] step %[[VAL_6]] {
 // CHECK:               %[[VAL_13:.*]] = arith.muli %[[VAL_11]], %[[VAL_4]] : index
@@ -63,9 +62,8 @@ func.func @add_dd(%arga: tensor<32x16xf32, #Tdd>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_9]], %[[VAL_10]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_10]] : memref<32x16xf32>)
 // CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_6]] {
 // CHECK:             scf.for %[[VAL_12:.*]] = %[[VAL_5]] to %[[VAL_4]] step %[[VAL_6]] {
 // CHECK:               %[[VAL_13:.*]] = arith.muli %[[VAL_11]], %[[VAL_4]] : index
@@ -103,9 +101,8 @@ func.func @mul_dd(%arga: tensor<32x16xf32, #Tdd>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_7]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_13]] : memref<32x16xf32>)
 // CHECK:           scf.for %[[VAL_14:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_7]] {
 // CHECK:             %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_14]]] : memref<?xindex>
 // CHECK:             %[[VAL_16:.*]] = arith.addi %[[VAL_14]], %[[VAL_7]] : index
@@ -165,9 +162,8 @@ func.func @add_ds(%arga: tensor<32x16xf32, #Tds>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_5]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_11]] : memref<32x16xf32>)
 // CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_12]]] : memref<?xindex>
 // CHECK:             %[[VAL_14:.*]] = arith.addi %[[VAL_12]], %[[VAL_5]] : index
@@ -207,9 +203,8 @@ func.func @mul_ds(%arga: tensor<32x16xf32, #Tds>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_6]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_13]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_16:.*]]:2 = scf.while (%[[VAL_17:.*]] = %[[VAL_14]], %[[VAL_18:.*]] = %[[VAL_6]]) : (index, index) -> (index, index) {
@@ -274,9 +269,8 @@ func.func @add_sd(%arga: tensor<32x16xf32, #Tsd>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_11]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_14:.*]] = %[[VAL_12]] to %[[VAL_13]] step %[[VAL_5]] {
@@ -319,9 +313,8 @@ func.func @mul_sd(%arga: tensor<32x16xf32, #Tsd>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_7]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_14:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_14]], %[[VAL_15]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_15]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_18:.*]]:2 = scf.while (%[[VAL_19:.*]] = %[[VAL_16]], %[[VAL_20:.*]] = %[[VAL_6]]) : (index, index) -> (index, index) {
@@ -411,9 +404,8 @@ func.func @add_ss(%arga: tensor<32x16xf32, #Tss>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_4]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_12]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_15:.*]] = %[[VAL_13]] to %[[VAL_14]] step %[[VAL_4]] {
@@ -459,9 +451,8 @@ func.func @mul_ss(%arga: tensor<32x16xf32, #Tss>, %argb: tensor<32x16xf32>, %arg
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_15:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_16:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_15]], %[[VAL_16]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_16]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_3]]] : memref<?xindex>
@@ -624,9 +615,8 @@ func.func @add_ss_ss(%arga: tensor<32x16xf32, #Tss>, %argb: tensor<32x16xf32, #T
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_15:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_16:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_15]], %[[VAL_16]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_16]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_3]]] : memref<?xindex>
@@ -720,9 +710,8 @@ func.func @mul_ss_ss(%arga: tensor<32x16xf32, #Tss>, %argb: tensor<32x16xf32, #T
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_7]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_7]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_14:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_14]], %[[VAL_15]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_15]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_18:.*]]:2 = scf.while (%[[VAL_19:.*]] = %[[VAL_16]], %[[VAL_20:.*]] = %[[VAL_5]]) : (index, index) -> (index, index) {
@@ -825,9 +814,8 @@ func.func @add_sd_ds(%arga: tensor<32x16xf32, #Tsd>, %argb: tensor<32x16xf32, #T
