[mlir][sparse] build proper insertion chain

[aartbik]

Summary

The alloc->insert/compress->load chain needs to be
properly represented with an SSA chain now in loops
and if statements to properly reflect the modifying
behavior (runtime support lib is forgiving on breaking
this, but the new codegen is not).

[aartbik]

rebased with main

[aartbik]

removed unused constants for now

[aartbik]

rebased with main to resolve conflict

mlir/lib/Dialect/SparseTensor/Transforms/CodegenUtils.h

 void genDenseTensorOrSparseConstantIterLoop(
     OpBuilder &builder, Location loc, Value src, unsigned rank,
     function_ref<void(OpBuilder &, Location, Value, ValueRange)> bodyBuilder);
 
 /// Populates given sizes array from dense tensor or sparse tensor constant.
 void sizesFromSrc(OpBuilder &builder, SmallVector<Value, 4> &sizes,
                   Location loc, Value src);
 
+/// Scans to top of generated loop.
+Operation *getTop(Operation *op);
+
 //===----------------------------------------------------------------------===//
 // Inlined constant generators.
 //
 // All these functions are just wrappers to improve code legibility;
 // therefore, we mark them as `inline` to avoid introducing any additional
 // overhead due to the legibility.
 //
 // TODO: Ideally these should move upstream, so that we don't

mlir/lib/Dialect/SparseTensor/Transforms/CodegenUtils.cpp

 
 void mlir::sparse_tensor::sizesFromSrc(OpBuilder &builder,
                                        SmallVector<Value, 4> &sizes,
                                        Location loc, Value src) {
   unsigned rank = src.getType().cast<ShapedType>().getRank();
   for (unsigned i = 0; i < rank; i++)
     sizes.push_back(linalg::createOrFoldDimOp(builder, loc, src, i));
 }
+
+Operation *mlir::sparse_tensor::getTop(Operation *op) {
+  for (; isa<scf::ForOp>(op->getParentOp()) ||
+         isa<scf::WhileOp>(op->getParentOp()) ||
+         isa<scf::ParallelOp>(op->getParentOp()) ||
+         isa<scf::IfOp>(op->getParentOp());
+       op = op->getParentOp())
+    ;
+  return op;
+}

mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorCodegen.cpp

 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 using namespace mlir;
 using namespace mlir::sparse_tensor;
 
 namespace {
 
+static constexpr uint64_t DimSizesIdx = 0;
+static constexpr uint64_t DimCursorIdx = 1;
 static constexpr uint64_t MemSizesIdx = 2;
 static constexpr uint64_t FieldsIdx = 3;
 
 //===----------------------------------------------------------------------===//
 // Helper methods.
 //===----------------------------------------------------------------------===//
 
 /// Returns the "tuple" value of the adapted tensor.
matchAndRewrite(CompressOp op, OpAdaptor adaptor,
     indices.push_back(index);
     genInsert(rewriter, loc, dstType, fields, indices, value);
     rewriter.create<memref::StoreOp>(loc, constantZero(rewriter, loc, eltType),
                                      values, index);
     rewriter.create<memref::StoreOp>(loc, constantI1(rewriter, loc, false),
                                      filled, index);
     rewriter.create<scf::YieldOp>(loc, fields);
     // Deallocate the buffers on exit of the full loop nest.
-    Operation *parent = op;
-    for (; isa<scf::ForOp>(parent->getParentOp()) ||
-           isa<scf::WhileOp>(parent->getParentOp()) ||
-           isa<scf::ParallelOp>(parent->getParentOp()) ||
-           isa<scf::IfOp>(parent->getParentOp());
-         parent = parent->getParentOp())
-      ;
+    Operation *parent = getTop(op);
     rewriter.setInsertionPointAfter(parent);
     rewriter.create<memref::DeallocOp>(loc, values);
     rewriter.create<memref::DeallocOp>(loc, filled);
     rewriter.create<memref::DeallocOp>(loc, added);
     // Replace operation with resulting memrefs.
     rewriter.replaceOp(op,
                        genTuple(rewriter, loc, dstType, loop->getResults()));
     return success();

mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp

for (unsigned i = 0; i < rank - 1; i++)
       rewriter.create<memref::StoreOp>(loc, adaptor.getIndices()[i], mref,
                                        constantIndex(rewriter, loc, i));
     SmallString<12> name{"expInsert", primaryTypeFunctionSuffix(elemTp)};
     createFuncCall(rewriter, loc, name, {},
                    {tensor, mref, values, filled, added, count},
                    EmitCInterface::On);
     rewriter.replaceOp(op, adaptor.getTensor());
     // Deallocate the buffers on exit of the loop nest.
-    Operation *parent = op;
-    for (; isa<scf::ForOp>(parent->getParentOp()) ||
-           isa<scf::WhileOp>(parent->getParentOp()) ||
-           isa<scf::ParallelOp>(parent->getParentOp()) ||
-           isa<scf::IfOp>(parent->getParentOp());
-         parent = parent->getParentOp())
-      ;
+    Operation *parent = getTop(op);
     rewriter.setInsertionPointAfter(parent);
     rewriter.create<memref::DeallocOp>(loc, values);
     rewriter.create<memref::DeallocOp>(loc, filled);
     rewriter.create<memref::DeallocOp>(loc, added);
     return success();
   }
 };
 

mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp

 
 // Code generation.
 struct CodeGen {
   CodeGen(SparsificationOptions o, ValueRange tensors, unsigned numTensors,
           unsigned numLoops, OpOperand *op, unsigned nest,
           std::vector<unsigned> &ts)
       : options(o), loopEmitter(tensors, /*isLastOutput=*/true,
                                 /*isSparseOut=*/op != nullptr),
-        sparseOut(op), outerParNest(nest), topSort(ts) {}
+        sparseOut(op), outerParNest(nest), topSort(ts) {
+    if (op)
+      insChain = op->get();
+  }
   /// Sparsification options.
   SparsificationOptions options;
   /// Loop emitter helper class.
   SparseTensorLoopEmitter loopEmitter;
   /// Current reduction, updated during code generation. When indices of a
   /// reduction are exhausted, all inner loops can use a scalarized reduction.
   unsigned redExp = -1u;
   Value redVal;
   Reduction redKind = kNoReduc;
   unsigned redCustom = -1u;
   // Sparse tensor as output. Implemented either through direct injective
   // insertion in lexicographic index order or through access pattern expansion
   // in the innermost loop nest (`expValues` through `expCount`).
   OpOperand *sparseOut;
   unsigned outerParNest;
+  Value insChain; // bookkeeping for insertion chain
   Value expValues;
   Value expFilled;
   Value expAdded;
   Value expCount;
   // Topsort (reference should remain in scope).
   std::vector<unsigned> &topSort;
 
   Value getLoopIdxValue(size_t loopIdx) const {
static void genInsertionStore(CodeGen &codegen, OpBuilder &builder,
   // Direct insertion in lexicographic index order.
   if (!codegen.expValues) {
     unsigned rank = op.getRank(t);
     SmallVector<Value, 4> indices;
     for (unsigned i = 0; i < rank; i++) {
       assert(codegen.loopEmitter.getLoopIV(i));
       indices.push_back(codegen.loopEmitter.getLoopIV(i));
     }
-    builder.create<InsertOp>(loc, rhs, t->get(), indices);
+    codegen.insChain =
+        builder.create<InsertOp>(loc, rhs, codegen.insChain, indices);
     return;
   }
   // Generates insertion code along expanded access pattern.
   //   if (!expFilled[i]) then
   //     expFilled[i] = true
   //     expAdded[inserts++] = i
   //   endif
   //   values[i] = rhs
static void genTensorStore(Merger &merger, CodeGen &codegen, OpBuilder &builder,
   // Store during insertion.
   OpOperand *t = op.getOutputOperand(0);
   if (t == codegen.sparseOut) {
     if (!rhs) {
       // Only unary and binary are allowed to return uninitialized rhs
       // to indicate missing output.
       assert(merger.exp(exp).kind == kUnary || merger.exp(exp).kind == kBinary);
     } else if (merger.exp(exp).kind == kSelect) {
-      scf::IfOp ifOp = builder.create<scf::IfOp>(loc, rhs);
+      // Select operation insertion.
+      Value insChain = codegen.insChain;
+      assert(insChain);
+      SmallVector<Type, 1> types;
+      types.push_back(codegen.insChain.getType());
+      scf::IfOp ifOp =
+          builder.create<scf::IfOp>(loc, types, rhs, /*else=*/true);
       builder.setInsertionPointToStart(&ifOp.getThenRegion().front());
       // Existing value was preserved to be used here.
       assert(merger.exp(exp).val);
       Value v0 = merger.exp(exp).val;
       genInsertionStore(codegen, builder, op, t, v0);
       merger.exp(exp).val = Value();
+      // Yield modified insertion chain along true branch.
+      builder.create<scf::YieldOp>(op.getLoc(), codegen.insChain);
+      // Yield original insertion chain along false branch.
+      builder.setInsertionPointToStart(&ifOp.getElseRegion().front());
+      builder.create<scf::YieldOp>(loc, insChain);
+      // Done with if statement.
+      codegen.insChain = ifOp->getResult(0);
       builder.setInsertionPointAfter(ifOp);
     } else {
       genInsertionStore(codegen, builder, op, t, rhs);
     }
     return;
   }
   // Actual store.
   SmallVector<Value, 4> args;

 /// Generates an expanded access pattern in innermost dimension.
 static void genExpansion(Merger &merger, CodeGen &codegen, OpBuilder &builder,
                          linalg::GenericOp op, unsigned at, bool atStart) {
   OpOperand *lhs = codegen.sparseOut;
   if (!lhs || codegen.outerParNest != op.getRank(lhs) - 1 ||
       at != codegen.outerParNest)
     return; // not needed at this level
   assert(codegen.redVal == nullptr);
-  // Generate start or end of an expanded access pattern.
+  // Generate start or end of an expanded access pattern. Note that because
+  // an expension does not rely on the ongoing contents of the sparse storage
+  // scheme, we can use the original tensor as incoming SSA value (which
+  // simplifies codegen a bit). If expansion on the actual contents is ever
+  // needed, we will need to use the SSA value in the insertion chain instead.
   Value tensor = lhs->get();
   Location loc = op.getLoc();
   if (atStart) {
     auto dynShape = {ShapedType::kDynamicSize};
     Type etp = tensor.getType().cast<ShapedType>().getElementType();
     Type t1 = MemRefType::get(dynShape, etp);
     Type t2 = MemRefType::get(dynShape, builder.getI1Type());
     Type t3 = MemRefType::get(dynShape, builder.getIndexType());
     Type t4 = builder.getIndexType();
     auto res =
         builder.create<ExpandOp>(loc, TypeRange({t1, t2, t3, t4}), tensor);
     assert(res.getNumResults() == 4);
     assert(!codegen.expValues);
     codegen.expValues = res.getResult(0);
     codegen.expFilled = res.getResult(1);
     codegen.expAdded = res.getResult(2);
     codegen.expCount = res.getResult(3);
   } else {
     assert(codegen.expValues);
     SmallVector<Value, 4> indices;
     for (unsigned i = 0; i < at; i++) {
       assert(codegen.loopEmitter.getLoopIV(i));
       indices.push_back(codegen.loopEmitter.getLoopIV(i));
     }
-    builder.create<CompressOp>(loc, codegen.expValues, codegen.expFilled,
-                               codegen.expAdded, codegen.expCount, tensor,
-                               indices);
+    codegen.insChain = builder.create<CompressOp>(
+        loc, codegen.expValues, codegen.expFilled, codegen.expAdded,
+        codegen.expCount, codegen.insChain, indices);
     codegen.expValues = codegen.expFilled = codegen.expAdded =
         codegen.expCount = Value();
   }
 }
 
 /// Returns parallelization strategy. Any implicit loop in the Linalg
 /// operation that is marked "parallel" is a candidate. Whether it is actually
 /// converted to a parallel operation depends on the requested strategy.
static Operation *genFor(Merger &merger, CodeGen &codegen, OpBuilder &builder,
   Location loc = op.getLoc();
   auto iteratorTypes = op.getIteratorTypesArray();
   bool isReduction = linalg::isReductionIterator(iteratorTypes[idx]);
   bool isSparse = isCompressedDLT(merger.getDimLevelType(tid, idx)) ||
                   isSingletonDLT(merger.getDimLevelType(tid, idx));
   bool isParallel = isParallelFor(codegen, isOuter, isReduction, isSparse);
   assert(!isParallel);
 
-  // Emit a sequential or vector loop.
+  // Emit a sequential for loop.
   SmallVector<Value, 4> operands;
   if (codegen.redVal)
     operands.push_back(codegen.redVal);
   if (codegen.expValues)
     operands.push_back(codegen.expCount);
+  if (codegen.insChain)
+    operands.push_back(codegen.insChain);
 
   Operation *loop = codegen.loopEmitter.enterLoopOverTensorAtDim(
       builder, loc, tid, dim, operands, isParallel, extraTids, extraDims);
 
-  // The operands should be updated by loop emitter already.
+  unsigned o = 0;
   if (codegen.redVal)
-    updateReduc(merger, codegen, operands.front());
+    updateReduc(merger, codegen, operands[o++]);
   if (codegen.expValues)
-    codegen.expCount = operands.back();
+    codegen.expCount = operands[o++];
+  if (codegen.insChain)
+    codegen.insChain = operands[o++];
+  assert(o == operands.size());
 
   return loop;
 }
 
 /// Emit a while-loop for co-iteration over multiple indices.
 static Operation *genWhile(Merger &merger, CodeGen &codegen, OpBuilder &builder,
                            linalg::GenericOp op, unsigned idx, bool needsUniv,
                            ArrayRef<size_t> condTids, ArrayRef<size_t> condDims,
                            ArrayRef<size_t> extraTids,
                            ArrayRef<size_t> extraDims) {
-
   SmallVector<Value, 4> operands;
 
   // Construct the while-loop with a parameter for each index.
   if (codegen.redVal)
     operands.push_back(codegen.redVal);
   if (codegen.expValues)
     operands.push_back(codegen.expCount);
+  if (codegen.insChain)
+    operands.push_back(codegen.insChain);
 
   Operation *loop = codegen.loopEmitter.enterCoIterationOverTensorsAtDims(
       builder, op.getLoc(), condTids, condDims, needsUniv, operands, extraTids,
       extraDims);
 
+  unsigned o = 0;
   if (codegen.redVal)
-    updateReduc(merger, codegen, operands.front());
+    updateReduc(merger, codegen, operands[o++]);
   if (codegen.expValues)
-    codegen.expCount = operands.back();
+    codegen.expCount = operands[o++];
+  if (codegen.insChain)
+    codegen.insChain = operands[o++];
+  assert(o == operands.size());
 
   return loop;
 }
 
 /// Generates a for-loop or a while-loop, depending on whether it implements
 /// singleton iteration or co-iteration over the given conjunction.
 static Operation *genLoop(Merger &merger, CodeGen &codegen, OpBuilder &builder,
                           linalg::GenericOp op, unsigned at, bool needsUniv,

 
 /// Generates the induction structure for a while-loop.
 static void finalizeWhileOp(Merger &merger, CodeGen &codegen,
                             OpBuilder &builder, linalg::GenericOp op,
                             unsigned idx, bool needsUniv, BitVector &induction,
                             scf::WhileOp whileOp) {
   Location loc = op.getLoc();
   // Finalize each else branch of all if statements.
-  if (codegen.redVal || codegen.expValues) {
+  if (codegen.redVal || codegen.expValues || codegen.insChain) {
     while (auto ifOp = dyn_cast_or_null<scf::IfOp>(
                builder.getInsertionBlock()->getParentOp())) {
       unsigned y = 0;
       SmallVector<Value, 4> yields;
       if (codegen.redVal) {
         yields.push_back(codegen.redVal);
         updateReduc(merger, codegen, ifOp.getResult(y++));
       }
       if (codegen.expValues) {
         yields.push_back(codegen.expCount);
         codegen.expCount = ifOp->getResult(y++);
       }
+      if (codegen.insChain) {
+        yields.push_back(codegen.insChain);
+        codegen.insChain = ifOp->getResult(y++);
+      }
       assert(y == yields.size());
       builder.create<scf::YieldOp>(loc, yields);
       builder.setInsertionPointAfter(ifOp);
     }
   }
   builder.setInsertionPointToEnd(&whileOp.getAfter().front());
 }
 
if (isCompressedDLT(merger.getDimLevelType(b)) ||
       clause = constantI1(builder, loc, true);
     }
     cond = cond ? builder.create<arith::AndIOp>(loc, cond, clause) : clause;
   }
   if (codegen.redVal)
     types.push_back(codegen.redVal.getType());
   if (codegen.expValues)
     types.push_back(builder.getIndexType());
+  if (codegen.insChain)
+    types.push_back(codegen.insChain.getType());
   scf::IfOp ifOp = builder.create<scf::IfOp>(loc, types, cond, /*else=*/true);
   builder.setInsertionPointToStart(&ifOp.getThenRegion().front());
   return ifOp;
 }
 
 /// Generates end of true branch of if-statement within a while-loop.
 static void endIf(Merger &merger, CodeGen &codegen, OpBuilder &builder,
                   linalg::GenericOp op, scf::IfOp ifOp, Operation *loop,
-                  Value redInput, Value cntInput) {
+                  Value redInput, Value cntInput, Value insInput) {
   SmallVector<Value, 4> operands;
   if (codegen.redVal) {
     operands.push_back(codegen.redVal);
     updateReduc(merger, codegen, redInput);
   }
   if (codegen.expValues) {
     operands.push_back(codegen.expCount);
     codegen.expCount = cntInput;
   }
+  if (codegen.insChain) {
+    operands.push_back(codegen.insChain);
+    codegen.insChain = insInput;
+  }
   if (!operands.empty())
     builder.create<scf::YieldOp>(op.getLoc(), operands);
   builder.setInsertionPointToStart(&ifOp.getElseRegion().front());
 }
 
 //===----------------------------------------------------------------------===//
 // Sparse compiler synthesis methods (loop sequence).
 //===----------------------------------------------------------------------===//
if (auto whileOp = dyn_cast<scf::WhileOp>(loop)) {
     needsUniv = false;
   }
 
   SmallVector<Value, 2> reduc;
   if (codegen.redVal)
     reduc.push_back(codegen.redVal);
   if (codegen.expValues)
     reduc.push_back(codegen.expCount);
+  if (codegen.insChain)
+    reduc.push_back(codegen.insChain);
 
   auto loopRet =
       codegen.loopEmitter.exitCurrentLoop(builder, op.getLoc(), reduc);
   assert(reduc.size() == loopRet.size());
 
+  unsigned o = 0;
   if (codegen.redVal)
-    updateReduc(merger, codegen, loopRet.front());
+    updateReduc(merger, codegen, loopRet[o++]);
   if (codegen.expValues)
-    codegen.expCount = loopRet.back();
+    codegen.expCount = loopRet[o++];
+  if (codegen.insChain)
+    codegen.insChain = loopRet[o++];
+  assert(o == loopRet.size());
 
   return needsUniv;
 }
 
 /// Ends a loop sequence at given level.
 static void endLoopSeq(Merger &merger, CodeGen &codegen, OpBuilder &builder,
                        linalg::GenericOp op, unsigned exp, unsigned at,
                        unsigned idx, unsigned ldx) {
if (at == codegen.topSort.size()) {
     return;
   }
 
   // Construct iteration lattices for current loop index, with L0 at top.
   unsigned idx = codegen.topSort[at];
   unsigned ldx = at == 0 ? -1u : codegen.topSort[at - 1];
   unsigned lts = merger.optimizeSet(merger.buildLattices(exp, idx));
 
+  // TODO: sort
+  // TODO: dedup
+
   // Start a loop sequence.
   bool needsUniv =
       startLoopSeq(merger, codegen, rewriter, op, exp, at, idx, ldx, lts);
 
   // Emit a loop for every lattice point L0 >= Li in this loop sequence.
   unsigned lsize = merger.set(lts).size();
   for (unsigned i = 0; i < lsize; i++) {
     // Start a loop.
     unsigned li = merger.set(lts)[i];
     Operation *loop =
         startLoop(merger, codegen, rewriter, op, at, li, needsUniv);
 
     // Visit all lattices points with Li >= Lj to generate the
     // loop-body, possibly with if statements for coiteration.
     Value redInput = codegen.redVal;
     Value cntInput = codegen.expCount;
+    Value insInput = codegen.insChain;
     bool isWhile = dyn_cast<scf::WhileOp>(loop) != nullptr;
     for (unsigned j = 0; j < lsize; j++) {
       unsigned lj = merger.set(lts)[j];
       unsigned ej = merger.lat(lj).exp;
       if (li == lj || merger.latGT(li, lj)) {
         // Recurse into body of each branch.
         if (isWhile) {
           scf::IfOp ifOp =
               genIf(merger, codegen, rewriter, op, idx, merger.lat(lj).simple);
           genStmt(merger, codegen, rewriter, op, ej, at + 1);
-          endIf(merger, codegen, rewriter, op, ifOp, loop, redInput, cntInput);
+          endIf(merger, codegen, rewriter, op, ifOp, loop, redInput, cntInput,
+                insInput);
         } else {
           genStmt(merger, codegen, rewriter, op, ej, at + 1);
         }
       }
     }
 
     // End a loop.
     needsUniv =
         endLoop(merger, codegen, rewriter, op, loop, idx, li, needsUniv);
   }
 
   // End a loop sequence.
   endLoopSeq(merger, codegen, rewriter, op, exp, at, idx, ldx);
 }
 
 /// Converts the result computed by the sparse kernel into the required form.
 static void genResult(Merger &merger, CodeGen &codegen, RewriterBase &rewriter,
                       linalg::GenericOp op) {
   OpOperand *lhs = op.getOutputOperand(0);
-  Type resType = lhs->get().getType();
+  Value tensor = lhs->get();
+  Type resType = tensor.getType();
   if (getSparseTensorEncoding(resType)) {
     // The sparse tensor rematerializes from the original sparse tensor's
-    // underlying sparse storage format.
-    rewriter.replaceOpWithNewOp<LoadOp>(op, resType, lhs->get(),
-                                        codegen.sparseOut == lhs);
+    // underlying sparse storage format. For an insertion chain, the
+    // tensor materializes from the chain with 'hasInserts' enabled.
+    bool hasInserts = codegen.sparseOut == lhs;
+    if (hasInserts)
+      tensor = codegen.insChain;
+    rewriter.replaceOpWithNewOp<LoadOp>(op, resType, tensor, hasInserts);
   } else {
     // To rematerialize an non-annotated tensor, simply load it
     // from the bufferized value.
     Value val = codegen.loopEmitter.getValBuffer().back(); // value array
     rewriter.replaceOpWithNewOp<bufferization::ToTensorOp>(op, resType, val);
   }
 }
 

mlir/test/Dialect/SparseTensor/sparse_broadcast.mlir

 // RUN: mlir-opt %s --sparsification --canonicalize --cse | FileCheck %s
 
 #DCSR = #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>
 #SparseTensor = #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ] }>
 
 #trait = {
   indexing_maps = [
     affine_map<(d0, d1, d2) -> (d0, d2)>,
     affine_map<(d0, d1, d2) -> (d0, d1, d2)>
   ],
   iterator_types = ["parallel", "parallel", "parallel"]
 }
 
 // CHECK-LABEL:  @main(
-//  CHECK-SAME:    %[[TMP_arg0:.*]]: tensor<4x5xi32, 
+//  CHECK-SAME:  %[[TMP_arg0:.*]]: tensor<4x5xi32,
 //   CHECK-DAG:  %[[TMP_c3:.*]] = arith.constant 3 : index
 //   CHECK-DAG:  %[[TMP_c0:.*]] = arith.constant 0 : index
 //   CHECK-DAG:  %[[TMP_c1:.*]] = arith.constant 1 : index
 //       CHECK:  %[[TMP_0:.*]] = bufferization.alloc_tensor()
 //       CHECK:  %[[TMP_1:.*]] = sparse_tensor.pointers %[[TMP_arg0]] {dimension = 0 : index}
 //       CHECK:  %[[TMP_2:.*]] = sparse_tensor.indices %[[TMP_arg0]] {dimension = 0 : index}
 //       CHECK:  %[[TMP_3:.*]] = sparse_tensor.pointers %[[TMP_arg0]] {dimension = 1 : index}
 //       CHECK:  %[[TMP_4:.*]] = sparse_tensor.indices %[[TMP_arg0]] {dimension = 1 : index}
 //       CHECK:  %[[TMP_5:.*]] = sparse_tensor.values %[[TMP_arg0]]
 //       CHECK:  %[[TMP_6:.*]] = memref.load %[[TMP_1]][%[[TMP_c0]]] : memref<?xindex>
 //       CHECK:  %[[TMP_7:.*]] = memref.load %[[TMP_1]][%[[TMP_c1]]] : memref<?xindex>
-//       CHECK:  scf.for %[[TMP_arg1:.*]] = %[[TMP_6]] to %[[TMP_7]] step %[[TMP_c1]] {
+//       CHECK:  %[[T:.*]] = scf.for %[[TMP_arg1:.*]] = %[[TMP_6]] to %[[TMP_7]] step %[[TMP_c1]] {{.*}} {
 //       CHECK:    %[[TMP_9:.*]] = memref.load %[[TMP_2]][%[[TMP_arg1]]] : memref<?xindex>
-//       CHECK:    scf.for %[[TMP_arg2:.*]] = %[[TMP_c0]] to %[[TMP_c3]] step %[[TMP_c1]] {
+//       CHECK:    %[[L1:.*]] = scf.for %[[TMP_arg2:.*]] = %[[TMP_c0]] to %[[TMP_c3]] step %[[TMP_c1]] {{.*}} {
 //       CHECK:      %[[TMP_10:.*]] = memref.load %[[TMP_3]][%[[TMP_arg1]]] : memref<?xindex>
 //       CHECK:      %[[TMP_11:.*]] = arith.addi %[[TMP_arg1]], %[[TMP_c1]] : index
 //       CHECK:      %[[TMP_12:.*]] = memref.load %[[TMP_3]][%[[TMP_11]]] : memref<?xindex>
-//       CHECK:      scf.for %[[TMP_arg3:.*]] = %[[TMP_10]] to %[[TMP_12]] step %[[TMP_c1]] {
+//       CHECK:      %[[L2:.*]] = scf.for %[[TMP_arg3:.*]] = %[[TMP_10]] to %[[TMP_12]] step %[[TMP_c1]] {{.*}} {
 //       CHECK:        %[[TMP_13:.*]] = memref.load %[[TMP_4]][%[[TMP_arg3]]] : memref<?xindex>
 //       CHECK:        %[[TMP_14:.*]] = memref.load %[[TMP_5]][%[[TMP_arg3]]] : memref<?xi32>
-//       CHECK:        %[[TMP_15:.*]] = sparse_tensor.insert %[[TMP_14]] into %[[TMP_0]][%[[TMP_9]], %[[TMP_arg2]], %[[TMP_13]]]
+//       CHECK:        %[[Y:.*]] = sparse_tensor.insert %[[TMP_14]] into %{{.*}}[%[[TMP_9]], %[[TMP_arg2]], %[[TMP_13]]]
+//       CHECK:        scf.yield %[[Y]]
 //       CHECK:      }
+//       CHECK:      scf.yield %[[L2]]
 //       CHECK:    }
+//       CHECK:    scf.yield %[[L1]]
 //       CHECK:  }
-//       CHECK:  %[[TMP_8:.*]] = sparse_tensor.load %[[TMP_0]] hasInserts
-//       CHECK:  return %[[TMP_8]] 
+//       CHECK:  %[[TMP_8:.*]] = sparse_tensor.load %[[T]] hasInserts
+//       CHECK:  return %[[TMP_8]]
 module @func_sparse {
   func.func public @main(%arg0: tensor<4x5xi32, #DCSR>) -> tensor<4x3x5xi32, #SparseTensor> {
     %0 = bufferization.alloc_tensor() : tensor<4x3x5xi32, #SparseTensor>
     %1 = linalg.generic #trait
     ins(%arg0 : tensor<4x5xi32, #DCSR>) outs(%0 : tensor<4x3x5xi32, #SparseTensor>) {
     ^bb0(%in: i32, %out: i32):
       linalg.yield %in : i32
     } -> tensor<4x3x5xi32, #SparseTensor>
     return %1 : tensor<4x3x5xi32, #SparseTensor>
   }
 }

mlir/test/Dialect/SparseTensor/sparse_expand.mlir

func.func @kernel(%arga: tensor<?x?xf64, #DCSC>) -> tensor<?xf64, #SV> {
   return %0 : tensor<?xf64, #SV>
 }
 