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_5]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_5]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<32x16xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<32x16xf32> to memref<32x16xf32>
+// CHECK:           linalg.fill ins(%{{.*}} : f32) outs(%[[VAL_13]] : memref<32x16xf32>)
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_16:.*]] = %[[VAL_14]] to %[[VAL_15]] step %[[VAL_5]] {
@@ -980,9 +968,8 @@ func.func @sum_reduction(%arga: tensor<10x20xf32, #Tds>, %argx: tensor<f32>) ->
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<?x?xf64, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf64>
 // CHECK-DAG:       %[[VAL_8:.*]] = tensor.dim %[[VAL_1]], %[[VAL_3]] : tensor<?x?xf64>
 // CHECK-DAG:       %[[VAL_9:.*]] = tensor.dim %[[VAL_1]], %[[VAL_4]] : tensor<?x?xf64>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<?x?xf64>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc(%[[VAL_8]], %[[VAL_9]]) : memref<?x?xf64>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<?x?xf64> to memref<?x?xf64>
+// CHECK:           linalg.fill ins(%{{.*}} : f64) outs(%[[VAL_11]] : memref<?x?xf64>)
 // CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_3]] to %[[VAL_8]] step %[[VAL_4]] {
 // CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_12]]] : memref<?xindex>
 // CHECK:             %[[VAL_14:.*]] = arith.addi %[[VAL_12]], %[[VAL_4]] : index
diff --git a/mlir/test/Dialect/SparseTensor/sparse_3d.mlir b/mlir/test/Dialect/SparseTensor/sparse_3d.mlir
index 521404bc18cad..bdc74e7d40be7 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_3d.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_3d.mlir
@@ -26,6 +26,7 @@
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 8 : index
@@ -33,9 +34,8 @@
 // CHECK-DAG:       %[[VAL_7:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_11]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_6]] to %[[VAL_3]] step %[[VAL_7]] {
 // CHECK:             scf.for %[[VAL_13:.*]] = %[[VAL_6]] to %[[VAL_4]] step %[[VAL_7]] {
 // CHECK:               %[[VAL_14:.*]] = arith.muli %[[VAL_12]], %[[VAL_4]] : index
@@ -68,6 +68,7 @@ func.func @add_ddd(%arga: tensor<32x16x8xf32, #Tddd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 8 : index
@@ -75,9 +76,8 @@ func.func @add_ddd(%arga: tensor<32x16x8xf32, #Tddd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_7:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_11]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_6]] to %[[VAL_3]] step %[[VAL_7]] {
 // CHECK:             scf.for %[[VAL_13:.*]] = %[[VAL_6]] to %[[VAL_4]] step %[[VAL_7]] {
 // CHECK:               %[[VAL_14:.*]] = arith.muli %[[VAL_12]], %[[VAL_4]] : index
@@ -110,6 +110,7 @@ func.func @mul_ddd(%arga: tensor<32x16x8xf32, #Tddd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 16 : index
@@ -121,9 +122,8 @@ func.func @mul_ddd(%arga: tensor<32x16x8xf32, #Tddd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_14:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_14]], %[[VAL_15]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_15]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_16:.*]] = %[[VAL_7]] to %[[VAL_4]] step %[[VAL_9]] {
 // CHECK:             scf.for %[[VAL_17:.*]] = %[[VAL_7]] to %[[VAL_5]] step %[[VAL_9]] {
 // CHECK:               %[[VAL_18:.*]] = arith.muli %[[VAL_16]], %[[VAL_5]] : index
@@ -180,6 +180,7 @@ func.func @add_dds(%arga: tensor<32x16x8xf32, #Tdds>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 16 : index
@@ -189,9 +190,8 @@ func.func @add_dds(%arga: tensor<32x16x8xf32, #Tdds>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_13]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_14:.*]] = %[[VAL_6]] to %[[VAL_4]] step %[[VAL_7]] {
 // CHECK:             scf.for %[[VAL_15:.*]] = %[[VAL_6]] to %[[VAL_5]] step %[[VAL_7]] {
 // CHECK:               %[[VAL_16:.*]] = arith.muli %[[VAL_14]], %[[VAL_5]] : index
@@ -226,6 +226,7 @@ func.func @mul_dds(%arga: tensor<32x16x8xf32, #Tdds>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 8 : index