 //
 // CHECK-SPARSE-LABEL: func @matmul1(
 // CHECK-SPARSE-DAG: %[[C0:.*]] = arith.constant 0 : index
 // CHECK-SPARSE-DAG: %[[C1:.*]] = arith.constant 1 : index
 // CHECK-SPARSE-DAG: %[[C8:.*]] = arith.constant 8 : index
-// CHECK-SPARSE: scf.for %{{.*}} = %[[C0]] to %[[C8]] step %[[C1]] {
+// CHECK-SPARSE: %[[T:.*]] = scf.for %{{.*}} = %[[C0]] to %[[C8]] step %[[C1]] {{.*}} {
 // CHECK-SPARSE:   %[[A:.*]], %[[B:.*]], %[[C:.*]], %{{.*}} = sparse_tensor.expand
 // CHECK-SPARSE:   %[[COUNT:.*]] = scf.for {{.*}} {
 // CHECK-SPARSE:     scf.for {{.*}} {
 // CHECK-SPARSE:     }
 // CHECK-SPARSE:   }
 // CHECK-SPARSE:   sparse_tensor.compress %[[A]], %[[B]], %[[C]], %[[COUNT]] into
 // CHECK-SPARSE: }
-// CHECK-SPARSE: %[[RET:.*]] = sparse_tensor.load %{{.*}} hasInserts
+// CHECK-SPARSE: %[[RET:.*]] = sparse_tensor.load %[[T]] hasInserts
 // CHECK-SPARSE: return %[[RET]]
 //
 // CHECK-CONVERT-LABEL: func @matmul1(
 // CHECK-CONVERT-DAG: %[[C0:.*]] = arith.constant 0 : index
 // CHECK-CONVERT-DAG: %[[C1:.*]] = arith.constant 1 : index
 // CHECK-CONVERT-DAG: %[[C4:.*]] = arith.constant 4 : index
 // CHECK-CONVERT-DAG: %[[C8:.*]] = arith.constant 8 : index
 // CHECK-CONVERT: %[[N:.*]] = call @newSparseTensor
 // CHECK-CONVERT: %[[A:.*]] = memref.alloc(%[[C4]]) : memref<?xf64>
 // CHECK-CONVERT: %[[B:.*]] = memref.alloc(%[[C4]]) : memref<?xi1>
 // CHECK-CONVERT: %[[C:.*]] = memref.alloc(%[[C4]]) : memref<?xindex>
 // CHECK-CONVERT: linalg.fill ins(%{{.*}} : f64) outs(%[[A]] : memref<?xf64>)
 // CHECK-CONVERT: linalg.fill ins(%{{.*}} : i1) outs(%[[B]] : memref<?xi1>)
-// CHECK-CONVERT: scf.for %{{.*}} = %[[C0]] to %[[C8]] step %[[C1]] {
+// CHECK-CONVERT: scf.for %{{.*}} = %[[C0]] to %[[C8]] step %[[C1]] {{.*}} {
 // CHECK-CONVERT:   scf.for {{.*}} {
 // CHECK-CONVERT:     scf.for {{.*}} {
 // CHECK-CONVERT:     }
 // CHECK-CONVERT:   }
 // CHECK-CONVERT:   call @expInsertF64
 // CHECK-CONVERT: }
 // CHECK-CONVERT: memref.dealloc %[[A]] : memref<?xf64>
 // CHECK-CONVERT: memref.dealloc %[[B]] : memref<?xi1>
func.func @matmul1(%A: tensor<8x2xf64, #CSR>,
   return %D: tensor<8x4xf64, #CSR>
 }
 
 //
 // CHECK-SPARSE-LABEL: func @matmul2(
 // CHECK-SPARSE-DAG: %[[C0:.*]] = arith.constant 0 : index
 // CHECK-SPARSE-DAG: %[[C1:.*]] = arith.constant 1 : index
 // CHECK-SPARSE-DAG: %[[C4:.*]] = arith.constant 4 : index
-// CHECK-SPARSE: scf.for %{{.*}} = %[[C0]] to %[[C4]] step %[[C1]] {
+// CHECK-SPARSE: %[[T:.*]] = scf.for %{{.*}} = %[[C0]] to %[[C4]] step %[[C1]] {{.*}} {
 // CHECK-SPARSE:   %[[A:.*]], %[[B:.*]], %[[C:.*]], %{{.*}} = sparse_tensor.expand
 // CHECK-SPARSE:   %[[COUNT:.*]] = scf.for {{.*}} {
 // CHECK-SPARSE:     scf.for {{.*}} {
 // CHECK-SPARSE:     }
 // CHECK-SPARSE:   }
 // CHECK-SPARSE:   sparse_tensor.compress %[[A]], %[[B]], %[[C]], %[[COUNT]]
 // CHECK-SPARSE: }
-// CHECK-SPARSE: %[[RET:.*]] = sparse_tensor.load %{{.*}} hasInserts
+// CHECK-SPARSE: %[[RET:.*]] = sparse_tensor.load %[[T]] hasInserts
 // CHECK-SPARSE: return %[[RET]]
 //
 // CHECK-CONVERT-LABEL: func @matmul2(
 // CHECK-CONVERT-DAG: %[[C0:.*]] = arith.constant 0 : index
 // CHECK-CONVERT-DAG: %[[C1:.*]] = arith.constant 1 : index
 // CHECK-CONVERT-DAG: %[[C4:.*]] = arith.constant 4 : index
 // CHECK-CONVERT-DAG: %[[C8:.*]] = arith.constant 8 : index
 // CHECK-CONVERT: %[[N:.*]] = call @newSparseTensor
 // CHECK-CONVERT: %[[A:.*]] = memref.alloc(%[[C8]]) : memref<?xf64>
 // CHECK-CONVERT: %[[B:.*]] = memref.alloc(%[[C8]]) : memref<?xi1>
 // CHECK-CONVERT: %[[C:.*]] = memref.alloc(%[[C8]]) : memref<?xindex>
 // CHECK-CONVERT: linalg.fill ins(%{{.*}} : f64) outs(%[[A]] : memref<?xf64>)
 // CHECK-CONVERT: linalg.fill ins(%{{.*}} : i1) outs(%[[B]] : memref<?xi1>)
-// CHECK-CONVERT: scf.for %{{.*}} = %[[C0]] to %[[C4]] step %[[C1]] {
+// CHECK-CONVERT: scf.for %{{.*}} = %[[C0]] to %[[C4]] step %[[C1]] {{.*}} {
 // CHECK-CONVERT:   scf.for {{.*}} {
 // CHECK-CONVERT:     scf.for {{.*}} {
 // CHECK-CONVERT:     }
 // CHECK-CONVERT:   }
 // CHECK-CONVERT:   call @expInsertF64
 // CHECK-CONVERT: }
 // CHECK-CONVERT: memref.dealloc %[[A]] : memref<?xf64>
 // CHECK-CONVERT: memref.dealloc %[[B]] : memref<?xi1>

mlir/test/Dialect/SparseTensor/sparse_fp_ops.mlir

 // CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 1 : index
 // CHECK:           %[[VAL_3:.*]] = bufferization.alloc_tensor() : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
 // CHECK:           %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_5:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xf64>
 // CHECK:           %[[VAL_7:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_1]]] : memref<?xindex>
 // CHECK:           %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_9:.*]] = %[[VAL_7]] to %[[VAL_8]] step %[[VAL_2]] {
+// CHECK:           %[[T:.*]] = scf.for %[[VAL_9:.*]] = %[[VAL_7]] to %[[VAL_8]] step %[[VAL_2]] {{.*}} {
 // CHECK:             %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_9]]] : memref<?xindex>
 // CHECK:             %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_9]]] : memref<?xf64>
 // CHECK:             %[[VAL_12:.*]] = math.absf %[[VAL_11]] : f64
 // CHECK:             %[[VAL_13:.*]] = math.ceil %[[VAL_12]] : f64
 // CHECK:             %[[VAL_14:.*]] = math.floor %[[VAL_13]] : f64
 // CHECK:             %[[VAL_15:.*]] = math.sqrt %[[VAL_14]] : f64
 // CHECK:             %[[VAL_16:.*]] = math.expm1 %[[VAL_15]] : f64
 // CHECK:             %[[VAL_17:.*]] = math.log1p %[[VAL_16]] : f64
 // CHECK:             %[[VAL_18:.*]] = math.sin %[[VAL_17]] : f64
 // CHECK:             %[[VAL_19:.*]] = math.tanh %[[VAL_18]] : f64
-// CHECK:             sparse_tensor.insert %[[VAL_19]] into %[[VAL_3]]{{\[}}%[[VAL_10]]] : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
+// CHECK:             %[[Y:.*]] = sparse_tensor.insert %[[VAL_19]] into %{{.*}}[%[[VAL_10]]] : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
+// CHECK:             scf.yield %[[Y]]
 // CHECK:           }
-// CHECK:           %[[VAL_20:.*]] = sparse_tensor.load %[[VAL_3]] hasInserts : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
+// CHECK:           %[[VAL_20:.*]] = sparse_tensor.load %[[T]] hasInserts : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
 // CHECK:           return %[[VAL_20]] : tensor<32xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
 // CHECK:         }
 func.func @zero_preserving_math(%arga: tensor<32xf64, #SV>) -> tensor<32xf64, #SV> {
   %c32 = arith.constant 32 : index
   %xinp = bufferization.alloc_tensor() : tensor<32xf64, #SV>
   %0 = linalg.generic #trait1
      ins(%arga: tensor<32xf64, #SV>)
     outs(%xinp: tensor<32xf64, #SV>) {

 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 1 : index
 // CHECK:           %[[VAL_3:.*]] = complex.constant [0.000000e+00, 1.000000e+00] : complex<f64>
 // CHECK:           %[[VAL_4:.*]] = bufferization.alloc_tensor() : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
 // CHECK:           %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xcomplex<f64>>
 // CHECK:           %[[VAL_8:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_1]]] : memref<?xindex>
 // CHECK:           %[[VAL_9:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_2]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_2]] {
+// CHECK:           %[[T:.*]] = scf.for %[[VAL_10:.*]] = %[[VAL_8]] to %[[VAL_9]] step %[[VAL_2]] {{.*}} {
 // CHECK:             %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xindex>
 // CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_10]]] : memref<?xcomplex<f64>>
 // CHECK:             %[[VAL_13:.*]] = complex.div %[[VAL_12]], %[[VAL_3]] : complex<f64>
-// CHECK:             sparse_tensor.insert %[[VAL_13]] into %[[VAL_4]]{{\[}}%[[VAL_11]]] : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
+// CHECK:             %[[Y:.*]] = sparse_tensor.insert %[[VAL_13]] into %{{.*}}[%[[VAL_11]]] : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
+// CHECK:             scf.yield %[[Y]]
 // CHECK:           }
-// CHECK:           %[[VAL_14:.*]] = sparse_tensor.load %[[VAL_4]] hasInserts : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
+// CHECK:           %[[VAL_14:.*]] = sparse_tensor.load %[[T]] hasInserts : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
 // CHECK:           return %[[VAL_14]] : tensor<32xcomplex<f64>, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>
 // CHECK:         }
 func.func @complex_divbyc(%arg0: tensor<32xcomplex<f64>, #SV>) -> tensor<32xcomplex<f64>, #SV> {
   %c = complex.constant [0.0, 1.0] : complex<f64>
   %init = bufferization.alloc_tensor() : tensor<32xcomplex<f64>, #SV>
   %0 = linalg.generic #traitc
      ins(%arg0: tensor<32xcomplex<f64>, #SV>)
     outs(%init: tensor<32xcomplex<f64>, #SV>) {
       ^bb(%a: complex<f64>, %x: complex<f64>):
         %0 = complex.div %a, %c : complex<f64>
         linalg.yield %0 : complex<f64>
   } -> tensor<32xcomplex<f64>, #SV>
   return %0 : tensor<32xcomplex<f64>, #SV>
 }