@@ -236,9 +237,8 @@ func.func @mul_dds(%arga: tensor<32x16x8xf32, #Tdds>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_8]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_14:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_13]], %[[VAL_14]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_14]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_15:.*]] = %[[VAL_7]] to %[[VAL_3]] step %[[VAL_8]] {
 // CHECK:             %[[VAL_16:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_15]]] : memref<?xindex>
 // CHECK:             %[[VAL_17:.*]] = arith.addi %[[VAL_15]], %[[VAL_8]] : index
@@ -299,6 +299,7 @@ func.func @add_dsd(%arga: tensor<32x16x8xf32, #Tdsd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 8 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 0 : index
@@ -307,9 +308,8 @@ func.func @add_dsd(%arga: tensor<32x16x8xf32, #Tdsd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_6]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_12]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_13:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_6]] {
 // CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_13]]] : memref<?xindex>
 // CHECK:             %[[VAL_15:.*]] = arith.addi %[[VAL_13]], %[[VAL_6]] : index
@@ -344,6 +344,7 @@ func.func @mul_dsd(%arga: tensor<32x16x8xf32, #Tdsd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 16 : index
@@ -357,9 +358,8 @@ func.func @mul_dsd(%arga: tensor<32x16x8xf32, #Tdsd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_16:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_17:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_16]], %[[VAL_17]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_17]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_18:.*]] = %[[VAL_8]] to %[[VAL_4]] step %[[VAL_9]] {
 // CHECK:             %[[VAL_19:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_18]]] : memref<?xindex>
 // CHECK:             %[[VAL_20:.*]] = arith.addi %[[VAL_18]], %[[VAL_9]] : index
@@ -444,6 +444,7 @@ func.func @add_dss(%arga: tensor<32x16x8xf32, #Tdss>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 0 : index
@@ -454,9 +455,8 @@ func.func @add_dss(%arga: tensor<32x16x8xf32, #Tdss>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_14:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_13]], %[[VAL_14]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_14]] : memref<32x16x8xf32>)
 // CHECK:           scf.for %[[VAL_15:.*]] = %[[VAL_5]] to %[[VAL_4]] step %[[VAL_6]] {
 // CHECK:             %[[VAL_16:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_15]]] : memref<?xindex>
 // CHECK:             %[[VAL_17:.*]] = arith.addi %[[VAL_15]], %[[VAL_6]] : index
@@ -493,6 +493,7 @@ func.func @mul_dss(%arga: tensor<32x16x8xf32, #Tdss>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 8 : index
@@ -503,9 +504,8 @@ func.func @mul_dss(%arga: tensor<32x16x8xf32, #Tdss>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_7]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_14:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_13]], %[[VAL_14]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_14]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_8]]] : memref<?xindex>
 // CHECK:           %[[VAL_17:.*]]:2 = scf.while (%[[VAL_18:.*]] = %[[VAL_15]], %[[VAL_19:.*]] = %[[VAL_7]]) : (index, index) -> (index, index) {
@@ -571,6 +571,7 @@ func.func @add_sdd(%arga: tensor<32x16x8xf32, #Tsdd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 8 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 0 : index
@@ -579,9 +580,8 @@ func.func @add_sdd(%arga: tensor<32x16x8xf32, #Tsdd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_5]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_11]], %[[VAL_12]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_12]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_15:.*]] = %[[VAL_13]] to %[[VAL_14]] step %[[VAL_6]] {
@@ -617,6 +617,7 @@ func.func @mul_sdd(%arga: tensor<32x16x8xf32, #Tsdd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 16 : index
@@ -630,9 +631,8 @@ func.func @mul_sdd(%arga: tensor<32x16x8xf32, #Tsdd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_16:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_17:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_16]], %[[VAL_17]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_17]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_8]]] : memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_9]]] : memref<?xindex>
 // CHECK:           %[[VAL_20:.*]]:2 = scf.while (%[[VAL_21:.*]] = %[[VAL_18]], %[[VAL_22:.*]] = %[[VAL_8]]) : (index, index) -> (index, index) {
@@ -722,6 +722,7 @@ func.func @add_sds(%arga: tensor<32x16x8xf32, #Tsds>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 0 : index
@@ -732,9 +733,8 @@ func.func @add_sds(%arga: tensor<32x16x8xf32, #Tsds>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "dense", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_14:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_13]], %[[VAL_14]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_14]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_6]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_17:.*]] = %[[VAL_15]] to %[[VAL_16]] step %[[VAL_6]] {
@@ -772,6 +772,7 @@ func.func @mul_sds(%arga: tensor<32x16x8xf32, #Tsds>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 16 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 8 : index
@@ -784,9 +785,8 @@ func.func @mul_sds(%arga: tensor<32x16x8xf32, #Tsds>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_8]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_14:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_15:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_16:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_15]], %[[VAL_16]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_16]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_7]]] : memref<?xindex>
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_8]]] : memref<?xindex>
 // CHECK:           %[[VAL_19:.*]]:2 = scf.while (%[[VAL_20:.*]] = %[[VAL_17]], %[[VAL_21:.*]] = %[[VAL_7]]) : (index, index) -> (index, index) {
@@ -880,6 +880,7 @@ func.func @add_ssd(%arga: tensor<32x16x8xf32, #Tssd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 8 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
@@ -889,9 +890,8 @@ func.func @add_ssd(%arga: tensor<32x16x8xf32, #Tssd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_5]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_13]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_16:.*]] = %[[VAL_14]] to %[[VAL_15]] step %[[VAL_5]] {
@@ -929,6 +929,7 @@ func.func @mul_ssd(%arga: tensor<32x16x8xf32, #Tssd>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 32 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 16 : index
@@ -944,9 +945,8 @@ func.func @mul_ssd(%arga: tensor<32x16x8xf32, #Tssd>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_16:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_17:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_18:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_19:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_18]], %[[VAL_19]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_19]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_20:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_8]]] : memref<?xindex>
 // CHECK:           %[[VAL_21:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_9]]] : memref<?xindex>
 // CHECK:           %[[VAL_22:.*]]:2 = scf.while (%[[VAL_23:.*]] = %[[VAL_20]], %[[VAL_24:.*]] = %[[VAL_8]]) : (index, index) -> (index, index) {
@@ -1064,6 +1064,7 @@ func.func @add_sss(%arga: tensor<32x16x8xf32, #Tsss>, %argb: tensor<32x16x8xf32>
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<32x16x8xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<32x16x8xf32>) -> tensor<32x16x8xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
@@ -1075,9 +1076,8 @@ func.func @add_sss(%arga: tensor<32x16x8xf32, #Tsss>, %argb: tensor<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16x8xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_1]] : memref<32x16x8xf32>
-// CHECK-DAG:       %[[VAL_14:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32x16x8xf32>
 // CHECK-DAG:       %[[VAL_15:.*]] = memref.alloc() : memref<32x16x8xf32>
-// CHECK:           memref.copy %[[VAL_14]], %[[VAL_15]] : memref<32x16x8xf32> to memref<32x16x8xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_15]] : memref<32x16x8xf32>)
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_18:.*]] = %[[VAL_16]] to %[[VAL_17]] step %[[VAL_5]] {