mlir/test/Dialect/SparseTensor/sparse_index.mlir

 // CHECK-DAG:       %[[VAL_5:.*]] = bufferization.alloc_tensor(%[[VAL_3]], %[[VAL_4]]) : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK:           %[[VAL_11:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_1]]] : memref<?xindex>
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_2]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_13:.*]] = %[[VAL_11]] to %[[VAL_12]] step %[[VAL_2]] {
+// CHECK:           %[[T:.*]] = scf.for %[[VAL_13:.*]] = %[[VAL_11]] to %[[VAL_12]] step %[[VAL_2]] {{.*}} {
 // CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_13]]] : memref<?xindex>
 // CHECK:             %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_13]]] : memref<?xindex>
 // CHECK:             %[[VAL_16:.*]] = arith.addi %[[VAL_13]], %[[VAL_2]] : index
 // CHECK:             %[[VAL_17:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_16]]] : memref<?xindex>
-// CHECK:             scf.for %[[VAL_18:.*]] = %[[VAL_15]] to %[[VAL_17]] step %[[VAL_2]] {
+// CHECK:             %[[L:.*]] = scf.for %[[VAL_18:.*]] = %[[VAL_15]] to %[[VAL_17]] step %[[VAL_2]] {{.*}} {
 // CHECK:               %[[VAL_19:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_18]]] : memref<?xindex>
 // CHECK:               %[[VAL_20:.*]] = arith.index_cast %[[VAL_19]] : index to i64
 // CHECK:               %[[VAL_21:.*]] = arith.index_cast %[[VAL_14]] : index to i64
 // CHECK:               %[[VAL_22:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_18]]] : memref<?xi64>
 // CHECK:               %[[VAL_23:.*]] = arith.muli %[[VAL_21]], %[[VAL_22]] : i64
 // CHECK:               %[[VAL_24:.*]] = arith.muli %[[VAL_20]], %[[VAL_23]] : i64
-// CHECK:               sparse_tensor.insert %[[VAL_24]] into %[[VAL_5]]{{\[}}%[[VAL_14]], %[[VAL_19]]] : tensor<?x?xi64, #sparse_tensor.encoding
+// CHECK:               %[[Y:.*]] = sparse_tensor.insert %[[VAL_24]] into %{{.*}}[%[[VAL_14]], %[[VAL_19]]] : tensor<?x?xi64, #sparse_tensor.encoding
+// CHECK:               scf.yield %[[Y]]
 // CHECK:             }
+// CHECK:             scf.yield %[[L]]
 // CHECK:           }
-// CHECK:           %[[VAL_25:.*]] = sparse_tensor.load %[[VAL_5]] hasInserts : tensor<?x?xi64, #sparse_tensor.encoding
+// CHECK:           %[[VAL_25:.*]] = sparse_tensor.load %[[T]] hasInserts : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK:           return %[[VAL_25]] : tensor<?x?xi64, #sparse_tensor.encoding
 // CHECK:         }
 func.func @sparse_index(%arga: tensor<?x?xi64, #SparseMatrix>)
                        -> tensor<?x?xi64, #SparseMatrix> {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 0 : index
   %0 = tensor.dim %arga, %c0 : tensor<?x?xi64, #SparseMatrix>
   %1 = tensor.dim %arga, %c1 : tensor<?x?xi64, #SparseMatrix>

mlir/test/Dialect/SparseTensor/sparse_kernels.mlir

 // RUN: mlir-opt %s \
 // RUN: --linalg-generalize-named-ops --linalg-fuse-elementwise-ops \
 // RUN: --sparsification | FileCheck %s
 
 #SparseVector = #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>
 
 #DCSR = #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>
 
 // CHECK-LABEL:   func.func @matmul1(
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<20x30xf32>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<10x30xf32>) -> tensor<10x30xf32> {
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 30 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>
+// CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>
 // CHECK:           %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<20x30xf32>
 // CHECK:           %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<10x30xf32>
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_15:.*]] = %[[VAL_13]] to %[[VAL_14]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_16:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_15]]] : memref<?xindex>
 // CHECK:             %[[VAL_17:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_15]]] : memref<?xindex>
 // CHECK:             %[[VAL_18:.*]] = arith.addi %[[VAL_15]], %[[VAL_5]] : index

 //
 // CHECK-LABEL:   func.func @matmul2(
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant false
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant true
-// CHECK:           %[[VAL_6:.*]] = bufferization.alloc_tensor() : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>
-// CHECK:           %[[VAL_12:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_13:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_14:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_15:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_16:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>
+// CHECK-DAG:       %[[VAL_6:.*]] = bufferization.alloc_tensor() : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<4x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>
+// CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_15:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_16:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<8x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>
 // CHECK:           %[[VAL_17:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_2]]] : memref<?xindex>
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_3]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_19:.*]] = %[[VAL_17]] to %[[VAL_18]] step %[[VAL_3]] {
-// CHECK:             %[[VAL_20:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_19]]] : memref<?xindex>
-// CHECK:             %[[VAL_21:.*]], %[[VAL_22:.*]], %[[VAL_23:.*]], %[[VAL_24:.*]] = sparse_tensor.expand %[[VAL_6]] : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>, memref<?xi1>, memref<?xindex>
-// CHECK:             %[[VAL_25:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_19]]] : memref<?xindex>
-// CHECK:             %[[VAL_26:.*]] = arith.addi %[[VAL_19]], %[[VAL_3]] : index
-// CHECK:             %[[VAL_27:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_26]]] : memref<?xindex>
-// CHECK:             %[[VAL_28:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_2]]] : memref<?xindex>
-// CHECK:             %[[VAL_29:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_3]]] : memref<?xindex>
-// CHECK:             %[[VAL_30:.*]]:3 = scf.while (%[[VAL_31:.*]] = %[[VAL_25]], %[[VAL_32:.*]] = %[[VAL_28]], %[[VAL_33:.*]] = %[[VAL_24]]) : (index, index, index) -> (index, index, index) {
-// CHECK:               %[[VAL_34:.*]] = arith.cmpi ult, %[[VAL_31]], %[[VAL_27]] : index
-// CHECK:               %[[VAL_35:.*]] = arith.cmpi ult, %[[VAL_32]], %[[VAL_29]] : index
-// CHECK:               %[[VAL_36:.*]] = arith.andi %[[VAL_34]], %[[VAL_35]] : i1
-// CHECK:               scf.condition(%[[VAL_36]]) %[[VAL_31]], %[[VAL_32]], %[[VAL_33]] : index, index, index
+// CHECK:           %[[VAL_19:.*]] = scf.for %[[VAL_20:.*]] = %[[VAL_17]] to %[[VAL_18]] step %[[VAL_3]] iter_args(%[[VAL_21:.*]] = %[[VAL_6]]) -> (tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:             %[[VAL_22:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_20]]] : memref<?xindex>
+// CHECK:             %[[VAL_23:.*]], %[[VAL_24:.*]], %[[VAL_25:.*]], %[[VAL_26:.*]] = sparse_tensor.expand %[[VAL_6]] : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>, memref<?xi1>, memref<?xindex>
+// CHECK:             %[[VAL_27:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_20]]] : memref<?xindex>
+// CHECK:             %[[VAL_28:.*]] = arith.addi %[[VAL_20]], %[[VAL_3]] : index
+// CHECK:             %[[VAL_29:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_28]]] : memref<?xindex>
+// CHECK:             %[[VAL_30:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_2]]] : memref<?xindex>
+// CHECK:             %[[VAL_31:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_3]]] : memref<?xindex>
+// CHECK:             %[[VAL_32:.*]]:4 = scf.while (%[[VAL_33:.*]] = %[[VAL_27]], %[[VAL_34:.*]] = %[[VAL_30]], %[[VAL_35:.*]] = %[[VAL_26]], %[[VAL_36:.*]] = %[[VAL_21]]) : (index, index, index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> (index, index, index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:               %[[VAL_37:.*]] = arith.cmpi ult, %[[VAL_33]], %[[VAL_29]] : index
+// CHECK:               %[[VAL_38:.*]] = arith.cmpi ult, %[[VAL_34]], %[[VAL_31]] : index
+// CHECK:               %[[VAL_39:.*]] = arith.andi %[[VAL_37]], %[[VAL_38]] : i1
+// CHECK:               scf.condition(%[[VAL_39]]) %[[VAL_33]], %[[VAL_34]], %[[VAL_35]], %[[VAL_36]] : index, index, index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             } do {
-// CHECK:             ^bb0(%[[VAL_37:.*]]: index, %[[VAL_38:.*]]: index, %[[VAL_39:.*]]: index):
-// CHECK:               %[[VAL_40:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_37]]] : memref<?xindex>
-// CHECK:               %[[VAL_41:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_38]]] : memref<?xindex>
-// CHECK:               %[[VAL_42:.*]] = arith.cmpi ult, %[[VAL_41]], %[[VAL_40]] : index
-// CHECK:               %[[VAL_43:.*]] = arith.select %[[VAL_42]], %[[VAL_41]], %[[VAL_40]] : index
-// CHECK:               %[[VAL_44:.*]] = arith.cmpi eq, %[[VAL_40]], %[[VAL_43]] : index
-// CHECK:               %[[VAL_45:.*]] = arith.cmpi eq, %[[VAL_41]], %[[VAL_43]] : index
-// CHECK:               %[[VAL_46:.*]] = arith.andi %[[VAL_44]], %[[VAL_45]] : i1
-// CHECK:               %[[VAL_47:.*]] = scf.if %[[VAL_46]] -> (index) {
-// CHECK:                 %[[VAL_48:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_37]]] : memref<?xf64>
-// CHECK:                 %[[VAL_49:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_38]]] : memref<?xindex>
-// CHECK:                 %[[VAL_50:.*]] = arith.addi %[[VAL_38]], %[[VAL_3]] : index
-// CHECK:                 %[[VAL_51:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_50]]] : memref<?xindex>
-// CHECK:                 %[[VAL_52:.*]] = scf.for %[[VAL_53:.*]] = %[[VAL_49]] to %[[VAL_51]] step %[[VAL_3]] iter_args(%[[VAL_54:.*]] = %[[VAL_39]]) -> (index) {
-// CHECK:                   %[[VAL_55:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_53]]] : memref<?xindex>
-// CHECK:                   %[[VAL_56:.*]] = memref.load %[[VAL_21]]{{\[}}%[[VAL_55]]] : memref<?xf64>
-// CHECK:                   %[[VAL_57:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_53]]] : memref<?xf64>
-// CHECK:                   %[[VAL_58:.*]] = arith.mulf %[[VAL_48]], %[[VAL_57]] : f64
-// CHECK:                   %[[VAL_59:.*]] = arith.addf %[[VAL_56]], %[[VAL_58]] : f64
-// CHECK:                   %[[VAL_60:.*]] = memref.load %[[VAL_22]]{{\[}}%[[VAL_55]]] : memref<?xi1>
-// CHECK:                   %[[VAL_61:.*]] = arith.cmpi eq, %[[VAL_60]], %[[VAL_4]] : i1
-// CHECK:                   %[[VAL_62:.*]] = scf.if %[[VAL_61]] -> (index) {
-// CHECK:                     memref.store %[[VAL_5]], %[[VAL_22]]{{\[}}%[[VAL_55]]] : memref<?xi1>
-// CHECK:                     memref.store %[[VAL_55]], %[[VAL_23]]{{\[}}%[[VAL_54]]] : memref<?xindex>
-// CHECK:                     %[[VAL_63:.*]] = arith.addi %[[VAL_54]], %[[VAL_3]] : index
-// CHECK:                     scf.yield %[[VAL_63]] : index
+// CHECK:             ^bb0(%[[VAL_40:.*]]: index, %[[VAL_41:.*]]: index, %[[VAL_42:.*]]: index, %[[VAL_43:.*]]: tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>):
+// CHECK:               %[[VAL_44:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_40]]] : memref<?xindex>
+// CHECK:               %[[VAL_45:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_41]]] : memref<?xindex>
+// CHECK:               %[[VAL_46:.*]] = arith.cmpi ult, %[[VAL_45]], %[[VAL_44]] : index
+// CHECK:               %[[VAL_47:.*]] = arith.select %[[VAL_46]], %[[VAL_45]], %[[VAL_44]] : index
+// CHECK:               %[[VAL_48:.*]] = arith.cmpi eq, %[[VAL_44]], %[[VAL_47]] : index
+// CHECK:               %[[VAL_49:.*]] = arith.cmpi eq, %[[VAL_45]], %[[VAL_47]] : index
+// CHECK:               %[[VAL_50:.*]] = arith.andi %[[VAL_48]], %[[VAL_49]] : i1
+// CHECK:               %[[VAL_51:.*]]:2 = scf.if %[[VAL_50]] -> (index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:                 %[[VAL_52:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_40]]] : memref<?xf64>
+// CHECK:                 %[[VAL_53:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_41]]] : memref<?xindex>
+// CHECK:                 %[[VAL_54:.*]] = arith.addi %[[VAL_41]], %[[VAL_3]] : index
+// CHECK:                 %[[VAL_55:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_54]]] : memref<?xindex>
+// CHECK:                 %[[VAL_56:.*]] = scf.for %[[VAL_57:.*]] = %[[VAL_53]] to %[[VAL_55]] step %[[VAL_3]] iter_args(%[[VAL_58:.*]] = %[[VAL_42]]) -> (index) {
+// CHECK:                   %[[VAL_59:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_57]]] : memref<?xindex>
+// CHECK:                   %[[VAL_60:.*]] = memref.load %[[VAL_23]]{{\[}}%[[VAL_59]]] : memref<?xf64>
+// CHECK:                   %[[VAL_61:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_57]]] : memref<?xf64>
+// CHECK:                   %[[VAL_62:.*]] = arith.mulf %[[VAL_52]], %[[VAL_61]] : f64
+// CHECK:                   %[[VAL_63:.*]] = arith.addf %[[VAL_60]], %[[VAL_62]] : f64
+// CHECK:                   %[[VAL_64:.*]] = memref.load %[[VAL_24]]{{\[}}%[[VAL_59]]] : memref<?xi1>
+// CHECK:                   %[[VAL_65:.*]] = arith.cmpi eq, %[[VAL_64]], %[[VAL_4]] : i1
+// CHECK:                   %[[VAL_66:.*]] = scf.if %[[VAL_65]] -> (index) {
+// CHECK:                     memref.store %[[VAL_5]], %[[VAL_24]]{{\[}}%[[VAL_59]]] : memref<?xi1>
+// CHECK:                     memref.store %[[VAL_59]], %[[VAL_25]]{{\[}}%[[VAL_58]]] : memref<?xindex>
+// CHECK:                     %[[VAL_67:.*]] = arith.addi %[[VAL_58]], %[[VAL_3]] : index
+// CHECK:                     scf.yield %[[VAL_67]] : index
 // CHECK:                   } else {
-// CHECK:                     scf.yield %[[VAL_54]] : index
+// CHECK:                     scf.yield %[[VAL_58]] : index
 // CHECK:                   }
-// CHECK:                   memref.store %[[VAL_59]], %[[VAL_21]]{{\[}}%[[VAL_55]]] : memref<?xf64>
-// CHECK:                   scf.yield %[[VAL_64:.*]] : index
+// CHECK:                   memref.store %[[VAL_63]], %[[VAL_23]]{{\[}}%[[VAL_59]]] : memref<?xf64>
+// CHECK:                   scf.yield %[[VAL_68:.*]] : index
 // CHECK:                 }
-// CHECK:                 scf.yield %[[VAL_65:.*]] : index
+// CHECK:                 scf.yield %[[VAL_69:.*]], %[[VAL_43]] : index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:               } else {
-// CHECK:                 scf.yield %[[VAL_39]] : index
+// CHECK:                 scf.yield %[[VAL_42]], %[[VAL_43]] : index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:               }
-// CHECK:               %[[VAL_66:.*]] = arith.cmpi eq, %[[VAL_40]], %[[VAL_43]] : index
-// CHECK:               %[[VAL_67:.*]] = arith.addi %[[VAL_37]], %[[VAL_3]] : index
-// CHECK:               %[[VAL_68:.*]] = arith.select %[[VAL_66]], %[[VAL_67]], %[[VAL_37]] : index
-// CHECK:               %[[VAL_69:.*]] = arith.cmpi eq, %[[VAL_41]], %[[VAL_43]] : index
-// CHECK:               %[[VAL_70:.*]] = arith.addi %[[VAL_38]], %[[VAL_3]] : index
-// CHECK:               %[[VAL_71:.*]] = arith.select %[[VAL_69]], %[[VAL_70]], %[[VAL_38]] : index
-// CHECK:               scf.yield %[[VAL_68]], %[[VAL_71]], %[[VAL_72:.*]] : index, index, index
+// CHECK:               %[[VAL_70:.*]] = arith.cmpi eq, %[[VAL_44]], %[[VAL_47]] : index
+// CHECK:               %[[VAL_71:.*]] = arith.addi %[[VAL_40]], %[[VAL_3]] : index
+// CHECK:               %[[VAL_72:.*]] = arith.select %[[VAL_70]], %[[VAL_71]], %[[VAL_40]] : index
+// CHECK:               %[[VAL_73:.*]] = arith.cmpi eq, %[[VAL_45]], %[[VAL_47]] : index
+// CHECK:               %[[VAL_74:.*]] = arith.addi %[[VAL_41]], %[[VAL_3]] : index
+// CHECK:               %[[VAL_75:.*]] = arith.select %[[VAL_73]], %[[VAL_74]], %[[VAL_41]] : index
+// CHECK:               scf.yield %[[VAL_72]], %[[VAL_75]], %[[VAL_76:.*]]#0, %[[VAL_76]]#1 : index, index, index, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             }
-// CHECK:             sparse_tensor.compress %[[VAL_21]], %[[VAL_22]], %[[VAL_23]], %[[VAL_73:.*]]#2 into %[[VAL_6]]{{\[}}%[[VAL_20]]] : memref<?xf64>, memref<?xi1>, memref<?xindex>, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:             %[[VAL_77:.*]] = sparse_tensor.compress %[[VAL_23]], %[[VAL_24]], %[[VAL_25]], %[[VAL_78:.*]]#2 into %[[VAL_78]]#3{{\[}}%[[VAL_22]]] : memref<?xf64>, memref<?xi1>, memref<?xindex>, tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:             scf.yield %[[VAL_77]] : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           }
-// CHECK:           %[[VAL_74:.*]] = sparse_tensor.load %[[VAL_6]] hasInserts : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           return %[[VAL_74]] : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_79:.*]] = sparse_tensor.load %[[VAL_80:.*]] hasInserts : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_79]] : tensor<4x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @matmul2(%A: tensor<4x8xf64, #DCSR>,
               %B: tensor<8x4xf64, #DCSR>) -> tensor<4x4xf64, #DCSR> {
   %c4 = arith.constant 4 : index
   %C = bufferization.alloc_tensor() : tensor<4x4xf64, #DCSR>
   %D = linalg.matmul
     ins(%A, %B: tensor<4x8xf64, #DCSR>, tensor<8x4xf64, #DCSR>)
        outs(%C: tensor<4x4xf64, #DCSR>) -> tensor<4x4xf64, #DCSR>
   return %D: tensor<4x4xf64, #DCSR>
 }
 