@@ -1314,6 +1314,7 @@ func.func @sum_reduction_inv(%arga: tensor<?x?x?xf32>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<20xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<30xf32>,
 // CHECK-SAME:      %[[VAL_3:.*]]: tensor<10x20x30xf32>) -> tensor<10x20x30xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 10 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 20 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant 30 : index
@@ -1322,9 +1323,8 @@ func.func @sum_reduction_inv(%arga: tensor<?x?x?xf32>,
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_1]] : memref<20xf32>
 // CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<30xf32>
-// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_3]] : memref<10x20x30xf32>
 // CHECK-DAG:       %[[VAL_13:.*]] = memref.alloc() : memref<10x20x30xf32>
-// CHECK:           memref.copy %[[VAL_12]], %[[VAL_13]] : memref<10x20x30xf32> to memref<10x20x30xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_13]] : memref<10x20x30xf32>)
 // CHECK:           scf.for %[[VAL_14:.*]] = %[[VAL_7]] to %[[VAL_4]] step %[[VAL_8]] {
 // CHECK:             %[[VAL_15:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_14]]] : memref<?xf32>
 // CHECK:             scf.for %[[VAL_16:.*]] = %[[VAL_7]] to %[[VAL_5]] step %[[VAL_8]] {
diff --git a/mlir/test/Dialect/SparseTensor/sparse_affine.mlir b/mlir/test/Dialect/SparseTensor/sparse_affine.mlir
index 3251e459a7c46..1520a0e0275db 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_affine.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_affine.mlir
@@ -70,6 +70,7 @@ func.func @mul_inv_dense1d(%arga: tensor<32xf32, #SpVec>,
 // CHECK-SAME:                             %[[VAL_0:.*]]: tensor<32xi32, #sparse_tensor.encoding<{{{.*}}}>>,
 // CHECK-SAME:                             %[[VAL_1:.*]]: tensor<34xi32>,
 // CHECK-SAME:                             %[[VAL_2:.*]]: tensor<32xi32>) -> tensor<32xi32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0 : i32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 2 : index
@@ -77,9 +78,8 @@ func.func @mul_inv_dense1d(%arga: tensor<32xf32, #SpVec>,
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_3]] : tensor<32xi32, #sparse_tensor.encoding<{{{.*}}}>>
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xi32, #sparse_tensor.encoding<{{{.*}}}>>
 // CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<34xi32>
-// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<32xi32>
 // CHECK-DAG:       %[[VAL_11:.*]] = memref.alloc() : memref<32xi32>
-// CHECK:           memref.copy %[[VAL_10]], %[[VAL_11]] : memref<32xi32> to memref<32xi32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : i32) outs(%[[VAL_11]] : memref<32xi32>)
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_14:.*]] = %[[VAL_12]] to %[[VAL_13]] step %[[VAL_4]] {
diff --git a/mlir/test/Dialect/SparseTensor/sparse_nd.mlir b/mlir/test/Dialect/SparseTensor/sparse_nd.mlir
index 3a7a32d953c25..1e1e7f62737e6 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_nd.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_nd.mlir
@@ -24,6 +24,7 @@
 // CHECK-SAME:              %[[VAL_0:.*]]: tensor<10x20x30x40x50x60x70x80xf32>,
 // CHECK-SAME:              %[[VAL_1:.*]]: tensor<80x70x60x50x40x30x20x10xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense", "compressed", "compressed", "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>>,
 // CHECK-SAME:              %[[VAL_2:.*]]: tensor<10x20x30x40x50x60x70x80xf32>) -> tensor<10x20x30x40x50x60x70x80xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 3 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 4 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 10 : index
@@ -40,9 +41,8 @@
 // CHECK-DAG:       %[[VAL_16:.*]] = sparse_tensor.pointers %[[VAL_1]], %[[VAL_4]] : tensor<80x70x60x50x40x30x20x10xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense", "compressed", "compressed", "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_17:.*]] = sparse_tensor.indices %[[VAL_1]], %[[VAL_4]] : tensor<80x70x60x50x40x30x20x10xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense", "compressed", "compressed", "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xindex>