 // CHECK-LABEL:   func.func @conv2d(
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<8x8xi32>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<6x6xi32>) -> tensor<6x6xi32> {
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 6 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_6:.*]] = bufferization.to_memref %[[VAL_0]] : memref<8x8xi32>
-// CHECK:           %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xi32>
+// CHECK-DAG:       %[[VAL_6:.*]] = bufferization.to_memref %[[VAL_0]] : memref<8x8xi32>
+// CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<3x3xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xi32>
 // CHECK:           %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<6x6xi32>
 // CHECK:           %[[VAL_13:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_15:.*]] = %[[VAL_13]] to %[[VAL_14]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_16:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_15]]] : memref<?xindex>
 // CHECK:             %[[VAL_17:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_15]]] : memref<?xindex>
 // CHECK:             %[[VAL_18:.*]] = arith.addi %[[VAL_15]], %[[VAL_5]] : index
 // CHECK:             %[[VAL_19:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_18]]] : memref<?xindex>

 // CHECK-LABEL:   func.func @quantized_matmul(
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<5x3xi8>,
 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<5x6xi64>) -> tensor<5x6xi64> {
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 5 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant 2 : i64
-// CHECK:           %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_0]] : memref<5x3xi8>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_10:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_11:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_12:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xi8>
+// CHECK-DAG:       %[[VAL_7:.*]] = bufferization.to_memref %[[VAL_0]] : memref<5x3xi8>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_12:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<3x6xi8, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xi8>
 // CHECK:           %[[VAL_13:.*]] = bufferization.to_memref %[[VAL_2]] : memref<5x6xi64>
 // CHECK:           %[[VAL_14:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_5]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_16:.*]] = %[[VAL_14]] to %[[VAL_15]] step %[[VAL_5]] {
 // CHECK:             %[[VAL_17:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_16]]] : memref<?xindex>
 // CHECK:             %[[VAL_18:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_16]]] : memref<?xindex>
 // CHECK:             %[[VAL_19:.*]] = arith.addi %[[VAL_16]], %[[VAL_5]] : index
 // CHECK:             %[[VAL_20:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_19]]] : memref<?xindex>

 }
 
 // CHECK-LABEL:   func.func @sparse_dot(
 // CHECK-SAME:      %[[VAL_0:.*0]]: tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>,
 // CHECK-SAME:      %[[VAL_1:.*1]]: tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>>,
 // CHECK-SAME:      %[[VAL_2:.*2]]: tensor<f32>) -> tensor<f32> {
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xf32>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xf32>
+// CHECK-DAG:       %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xf32>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_10:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<1024xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed" ] }>> to memref<?xf32>
 // CHECK:           %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_2]] : memref<f32>
 // CHECK:           %[[VAL_12:.*]] = memref.load %[[VAL_11]][] : memref<f32>
 // 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>
 // CHECK:           %[[VAL_16:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_17:.*]]:3 = scf.while (%[[VAL_18:.*]] = %[[VAL_13]], %[[VAL_19:.*]] = %[[VAL_15]], %[[VAL_20:.*]] = %[[VAL_12]]) : (index, index, f32) -> (index, index, f32) {
 // CHECK:             %[[VAL_21:.*]] = arith.cmpi ult, %[[VAL_18]], %[[VAL_14]] : index

mlir/test/Dialect/SparseTensor/sparse_out.mlir

#trait_scale_inpl = {
   indexing_maps = [
     affine_map<(i,j) -> (i,j)>   // X (out)
   ],
   iterator_types = ["parallel", "parallel"],
   doc = "X(i,j) *= 2 or X(i,j) += X(i,j)"
 }
 
 // CHECK-LABEL:   func.func @sparse_simply_dynamic1(
-// CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>> {
-// CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 2.000000e+00 : f32
-// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
-// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>> to memref<?xindex>
-// CHECK:           %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>> to memref<?xindex>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>> to memref<?xf32>
-// 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_6]]{{\[}}%[[VAL_11]]] : memref<?xindex>
-// CHECK:             %[[VAL_13:.*]] = arith.addi %[[VAL_11]], %[[VAL_3]] : index
-// CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_13]]] : memref<?xindex>
-// CHECK:             scf.for %[[VAL_15:.*]] = %[[VAL_12]] to %[[VAL_14]] step %[[VAL_3]] {
-// CHECK:               %[[VAL_16:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_15]]] : memref<?xf32>
-// CHECK:               %[[VAL_17:.*]] = arith.mulf %[[VAL_16]], %[[VAL_1]] : f32
-// CHECK:               memref.store %[[VAL_17]], %[[VAL_8]]{{\[}}%[[VAL_15]]] : memref<?xf32>
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
+// CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 0 : index
+// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 1 : index
+// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 2.000000e+00 : f32
+// CHECK-DAG:       %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>
+// CHECK:           %[[VAL_7:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_1]]] : memref<?xindex>
+// CHECK:           %[[VAL_8:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_2]]] : memref<?xindex>
+// CHECK:           scf.for %[[VAL_9:.*]] = %[[VAL_7]] to %[[VAL_8]] step %[[VAL_2]] {
+// CHECK:             %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_9]]] : memref<?xindex>
+// CHECK:             %[[VAL_11:.*]] = arith.addi %[[VAL_9]], %[[VAL_2]] : index
+// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_11]]] : memref<?xindex>
+// CHECK:             scf.for %[[VAL_13:.*]] = %[[VAL_10]] to %[[VAL_12]] step %[[VAL_2]] {
+// CHECK:               %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_13]]] : memref<?xf32>
+// CHECK:               %[[VAL_15:.*]] = arith.mulf %[[VAL_14]], %[[VAL_3]] : f32
+// CHECK:               memref.store %[[VAL_15]], %[[VAL_6]]{{\[}}%[[VAL_13]]] : memref<?xf32>
 // CHECK:             }
 // CHECK:           }
-// CHECK:           %[[VAL_18:.*]] = sparse_tensor.load %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
-// CHECK:           return %[[VAL_18]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
+// CHECK:           %[[VAL_16:.*]] = sparse_tensor.load %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_16]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @sparse_simply_dynamic1(%argx: tensor<32x16xf32, #DCSR>) -> tensor<32x16xf32, #DCSR> {
   %c = arith.constant 2.0 : f32
   %0 = linalg.generic #trait_scale_inpl
     outs(%argx: tensor<32x16xf32, #DCSR>) {
       ^bb(%x: f32):
         %1 = arith.mulf %x, %c : f32
         linalg.yield %1 : f32
   } -> tensor<32x16xf32, #DCSR>
   return %0 : tensor<32x16xf32, #DCSR>
 }
 
 // CHECK-LABEL:   func.func @sparse_simply_dynamic2(
-// CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
 // CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_3:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
-// CHECK:           %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
-// CHECK:           %[[VAL_5:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
+// CHECK-DAG:       %[[VAL_3:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_4:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_5:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>
 // CHECK:           %[[VAL_6:.*]] = memref.load %[[VAL_3]]{{\[}}%[[VAL_1]]] : memref<?xindex>
 // CHECK:           %[[VAL_7:.*]] = memref.load %[[VAL_3]]{{\[}}%[[VAL_2]]] : memref<?xindex>
 // CHECK:           scf.for %[[VAL_8:.*]] = %[[VAL_6]] to %[[VAL_7]] step %[[VAL_2]] {
 // CHECK:             %[[VAL_9:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_8]]] : memref<?xindex>
 // CHECK:             %[[VAL_10:.*]] = arith.addi %[[VAL_8]], %[[VAL_2]] : index
 // CHECK:             %[[VAL_11:.*]] = memref.load %[[VAL_4]]{{\[}}%[[VAL_10]]] : memref<?xindex>
 // CHECK:             scf.for %[[VAL_12:.*]] = %[[VAL_9]] to %[[VAL_11]] step %[[VAL_2]] {
 // CHECK:               %[[VAL_13:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_12]]] : memref<?xf32>
 // CHECK:               %[[VAL_14:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_12]]] : memref<?xf32>
 // CHECK:               %[[VAL_15:.*]] = arith.addf %[[VAL_13]], %[[VAL_14]] : f32
 // CHECK:               memref.store %[[VAL_15]], %[[VAL_5]]{{\[}}%[[VAL_12]]] : memref<?xf32>
 // CHECK:             }
 // CHECK:           }
-// CHECK:           %[[VAL_16:.*]] = sparse_tensor.load %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
-// CHECK:           return %[[VAL_16]] : tensor<32x16xf32, #sparse_tensor.encoding<{{.*}}>>
+// CHECK:           %[[VAL_16:.*]] = sparse_tensor.load %[[VAL_0]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_16]] : tensor<32x16xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @sparse_simply_dynamic2(%argx: tensor<32x16xf32, #DCSR>) -> tensor<32x16xf32, #DCSR> {
   %0 = linalg.generic #trait_scale_inpl
     outs(%argx: tensor<32x16xf32, #DCSR>) {
       ^bb(%x: f32):
         %1 = arith.addf %x, %x : f32
         linalg.yield %1 : f32
   } -> tensor<32x16xf32, #DCSR>

 }
 
 // CHECK-LABEL:   func.func @sparse_truly_dynamic(
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>>) -> tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
 // CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 10 : index
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 2.000000e+00 : f32
-// CHECK:           %[[VAL_5:.*]] = bufferization.alloc_tensor() : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>> to memref<?xf32>
-// CHECK:           scf.for %[[VAL_9:.*]] = %[[VAL_2]] to %[[VAL_1]] step %[[VAL_3]] {
-// CHECK:             %[[VAL_10:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_9]]] : memref<?xindex>
-// CHECK:             %[[VAL_11:.*]] = arith.addi %[[VAL_9]], %[[VAL_3]] : index
-// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_11]]] : memref<?xindex>
-// CHECK:             scf.for %[[VAL_13:.*]] = %[[VAL_10]] to %[[VAL_12]] step %[[VAL_3]] {
-// CHECK:               %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_13]]] : memref<?xindex>
-// CHECK:               %[[VAL_15:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_13]]] : memref<?xf32>
-// CHECK:               %[[VAL_16:.*]] = arith.mulf %[[VAL_15]], %[[VAL_4]] : f32
-// CHECK:               sparse_tensor.insert %[[VAL_16]] into %[[VAL_5]]{{\[}}%[[VAL_9]], %[[VAL_14]]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK-DAG:       %[[VAL_5:.*]] = bufferization.alloc_tensor() : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "dense", "compressed" ] }>> to memref<?xf32>
+// CHECK:           %[[VAL_9:.*]] = scf.for %[[VAL_10:.*]] = %[[VAL_2]] to %[[VAL_1]] step %[[VAL_3]] iter_args(%[[VAL_11:.*]] = %[[VAL_5]]) -> (tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:             %[[VAL_12:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_10]]] : memref<?xindex>
+// CHECK:             %[[VAL_13:.*]] = arith.addi %[[VAL_10]], %[[VAL_3]] : index
+// CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_13]]] : memref<?xindex>
+// CHECK:             %[[VAL_15:.*]] = scf.for %[[VAL_16:.*]] = %[[VAL_12]] to %[[VAL_14]] step %[[VAL_3]] iter_args(%[[VAL_17:.*]] = %[[VAL_11]]) -> (tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:               %[[VAL_18:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_16]]] : memref<?xindex>
+// CHECK:               %[[VAL_19:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_16]]] : memref<?xf32>
+// CHECK:               %[[VAL_20:.*]] = arith.mulf %[[VAL_19]], %[[VAL_4]] : f32
+// CHECK:               %[[VAL_21:.*]] = sparse_tensor.insert %[[VAL_20]] into %[[VAL_17]]{{\[}}%[[VAL_10]], %[[VAL_18]]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:               scf.yield %[[VAL_21]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             }
+// CHECK:             scf.yield %[[VAL_22:.*]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           }
-// CHECK:           %[[VAL_17:.*]] = sparse_tensor.load %[[VAL_5]] hasInserts : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           return %[[VAL_17]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_23:.*]] = sparse_tensor.load %[[VAL_24:.*]] hasInserts : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_23]] : tensor<10x20xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @sparse_truly_dynamic(%arga: tensor<10x20xf32, #CSR>) -> tensor<10x20xf32, #DCSR> {
   %s = arith.constant 2.0 : f32
   %xm = bufferization.alloc_tensor() : tensor<10x20xf32, #DCSR>
   %0 = linalg.generic #trait_scale
      ins(%arga: tensor<10x20xf32, #CSR>)
       outs(%xm: tensor<10x20xf32, #DCSR>) {
       ^bb(%a: f32, %x: f32):