 // CHECK-DAG:       %[[VAL_18:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<80x70x60x50x40x30x20x10xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "dense", "dense", "compressed", "compressed", "dense", "dense", "dense" ], pointerBitWidth = 0, indexBitWidth = 0 }>> to memref<?xf32>
-// CHECK-DAG:       %[[VAL_19:.*]] = bufferization.to_memref %[[VAL_2]] : memref<10x20x30x40x50x60x70x80xf32>
 // CHECK-DAG:       %[[VAL_20:.*]] = memref.alloc() : memref<10x20x30x40x50x60x70x80xf32>
-// CHECK:           memref.copy %[[VAL_19]], %[[VAL_20]] : memref<10x20x30x40x50x60x70x80xf32> to memref<10x20x30x40x50x60x70x80xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_20]] : memref<10x20x30x40x50x60x70x80xf32>
 // CHECK:           scf.for %[[VAL_21:.*]] = %[[VAL_11]] to %[[VAL_10]] step %[[VAL_12]] {
 // CHECK:             scf.for %[[VAL_22:.*]] = %[[VAL_11]] to %[[VAL_9]] step %[[VAL_12]] {
 // CHECK:               %[[VAL_23:.*]] = arith.muli %[[VAL_21]], %[[VAL_9]] : index
diff --git a/mlir/test/Dialect/SparseTensor/sparse_outbuf.mlir b/mlir/test/Dialect/SparseTensor/sparse_outbuf.mlir
new file mode 100644
index 0000000000000..8f9bd40355a97
--- /dev/null
+++ b/mlir/test/Dialect/SparseTensor/sparse_outbuf.mlir
@@ -0,0 +1,148 @@
+// RUN: mlir-opt %s -sparsification | FileCheck %s
+
+#SV = #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>
+
+#trait = {
+  indexing_maps = [
+    affine_map<(i) -> (i)>,  // A (in)
+    affine_map<(i) -> (i)>   // X (out)
+  ],
+  iterator_types = ["parallel"]
+}
+
+// CHECK-LABEL:   func.func @allout_inplace(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<10xi32, #{{.*}}>,
+// CHECK-SAME:      %[[VAL_1:.*]]: tensor<10xf32> {linalg.inplaceable = true}) -> tensor<10xf32> {
+// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
+// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 1 : index
+// CHECK:           %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<10xi32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<10xi32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10xi32, #{{.*}}> to memref<?xi32>
+// CHECK:           %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<10xf32>
+// CHECK:           linalg.fill ins(%[[VAL_3]] : f32) outs(%[[VAL_8]] : memref<10xf32>)
+// CHECK:           %[[VAL_9:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_2]]] : memref<?xindex>
+// CHECK:           %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_4]]] : memref<?xindex>
+// CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_4]] {
+// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xindex>
+// CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<?xi32>
+// CHECK:             %[[VAL_14:.*]] = arith.sitofp %[[VAL_13]] : i32 to f32
+// CHECK:             memref.store %[[VAL_14]], %[[VAL_8]]{{\[}}%[[VAL_12]]] : memref<10xf32>
+// CHECK:           }
+// CHECK:           %[[VAL_15:.*]] = bufferization.to_tensor %[[VAL_8]] : memref<10xf32>
+// CHECK:           return %[[VAL_15]] : tensor<10xf32>
+// CHECK:         }
+func.func @allout_inplace(%arga: tensor<10xi32, #SV>,
+                          %argb: tensor<10xf32> {linalg.inplaceable = true}) -> tensor<10xf32> {
+  %0 = linalg.generic #trait
+  ins(%arga: tensor<10xi32, #SV>)
+  outs(%argb: tensor<10xf32>) {
+    ^bb(%a: i32, %x : f32):
+      %cst = arith.sitofp %a : i32 to f32
+      linalg.yield %cst : f32
+  } -> tensor<10xf32>
+  return %0 : tensor<10xf32>
+}
+
+// CHECK-LABEL:   func.func @allout_materialize(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<10xi32, #{{.*}}>) -> tensor<10xf32> {
+// CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 0 : index
+// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
+// CHECK:           %[[VAL_4:.*]] = bufferization.alloc_tensor() : tensor<10xf32>
+// CHECK:           %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_1]] : tensor<10xi32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_1]] : tensor<10xi32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10xi32, #{{.*}}> to memref<?xi32>
+// CHECK:           %[[VAL_8:.*]] = memref.alloc() : memref<10xf32>
+// CHECK:           linalg.fill ins(%[[VAL_2]] : f32) outs(%[[VAL_8]] : memref<10xf32>)
+// CHECK:           %[[VAL_9:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_1]]] : memref<?xindex>
+// CHECK:           %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_3]]] : memref<?xindex>
+// CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_3]] {
+// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xindex>
+// CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<?xi32>
+// CHECK:             %[[VAL_14:.*]] = arith.sitofp %[[VAL_13]] : i32 to f32
+// CHECK:             memref.store %[[VAL_14]], %[[VAL_8]]{{\[}}%[[VAL_12]]] : memref<10xf32>
+// CHECK:           }
+// CHECK:           %[[VAL_15:.*]] = bufferization.to_tensor %[[VAL_8]] : memref<10xf32>
+// CHECK:           return %[[VAL_15]] : tensor<10xf32>
+// CHECK:         }
+func.func @allout_materialize(%arga: tensor<10xi32, #SV>) -> tensor<10xf32> {
+  %m = bufferization.alloc_tensor() : tensor<10xf32>
+  %0 = linalg.generic #trait
+  ins(%arga: tensor<10xi32, #SV>)
+  outs(%m: tensor<10xf32>) {
+    ^bb(%a: i32, %x : f32):
+      %cst = arith.sitofp %a : i32 to f32
+      linalg.yield %cst : f32
+  } -> tensor<10xf32>
+  return %0 : tensor<10xf32>
+}
+
+// CHECK-LABEL:   func.func @update_notinplace(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<10xf32, #{{.*}}>,
+// CHECK-SAME:      %[[VAL_1:.*]]: tensor<10xf32> {linalg.inplaceable = false}) -> tensor<10xf32> {
+// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
+// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
+// CHECK:           %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<10xf32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<10xf32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10xf32, #{{.*}}> to memref<?xf32>
+// CHECK:           %[[VAL_7:.*]] = memref.alloc() : memref<10xf32>
+// CHECK:           %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<10xf32>
+// CHECK:           memref.copy %[[VAL_8]], %[[VAL_7]] : memref<10xf32> to memref<10xf32>
+// CHECK:           %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
+// CHECK:           %[[VAL_10:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
+// CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_9]] to %[[VAL_10]] step %[[VAL_3]] {
+// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_11]]] : memref<?xindex>
+// CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xf32>
+// CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_12]]] : memref<10xf32>
+// CHECK:             %[[VAL_15:.*]] = arith.addf %[[VAL_13]], %[[VAL_14]] : f32
+// CHECK:             memref.store %[[VAL_15]], %[[VAL_7]]{{\[}}%[[VAL_12]]] : memref<10xf32>
+// CHECK:           }
+// CHECK:           %[[VAL_16:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<10xf32>
+// CHECK:           return %[[VAL_16]] : tensor<10xf32>
+// CHECK:         }
+func.func @update_notinplace(%arga: tensor<10xf32, #SV>,
+                             %argb: tensor<10xf32> {linalg.inplaceable = false}) -> tensor<10xf32> {
+  %0 = linalg.generic #trait
+  ins(%arga: tensor<10xf32, #SV>)
+  outs(%argb: tensor<10xf32>) {
+    ^bb(%a: f32, %x : f32):
+      %up = arith.addf %a, %x : f32
+      linalg.yield %up : f32
+  } -> tensor<10xf32>
+  return %0 : tensor<10xf32>
+}
+
+// CHECK-LABEL:   func.func @update_inplace(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<10xf32, #{{.*}}>,
+// CHECK-SAME:      %[[VAL_1:.*]]: tensor<10xf32> {linalg.inplaceable = true}) -> tensor<10xf32> {
+// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
+// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
+// CHECK:           %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]], %[[VAL_2]] : tensor<10xf32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]], %[[VAL_2]] : tensor<10xf32, #{{.*}}> to memref<?xindex>
+// CHECK:           %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10xf32, #{{.*}}> to memref<?xf32>
+// CHECK:           %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_1]] : memref<10xf32>
+// CHECK:           %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
+// CHECK:           %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_3]]] : memref<?xindex>
+// CHECK:           scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_3]] {
+// CHECK:             %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_10]]] : memref<?xindex>
+// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xf32>
+// CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<10xf32>
+// CHECK:             %[[VAL_14:.*]] = arith.addf %[[VAL_12]], %[[VAL_13]] : f32
+// CHECK:             memref.store %[[VAL_14]], %[[VAL_7]]{{\[}}%[[VAL_11]]] : memref<10xf32>
+// CHECK:           }
+// CHECK:           %[[VAL_15:.*]] = bufferization.to_tensor %[[VAL_7]] : memref<10xf32>
+// CHECK:           return %[[VAL_15]] : tensor<10xf32>
+// CHECK:         }
+func.func @update_inplace(%arga: tensor<10xf32, #SV>,
+                          %argb: tensor<10xf32> {linalg.inplaceable = true}) -> tensor<10xf32> {
+  %0 = linalg.generic #trait
+  ins(%arga: tensor<10xf32, #SV>)
+  outs(%argb: tensor<10xf32>) {
+    ^bb(%a: f32, %x : f32):
+      %up = arith.addf %a, %x : f32
+      linalg.yield %up : f32
+  } -> tensor<10xf32>
+  return %0 : tensor<10xf32>
+}
+
diff --git a/mlir/test/Dialect/SparseTensor/sparse_perm.mlir b/mlir/test/Dialect/SparseTensor/sparse_perm.mlir
index b30f56260a53d..1c7bccebaedf5 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_perm.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_perm.mlir
@@ -17,15 +17,15 @@
 // CHECK-LABEL:   func @sparse_static_dims(
 // CHECK-SAME:                          %[[VAL_0:.*]]: tensor<10x20x30xf32, #sparse_tensor.encoding<{{{.*}}}>>,
 // CHECK-SAME:                          %[[VAL_1:.*]]: tensor<20x30x10xf32>) -> tensor<20x30x10xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 20 : index
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 30 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 10 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10x20x30xf32, #sparse_tensor.encoding<{{{.*}}}>>
-// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.to_memref %[[VAL_1]] : memref<20x30x10xf32>
 // CHECK-DAG:       %[[VAL_9:.*]] = memref.alloc() : memref<20x30x10xf32>
-// CHECK:           memref.copy %[[VAL_8]], %[[VAL_9]] : memref<20x30x10xf32> to memref<20x30x10xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_9]] : memref<20x30x10xf32>)
 // CHECK:           scf.for %[[VAL_10:.*]] = %[[VAL_5]] to %[[VAL_3]] step %[[VAL_6]] {
 // CHECK:             scf.for %[[VAL_11:.*]] = %[[VAL_5]] to %[[VAL_4]] step %[[VAL_6]] {
 // CHECK:               %[[VAL_12:.*]] = arith.muli %[[VAL_10]], %[[VAL_4]] : index
@@ -42,7 +42,7 @@
 // CHECK:           return %[[VAL_18]] : tensor<20x30x10xf32>
 // CHECK:         }
 func.func @sparse_static_dims(%arga: tensor<10x20x30xf32, #X>,
-                         %argx: tensor<20x30x10xf32>) -> tensor<20x30x10xf32> {
+                              %argx: tensor<20x30x10xf32>) -> tensor<20x30x10xf32> {
   %0 = linalg.generic #trait
     ins(%arga: tensor<10x20x30xf32, #X>)
     outs(%argx: tensor<20x30x10xf32>) {
@@ -55,6 +55,7 @@ func.func @sparse_static_dims(%arga: tensor<10x20x30xf32, #X>,
 // CHECK-LABEL:   func @sparse_dynamic_dims(
 // CHECK-SAME:                          %[[VAL_0:.*]]: tensor<?x?x?xf32, #sparse_tensor.encoding<{{{.*}}}>>,
 // CHECK-SAME:                          %[[VAL_1:.*]]: tensor<?x?x?xf32>) -> tensor<?x?x?xf32> {
+// CHECK-DAG:       %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 2 : index
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 1 : index
@@ -62,9 +63,8 @@ func.func @sparse_static_dims(%arga: tensor<10x20x30xf32, #X>,
 // CHECK-DAG:       %[[VAL_6:.*]] = tensor.dim %[[VAL_1]], %[[VAL_3]] : tensor<?x?x?xf32>
 // CHECK-DAG:       %[[VAL_7:.*]] = tensor.dim %[[VAL_1]], %[[VAL_4]] : tensor<?x?x?xf32>
 // CHECK-DAG:       %[[VAL_8:.*]] = tensor.dim %[[VAL_1]], %[[VAL_2]] : tensor<?x?x?xf32>
-// CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<?x?x?xf32>
 // CHECK-DAG:       %[[VAL_10:.*]] = memref.alloc(%[[VAL_6]], %[[VAL_7]], %[[VAL_8]]) : memref<?x?x?xf32>
-// CHECK:           memref.copy %[[VAL_9]], %[[VAL_10]] : memref<?x?x?xf32> to memref<?x?x?xf32>
+// CHECK:           linalg.fill ins(%[[ZERO]] : f32) outs(%[[VAL_10]] : memref<?x?x?xf32>)
 // CHECK:           scf.for %[[VAL_11:.*]] = %[[VAL_3]] to %[[VAL_7]] step %[[VAL_4]] {
 // CHECK:             scf.for %[[VAL_12:.*]] = %[[VAL_3]] to %[[VAL_8]] step %[[VAL_4]] {
 // CHECK:               %[[VAL_13:.*]] = arith.muli %[[VAL_8]], %[[VAL_11]] : index