 // CHECK:           %[[VAL_18:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<?x?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 2 : index} : tensor<?x?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_20:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 2 : index} : tensor<?x?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_21:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<?x?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed", "compressed" ] }>> to memref<?xi32>
 // CHECK:           %[[VAL_22:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_2]]] : memref<?xindex>
 // CHECK:           %[[VAL_23:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_3]]] : memref<?xindex>
 // CHECK:           %[[VAL_24:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_2]]] : memref<?xindex>
 // CHECK:           %[[VAL_25:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_3]]] : memref<?xindex>
-// CHECK:           %[[VAL_26:.*]]:2 = scf.while (%[[VAL_27:.*]] = %[[VAL_22]], %[[VAL_28:.*]] = %[[VAL_24]]) : (index, index) -> (index, index) {
-// CHECK:             %[[VAL_29:.*]] = arith.cmpi ult, %[[VAL_27]], %[[VAL_23]] : index
-// CHECK:             %[[VAL_30:.*]] = arith.cmpi ult, %[[VAL_28]], %[[VAL_25]] : index
-// CHECK:             %[[VAL_31:.*]] = arith.andi %[[VAL_29]], %[[VAL_30]] : i1
-// CHECK:             scf.condition(%[[VAL_31]]) %[[VAL_27]], %[[VAL_28]] : index, index
+// CHECK:           %[[VAL_26:.*]]:3 = scf.while (%[[VAL_27:.*]] = %[[VAL_22]], %[[VAL_28:.*]] = %[[VAL_24]], %[[VAL_29:.*]] = %[[VAL_7]]) : (index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> (index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:             %[[VAL_30:.*]] = arith.cmpi ult, %[[VAL_27]], %[[VAL_23]] : index
+// CHECK:             %[[VAL_31:.*]] = arith.cmpi ult, %[[VAL_28]], %[[VAL_25]] : index
+// CHECK:             %[[VAL_32:.*]] = arith.andi %[[VAL_30]], %[[VAL_31]] : i1
+// CHECK:             scf.condition(%[[VAL_32]]) %[[VAL_27]], %[[VAL_28]], %[[VAL_29]] : index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           } do {
-// CHECK:           ^bb0(%[[VAL_32:.*]]: index, %[[VAL_33:.*]]: index):
-// CHECK:             %[[VAL_34:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_32]]] : memref<?xindex>
-// CHECK:             %[[VAL_35:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_33]]] : memref<?xindex>
-// CHECK:             %[[VAL_36:.*]] = arith.cmpi ult, %[[VAL_35]], %[[VAL_34]] : index
-// CHECK:             %[[VAL_37:.*]] = arith.select %[[VAL_36]], %[[VAL_35]], %[[VAL_34]] : index
-// CHECK:             %[[VAL_38:.*]] = arith.cmpi eq, %[[VAL_34]], %[[VAL_37]] : index
-// CHECK:             %[[VAL_39:.*]] = arith.cmpi eq, %[[VAL_35]], %[[VAL_37]] : index
-// CHECK:             %[[VAL_40:.*]] = arith.andi %[[VAL_38]], %[[VAL_39]] : i1
-// CHECK:             scf.if %[[VAL_40]] {
-// CHECK:               %[[VAL_41:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_32]]] : memref<?xindex>
-// CHECK:               %[[VAL_42:.*]] = arith.addi %[[VAL_32]], %[[VAL_3]] : index
-// CHECK:               %[[VAL_43:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_42]]] : memref<?xindex>
-// CHECK:               %[[VAL_44:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_33]]] : memref<?xindex>
+// CHECK:           ^bb0(%[[VAL_33:.*]]: index, %[[VAL_34:.*]]: index, %[[VAL_35:.*]]: tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>):
+// CHECK:             %[[VAL_36:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_33]]] : memref<?xindex>
+// CHECK:             %[[VAL_37:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_34]]] : memref<?xindex>
+// CHECK:             %[[VAL_38:.*]] = arith.cmpi ult, %[[VAL_37]], %[[VAL_36]] : index
+// CHECK:             %[[VAL_39:.*]] = arith.select %[[VAL_38]], %[[VAL_37]], %[[VAL_36]] : index
+// CHECK:             %[[VAL_40:.*]] = arith.cmpi eq, %[[VAL_36]], %[[VAL_39]] : index
+// CHECK:             %[[VAL_41:.*]] = arith.cmpi eq, %[[VAL_37]], %[[VAL_39]] : index
+// CHECK:             %[[VAL_42:.*]] = arith.andi %[[VAL_40]], %[[VAL_41]] : i1
+// CHECK:             %[[VAL_43:.*]] = scf.if %[[VAL_42]] -> (tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:               %[[VAL_44:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_33]]] : memref<?xindex>
 // CHECK:               %[[VAL_45:.*]] = arith.addi %[[VAL_33]], %[[VAL_3]] : index
-// CHECK:               %[[VAL_46:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_45]]] : memref<?xindex>
-// CHECK:               %[[VAL_47:.*]]:2 = scf.while (%[[VAL_48:.*]] = %[[VAL_41]], %[[VAL_49:.*]] = %[[VAL_44]]) : (index, index) -> (index, index) {
-// CHECK:                 %[[VAL_50:.*]] = arith.cmpi ult, %[[VAL_48]], %[[VAL_43]] : index
-// CHECK:                 %[[VAL_51:.*]] = arith.cmpi ult, %[[VAL_49]], %[[VAL_46]] : index
-// CHECK:                 %[[VAL_52:.*]] = arith.andi %[[VAL_50]], %[[VAL_51]] : i1
-// CHECK:                 scf.condition(%[[VAL_52]]) %[[VAL_48]], %[[VAL_49]] : index, index
+// CHECK:               %[[VAL_46:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_45]]] : memref<?xindex>
+// CHECK:               %[[VAL_47:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_34]]] : memref<?xindex>
+// CHECK:               %[[VAL_48:.*]] = arith.addi %[[VAL_34]], %[[VAL_3]] : index
+// CHECK:               %[[VAL_49:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_48]]] : memref<?xindex>
+// CHECK:               %[[VAL_50:.*]]:3 = scf.while (%[[VAL_51:.*]] = %[[VAL_44]], %[[VAL_52:.*]] = %[[VAL_47]], %[[VAL_53:.*]] = %[[VAL_35]]) : (index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> (index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:                 %[[VAL_54:.*]] = arith.cmpi ult, %[[VAL_51]], %[[VAL_46]] : index
+// CHECK:                 %[[VAL_55:.*]] = arith.cmpi ult, %[[VAL_52]], %[[VAL_49]] : index
+// CHECK:                 %[[VAL_56:.*]] = arith.andi %[[VAL_54]], %[[VAL_55]] : i1
+// CHECK:                 scf.condition(%[[VAL_56]]) %[[VAL_51]], %[[VAL_52]], %[[VAL_53]] : index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:               } do {
-// CHECK:               ^bb0(%[[VAL_53:.*]]: index, %[[VAL_54:.*]]: index):
-// CHECK:                 %[[VAL_55:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_53]]] : memref<?xindex>
-// CHECK:                 %[[VAL_56:.*]] = memref.load %[[VAL_18]]{{\[}}%[[VAL_54]]] : memref<?xindex>
-// CHECK:                 %[[VAL_57:.*]] = arith.cmpi ult, %[[VAL_56]], %[[VAL_55]] : index
-// CHECK:                 %[[VAL_58:.*]] = arith.select %[[VAL_57]], %[[VAL_56]], %[[VAL_55]] : index
-// CHECK:                 %[[VAL_59:.*]] = arith.cmpi eq, %[[VAL_55]], %[[VAL_58]] : index
-// CHECK:                 %[[VAL_60:.*]] = arith.cmpi eq, %[[VAL_56]], %[[VAL_58]] : index
-// CHECK:                 %[[VAL_61:.*]] = arith.andi %[[VAL_59]], %[[VAL_60]] : i1
-// CHECK:                 scf.if %[[VAL_61]] {
-// CHECK:                   %[[VAL_62:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_53]]] : memref<?xindex>
-// CHECK:                   %[[VAL_63:.*]] = arith.addi %[[VAL_53]], %[[VAL_3]] : index
-// CHECK:                   %[[VAL_64:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_63]]] : memref<?xindex>
-// CHECK:                   %[[VAL_65:.*]] = memref.load %[[VAL_19]]{{\[}}%[[VAL_54]]] : memref<?xindex>
-// CHECK:                   %[[VAL_66:.*]] = arith.addi %[[VAL_54]], %[[VAL_3]] : index
-// CHECK:                   %[[VAL_67:.*]] = memref.load %[[VAL_19]]{{\[}}%[[VAL_66]]] : memref<?xindex>
-// CHECK:                   %[[VAL_68:.*]]:3 = scf.while (%[[VAL_69:.*]] = %[[VAL_62]], %[[VAL_70:.*]] = %[[VAL_65]], %[[VAL_71:.*]] = %[[VAL_4]]) : (index, index, i32) -> (index, index, i32) {
-// CHECK:                     %[[VAL_72:.*]] = arith.cmpi ult, %[[VAL_69]], %[[VAL_64]] : index
-// CHECK:                     %[[VAL_73:.*]] = arith.cmpi ult, %[[VAL_70]], %[[VAL_67]] : index
-// CHECK:                     %[[VAL_74:.*]] = arith.andi %[[VAL_72]], %[[VAL_73]] : i1
-// CHECK:                     scf.condition(%[[VAL_74]]) %[[VAL_69]], %[[VAL_70]], %[[VAL_71]] : index, index, i32
+// CHECK:               ^bb0(%[[VAL_57:.*]]: index, %[[VAL_58:.*]]: index, %[[VAL_59:.*]]: tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>):
+// CHECK:                 %[[VAL_60:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_57]]] : memref<?xindex>
+// CHECK:                 %[[VAL_61:.*]] = memref.load %[[VAL_18]]{{\[}}%[[VAL_58]]] : memref<?xindex>
+// CHECK:                 %[[VAL_62:.*]] = arith.cmpi ult, %[[VAL_61]], %[[VAL_60]] : index
+// CHECK:                 %[[VAL_63:.*]] = arith.select %[[VAL_62]], %[[VAL_61]], %[[VAL_60]] : index
+// CHECK:                 %[[VAL_64:.*]] = arith.cmpi eq, %[[VAL_60]], %[[VAL_63]] : index
+// CHECK:                 %[[VAL_65:.*]] = arith.cmpi eq, %[[VAL_61]], %[[VAL_63]] : index
+// CHECK:                 %[[VAL_66:.*]] = arith.andi %[[VAL_64]], %[[VAL_65]] : i1
+// CHECK:                 %[[VAL_67:.*]] = scf.if %[[VAL_66]] -> (tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:                   %[[VAL_68:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_57]]] : memref<?xindex>
+// CHECK:                   %[[VAL_69:.*]] = arith.addi %[[VAL_57]], %[[VAL_3]] : index
+// CHECK:                   %[[VAL_70:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_69]]] : memref<?xindex>
+// CHECK:                   %[[VAL_71:.*]] = memref.load %[[VAL_19]]{{\[}}%[[VAL_58]]] : memref<?xindex>
+// CHECK:                   %[[VAL_72:.*]] = arith.addi %[[VAL_58]], %[[VAL_3]] : index
+// CHECK:                   %[[VAL_73:.*]] = memref.load %[[VAL_19]]{{\[}}%[[VAL_72]]] : memref<?xindex>
+// CHECK:                   %[[VAL_74:.*]]:4 = scf.while (%[[VAL_75:.*]] = %[[VAL_68]], %[[VAL_76:.*]] = %[[VAL_71]], %[[VAL_77:.*]] = %[[VAL_4]], %[[VAL_78:.*]] = %[[VAL_59]]) : (index, index, i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> (index, index, i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:                     %[[VAL_79:.*]] = arith.cmpi ult, %[[VAL_75]], %[[VAL_70]] : index
+// CHECK:                     %[[VAL_80:.*]] = arith.cmpi ult, %[[VAL_76]], %[[VAL_73]] : index
+// CHECK:                     %[[VAL_81:.*]] = arith.andi %[[VAL_79]], %[[VAL_80]] : i1
+// CHECK:                     scf.condition(%[[VAL_81]]) %[[VAL_75]], %[[VAL_76]], %[[VAL_77]], %[[VAL_78]] : index, index, i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:                   } do {
-// CHECK:                   ^bb0(%[[VAL_75:.*]]: index, %[[VAL_76:.*]]: index, %[[VAL_77:.*]]: i32):
-// CHECK:                     %[[VAL_78:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_75]]] : memref<?xindex>
-// CHECK:                     %[[VAL_79:.*]] = memref.load %[[VAL_20]]{{\[}}%[[VAL_76]]] : memref<?xindex>
-// CHECK:                     %[[VAL_80:.*]] = arith.cmpi ult, %[[VAL_79]], %[[VAL_78]] : index
-// CHECK:                     %[[VAL_81:.*]] = arith.select %[[VAL_80]], %[[VAL_79]], %[[VAL_78]] : index
-// CHECK:                     %[[VAL_82:.*]] = arith.cmpi eq, %[[VAL_78]], %[[VAL_81]] : index
-// CHECK:                     %[[VAL_83:.*]] = arith.cmpi eq, %[[VAL_79]], %[[VAL_81]] : index
-// CHECK:                     %[[VAL_84:.*]] = arith.andi %[[VAL_82]], %[[VAL_83]] : i1
-// CHECK:                     %[[VAL_85:.*]] = scf.if %[[VAL_84]] -> (i32) {
-// CHECK:                       %[[VAL_86:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_75]]] : memref<?xi32>
-// CHECK:                       %[[VAL_87:.*]] = memref.load %[[VAL_21]]{{\[}}%[[VAL_76]]] : memref<?xi32>
-// CHECK:                       %[[VAL_88:.*]] = arith.muli %[[VAL_86]], %[[VAL_87]] : i32
-// CHECK:                       %[[VAL_89:.*]] = arith.addi %[[VAL_77]], %[[VAL_88]] : i32
-// CHECK:                       scf.yield %[[VAL_89]] : i32
+// CHECK:                   ^bb0(%[[VAL_82:.*]]: index, %[[VAL_83:.*]]: index, %[[VAL_84:.*]]: i32, %[[VAL_85:.*]]: tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>):
+// CHECK:                     %[[VAL_86:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_82]]] : memref<?xindex>
+// CHECK:                     %[[VAL_87:.*]] = memref.load %[[VAL_20]]{{\[}}%[[VAL_83]]] : memref<?xindex>
+// CHECK:                     %[[VAL_88:.*]] = arith.cmpi ult, %[[VAL_87]], %[[VAL_86]] : index
+// CHECK:                     %[[VAL_89:.*]] = arith.select %[[VAL_88]], %[[VAL_87]], %[[VAL_86]] : index
+// CHECK:                     %[[VAL_90:.*]] = arith.cmpi eq, %[[VAL_86]], %[[VAL_89]] : index
+// CHECK:                     %[[VAL_91:.*]] = arith.cmpi eq, %[[VAL_87]], %[[VAL_89]] : index
+// CHECK:                     %[[VAL_92:.*]] = arith.andi %[[VAL_90]], %[[VAL_91]] : i1
+// CHECK:                     %[[VAL_93:.*]]:2 = scf.if %[[VAL_92]] -> (i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:                       %[[VAL_94:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_82]]] : memref<?xi32>
+// CHECK:                       %[[VAL_95:.*]] = memref.load %[[VAL_21]]{{\[}}%[[VAL_83]]] : memref<?xi32>
+// CHECK:                       %[[VAL_96:.*]] = arith.muli %[[VAL_94]], %[[VAL_95]] : i32
+// CHECK:                       %[[VAL_97:.*]] = arith.addi %[[VAL_84]], %[[VAL_96]] : i32
+// CHECK:                       scf.yield %[[VAL_97]], %[[VAL_85]] : i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:                     } else {
-// CHECK:                       scf.yield %[[VAL_77]] : i32
+// CHECK:                       scf.yield %[[VAL_84]], %[[VAL_85]] : i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:                     }
-// CHECK:                     %[[VAL_90:.*]] = arith.cmpi eq, %[[VAL_78]], %[[VAL_81]] : index
-// CHECK:                     %[[VAL_91:.*]] = arith.addi %[[VAL_75]], %[[VAL_3]] : index
-// CHECK:                     %[[VAL_92:.*]] = arith.select %[[VAL_90]], %[[VAL_91]], %[[VAL_75]] : index
-// CHECK:                     %[[VAL_93:.*]] = arith.cmpi eq, %[[VAL_79]], %[[VAL_81]] : index
-// CHECK:                     %[[VAL_94:.*]] = arith.addi %[[VAL_76]], %[[VAL_3]] : index
-// CHECK:                     %[[VAL_95:.*]] = arith.select %[[VAL_93]], %[[VAL_94]], %[[VAL_76]] : index
-// CHECK:                     scf.yield %[[VAL_92]], %[[VAL_95]], %[[VAL_96:.*]] : index, index, i32
+// CHECK:                     %[[VAL_98:.*]] = arith.cmpi eq, %[[VAL_86]], %[[VAL_89]] : index
+// CHECK:                     %[[VAL_99:.*]] = arith.addi %[[VAL_82]], %[[VAL_3]] : index
+// CHECK:                     %[[VAL_100:.*]] = arith.select %[[VAL_98]], %[[VAL_99]], %[[VAL_82]] : index
+// CHECK:                     %[[VAL_101:.*]] = arith.cmpi eq, %[[VAL_87]], %[[VAL_89]] : index
+// CHECK:                     %[[VAL_102:.*]] = arith.addi %[[VAL_83]], %[[VAL_3]] : index
+// CHECK:                     %[[VAL_103:.*]] = arith.select %[[VAL_101]], %[[VAL_102]], %[[VAL_83]] : index
+// CHECK:                     scf.yield %[[VAL_100]], %[[VAL_103]], %[[VAL_104:.*]]#0, %[[VAL_104]]#1 : index, index, i32, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:                   }
-// CHECK:                   sparse_tensor.insert %[[VAL_97:.*]]#2 into %[[VAL_7]]{{\[}}%[[VAL_37]], %[[VAL_58]]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:                   %[[VAL_105:.*]] = sparse_tensor.insert %[[VAL_106:.*]]#2 into %[[VAL_106]]#3{{\[}}%[[VAL_39]], %[[VAL_63]]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:                   scf.yield %[[VAL_105]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:                 } else {
+// CHECK:                   scf.yield %[[VAL_59]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:                 }
-// CHECK:                 %[[VAL_98:.*]] = arith.cmpi eq, %[[VAL_55]], %[[VAL_58]] : index
-// CHECK:                 %[[VAL_99:.*]] = arith.addi %[[VAL_53]], %[[VAL_3]] : index
-// CHECK:                 %[[VAL_100:.*]] = arith.select %[[VAL_98]], %[[VAL_99]], %[[VAL_53]] : index
-// CHECK:                 %[[VAL_101:.*]] = arith.cmpi eq, %[[VAL_56]], %[[VAL_58]] : index
-// CHECK:                 %[[VAL_102:.*]] = arith.addi %[[VAL_54]], %[[VAL_3]] : index
-// CHECK:                 %[[VAL_103:.*]] = arith.select %[[VAL_101]], %[[VAL_102]], %[[VAL_54]] : index
-// CHECK:                 scf.yield %[[VAL_100]], %[[VAL_103]] : index, index
+// CHECK:                 %[[VAL_107:.*]] = arith.cmpi eq, %[[VAL_60]], %[[VAL_63]] : index
+// CHECK:                 %[[VAL_108:.*]] = arith.addi %[[VAL_57]], %[[VAL_3]] : index
+// CHECK:                 %[[VAL_109:.*]] = arith.select %[[VAL_107]], %[[VAL_108]], %[[VAL_57]] : index
+// CHECK:                 %[[VAL_110:.*]] = arith.cmpi eq, %[[VAL_61]], %[[VAL_63]] : index
+// CHECK:                 %[[VAL_111:.*]] = arith.addi %[[VAL_58]], %[[VAL_3]] : index
+// CHECK:                 %[[VAL_112:.*]] = arith.select %[[VAL_110]], %[[VAL_111]], %[[VAL_58]] : index
+// CHECK:                 scf.yield %[[VAL_109]], %[[VAL_112]], %[[VAL_113:.*]] : index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:               }
+// CHECK:               scf.yield %[[VAL_114:.*]]#2 : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             } else {
+// CHECK:               scf.yield %[[VAL_35]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             }
-// CHECK:             %[[VAL_104:.*]] = arith.cmpi eq, %[[VAL_34]], %[[VAL_37]] : index
-// CHECK:             %[[VAL_105:.*]] = arith.addi %[[VAL_32]], %[[VAL_3]] : index
-// CHECK:             %[[VAL_106:.*]] = arith.select %[[VAL_104]], %[[VAL_105]], %[[VAL_32]] : index
-// CHECK:             %[[VAL_107:.*]] = arith.cmpi eq, %[[VAL_35]], %[[VAL_37]] : index
-// CHECK:             %[[VAL_108:.*]] = arith.addi %[[VAL_33]], %[[VAL_3]] : index
-// CHECK:             %[[VAL_109:.*]] = arith.select %[[VAL_107]], %[[VAL_108]], %[[VAL_33]] : index
-// CHECK:             scf.yield %[[VAL_106]], %[[VAL_109]] : index, index
+// CHECK:             %[[VAL_115:.*]] = arith.cmpi eq, %[[VAL_36]], %[[VAL_39]] : index
+// CHECK:             %[[VAL_116:.*]] = arith.addi %[[VAL_33]], %[[VAL_3]] : index
+// CHECK:             %[[VAL_117:.*]] = arith.select %[[VAL_115]], %[[VAL_116]], %[[VAL_33]] : index
+// CHECK:             %[[VAL_118:.*]] = arith.cmpi eq, %[[VAL_37]], %[[VAL_39]] : index
+// CHECK:             %[[VAL_119:.*]] = arith.addi %[[VAL_34]], %[[VAL_3]] : index
+// CHECK:             %[[VAL_120:.*]] = arith.select %[[VAL_118]], %[[VAL_119]], %[[VAL_34]] : index
+// CHECK:             scf.yield %[[VAL_117]], %[[VAL_120]], %[[VAL_121:.*]] : index, index, tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           }
-// CHECK:           %[[VAL_110:.*]] = sparse_tensor.load %[[VAL_7]] hasInserts : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           return %[[VAL_110]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_122:.*]] = sparse_tensor.load %[[VAL_123:.*]]#2 hasInserts : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_122]] : tensor<?x?xi32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @sumred(%arga: tensor<?x?x?xi32, #SparseTensor>,
              %argb: tensor<?x?x?xi32, #SparseTensor>) -> tensor<?x?xi32, #DCSR> {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 1 : index
   %d0 = tensor.dim %arga, %c0 : tensor<?x?x?xi32, #SparseTensor>
   %d1 = tensor.dim %arga, %c1 : tensor<?x?x?xi32, #SparseTensor>
   %xinit = bufferization.alloc_tensor(%d0, %d1) : tensor<?x?xi32, #DCSR>
indexing_maps = [
     affine_map<(i,j,k) -> (k,j)>, // B
     affine_map<(i,j,k) -> (i,j)>  // C (out)
   ],
   iterator_types = ["parallel", "parallel", "reduction"],
   doc = "C(i,j) = SUM_k A(i,k) * B(k,j)"
 }
 
 // CHECK-LABEL:   func.func @matmat(
-// CHECK-SAME:                      %[[VAL_0:.*]]: tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>,
-// CHECK-SAME:                      %[[VAL_1:.*]]: tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
-// CHECK:           %[[VAL_2:.*]] = arith.constant 0 : index
-// CHECK:           %[[VAL_3:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_4:.*]] = arith.constant false
-// CHECK:           %[[VAL_5:.*]] = arith.constant true
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>,
+// CHECK-SAME:      %[[VAL_1:.*]]: tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
+// CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 0 : index
+// CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 1 : index
+// CHECK-DAG:       %[[VAL_4:.*]] = arith.constant false
+// CHECK-DAG:       %[[VAL_5:.*]] = arith.constant true
 // CHECK:           %[[VAL_6:.*]] = tensor.dim %[[VAL_0]], %[[VAL_2]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           %[[VAL_7:.*]] = tensor.dim %[[VAL_1]], %[[VAL_3]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           %[[VAL_8:.*]] = bufferization.alloc_tensor(%[[VAL_6]], %[[VAL_7]]) : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           %[[VAL_9:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_10:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_11:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_12:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_13:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>
 // CHECK:           %[[VAL_14:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 0 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_15:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 0 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_16:.*]] = sparse_tensor.pointers %[[VAL_1]] {dimension = 1 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_17:.*]] = sparse_tensor.indices %[[VAL_1]] {dimension = 1 : index} : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
 // CHECK:           %[[VAL_18:.*]] = sparse_tensor.values %[[VAL_1]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_2]]] : memref<?xindex>
 // CHECK:           %[[VAL_20:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_3]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_21:.*]] = %[[VAL_19]] to %[[VAL_20]] step %[[VAL_3]] {
-// CHECK:             %[[VAL_22:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_21]]] : memref<?xindex>
-// CHECK:             %[[VAL_23:.*]], %[[VAL_24:.*]], %[[VAL_25:.*]], %[[VAL_26:.*]] = sparse_tensor.expand %[[VAL_8]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>, memref<?xi1>, memref<?xindex>
-// CHECK:             %[[VAL_27:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_21]]] : memref<?xindex>
-// CHECK:             %[[VAL_28:.*]] = arith.addi %[[VAL_21]], %[[VAL_3]] : index
-// CHECK:             %[[VAL_29:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_28]]] : memref<?xindex>
-// CHECK:             %[[VAL_30:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_2]]] : memref<?xindex>
-// CHECK:             %[[VAL_31:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_3]]] : memref<?xindex>
-// CHECK:             %[[VAL_32:.*]]:3 = scf.while (%[[VAL_33:.*]] = %[[VAL_27]], %[[VAL_34:.*]] = %[[VAL_30]], %[[VAL_35:.*]] = %[[VAL_26]]) : (index, index, index) -> (index, index, index) {
-// CHECK:               %[[VAL_36:.*]] = arith.cmpi ult, %[[VAL_33]], %[[VAL_29]] : index
-// CHECK:               %[[VAL_37:.*]] = arith.cmpi ult, %[[VAL_34]], %[[VAL_31]] : index
-// CHECK:               %[[VAL_38:.*]] = arith.andi %[[VAL_36]], %[[VAL_37]] : i1
-// CHECK:               scf.condition(%[[VAL_38]]) %[[VAL_33]], %[[VAL_34]], %[[VAL_35]] : index, index, index
+// CHECK:           %[[VAL_21:.*]] = scf.for %[[VAL_22:.*]] = %[[VAL_19]] to %[[VAL_20]] step %[[VAL_3]] iter_args(%[[VAL_23:.*]] = %[[VAL_8]]) -> (tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:             %[[VAL_24:.*]] = memref.load %[[VAL_10]]{{\[}}%[[VAL_22]]] : memref<?xindex>
+// CHECK:             %[[VAL_25:.*]], %[[VAL_26:.*]], %[[VAL_27:.*]], %[[VAL_28:.*]] = sparse_tensor.expand %[[VAL_8]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf32>, memref<?xi1>, memref<?xindex>
+// CHECK:             %[[VAL_29:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_22]]] : memref<?xindex>
+// CHECK:             %[[VAL_30:.*]] = arith.addi %[[VAL_22]], %[[VAL_3]] : index
+// CHECK:             %[[VAL_31:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_30]]] : memref<?xindex>
+// CHECK:             %[[VAL_32:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_2]]] : memref<?xindex>
+// CHECK:             %[[VAL_33:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_3]]] : memref<?xindex>
+// CHECK:             %[[VAL_34:.*]]:4 = scf.while (%[[VAL_35:.*]] = %[[VAL_29]], %[[VAL_36:.*]] = %[[VAL_32]], %[[VAL_37:.*]] = %[[VAL_28]], %[[VAL_38:.*]] = %[[VAL_23]]) : (index, index, index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> (index, index, index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:               %[[VAL_39:.*]] = arith.cmpi ult, %[[VAL_35]], %[[VAL_31]] : index
+// CHECK:               %[[VAL_40:.*]] = arith.cmpi ult, %[[VAL_36]], %[[VAL_33]] : index
+// CHECK:               %[[VAL_41:.*]] = arith.andi %[[VAL_39]], %[[VAL_40]] : i1
+// CHECK:               scf.condition(%[[VAL_41]]) %[[VAL_35]], %[[VAL_36]], %[[VAL_37]], %[[VAL_38]] : index, index, index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             } do {
-// CHECK:             ^bb0(%[[VAL_39:.*]]: index, %[[VAL_40:.*]]: index, %[[VAL_41:.*]]: index):
-// CHECK:               %[[VAL_42:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_39]]] : memref<?xindex>
-// CHECK:               %[[VAL_43:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_40]]] : memref<?xindex>
-// CHECK:               %[[VAL_44:.*]] = arith.cmpi ult, %[[VAL_43]], %[[VAL_42]] : index
-// CHECK:               %[[VAL_45:.*]] = arith.select %[[VAL_44]], %[[VAL_43]], %[[VAL_42]] : index
-// CHECK:               %[[VAL_46:.*]] = arith.cmpi eq, %[[VAL_42]], %[[VAL_45]] : index
-// CHECK:               %[[VAL_47:.*]] = arith.cmpi eq, %[[VAL_43]], %[[VAL_45]] : index
-// CHECK:               %[[VAL_48:.*]] = arith.andi %[[VAL_46]], %[[VAL_47]] : i1
-// CHECK:               %[[VAL_49:.*]] = scf.if %[[VAL_48]] -> (index) {
-// CHECK:                 %[[VAL_50:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_39]]] : memref<?xf32>
-// CHECK:                 %[[VAL_51:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_40]]] : memref<?xindex>
-// CHECK:                 %[[VAL_52:.*]] = arith.addi %[[VAL_40]], %[[VAL_3]] : index
-// CHECK:                 %[[VAL_53:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_52]]] : memref<?xindex>
-// CHECK:                 %[[VAL_54:.*]] = scf.for %[[VAL_55:.*]] = %[[VAL_51]] to %[[VAL_53]] step %[[VAL_3]] iter_args(%[[VAL_56:.*]] = %[[VAL_41]]) -> (index) {
-// CHECK:                   %[[VAL_57:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_55]]] : memref<?xindex>
-// CHECK:                   %[[VAL_58:.*]] = memref.load %[[VAL_23]]{{\[}}%[[VAL_57]]] : memref<?xf32>
-// CHECK:                   %[[VAL_59:.*]] = memref.load %[[VAL_18]]{{\[}}%[[VAL_55]]] : memref<?xf32>
-// CHECK:                   %[[VAL_60:.*]] = arith.mulf %[[VAL_50]], %[[VAL_59]] : f32
-// CHECK:                   %[[VAL_61:.*]] = arith.addf %[[VAL_58]], %[[VAL_60]] : f32
-// CHECK:                   %[[VAL_62:.*]] = memref.load %[[VAL_24]]{{\[}}%[[VAL_57]]] : memref<?xi1>
-// CHECK:                   %[[VAL_63:.*]] = arith.cmpi eq, %[[VAL_62]], %[[VAL_4]] : i1
-// CHECK:                   %[[VAL_64:.*]] = scf.if %[[VAL_63]] -> (index) {
-// CHECK:                     memref.store %[[VAL_5]], %[[VAL_24]]{{\[}}%[[VAL_57]]] : memref<?xi1>
-// CHECK:                     memref.store %[[VAL_57]], %[[VAL_25]]{{\[}}%[[VAL_56]]] : memref<?xindex>
-// CHECK:                     %[[VAL_65:.*]] = arith.addi %[[VAL_56]], %[[VAL_3]] : index
-// CHECK:                     scf.yield %[[VAL_65]] : index
+// CHECK:             ^bb0(%[[VAL_42:.*]]: index, %[[VAL_43:.*]]: index, %[[VAL_44:.*]]: index, %[[VAL_45:.*]]: tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>):
+// CHECK:               %[[VAL_46:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_42]]] : memref<?xindex>
+// CHECK:               %[[VAL_47:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_43]]] : memref<?xindex>
+// CHECK:               %[[VAL_48:.*]] = arith.cmpi ult, %[[VAL_47]], %[[VAL_46]] : index
+// CHECK:               %[[VAL_49:.*]] = arith.select %[[VAL_48]], %[[VAL_47]], %[[VAL_46]] : index
+// CHECK:               %[[VAL_50:.*]] = arith.cmpi eq, %[[VAL_46]], %[[VAL_49]] : index
+// CHECK:               %[[VAL_51:.*]] = arith.cmpi eq, %[[VAL_47]], %[[VAL_49]] : index
+// CHECK:               %[[VAL_52:.*]] = arith.andi %[[VAL_50]], %[[VAL_51]] : i1
+// CHECK:               %[[VAL_53:.*]]:2 = scf.if %[[VAL_52]] -> (index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:                 %[[VAL_54:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_42]]] : memref<?xf32>
+// CHECK:                 %[[VAL_55:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_43]]] : memref<?xindex>
+// CHECK:                 %[[VAL_56:.*]] = arith.addi %[[VAL_43]], %[[VAL_3]] : index
+// CHECK:                 %[[VAL_57:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_56]]] : memref<?xindex>
+// CHECK:                 %[[VAL_58:.*]] = scf.for %[[VAL_59:.*]] = %[[VAL_55]] to %[[VAL_57]] step %[[VAL_3]] iter_args(%[[VAL_60:.*]] = %[[VAL_44]]) -> (index) {
+// CHECK:                   %[[VAL_61:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_59]]] : memref<?xindex>
+// CHECK:                   %[[VAL_62:.*]] = memref.load %[[VAL_25]]{{\[}}%[[VAL_61]]] : memref<?xf32>
+// CHECK:                   %[[VAL_63:.*]] = memref.load %[[VAL_18]]{{\[}}%[[VAL_59]]] : memref<?xf32>
+// CHECK:                   %[[VAL_64:.*]] = arith.mulf %[[VAL_54]], %[[VAL_63]] : f32
+// CHECK:                   %[[VAL_65:.*]] = arith.addf %[[VAL_62]], %[[VAL_64]] : f32
+// CHECK:                   %[[VAL_66:.*]] = memref.load %[[VAL_26]]{{\[}}%[[VAL_61]]] : memref<?xi1>
+// CHECK:                   %[[VAL_67:.*]] = arith.cmpi eq, %[[VAL_66]], %[[VAL_4]] : i1
+// CHECK:                   %[[VAL_68:.*]] = scf.if %[[VAL_67]] -> (index) {
+// CHECK:                     memref.store %[[VAL_5]], %[[VAL_26]]{{\[}}%[[VAL_61]]] : memref<?xi1>
+// CHECK:                     memref.store %[[VAL_61]], %[[VAL_27]]{{\[}}%[[VAL_60]]] : memref<?xindex>
+// CHECK:                     %[[VAL_69:.*]] = arith.addi %[[VAL_60]], %[[VAL_3]] : index
+// CHECK:                     scf.yield %[[VAL_69]] : index
 // CHECK:                   } else {
-// CHECK:                     scf.yield %[[VAL_56]] : index
+// CHECK:                     scf.yield %[[VAL_60]] : index
 // CHECK:                   }
-// CHECK:                   memref.store %[[VAL_61]], %[[VAL_23]]{{\[}}%[[VAL_57]]] : memref<?xf32>
-// CHECK:                   scf.yield %[[VAL_66:.*]] : index
+// CHECK:                   memref.store %[[VAL_65]], %[[VAL_25]]{{\[}}%[[VAL_61]]] : memref<?xf32>
+// CHECK:                   scf.yield %[[VAL_70:.*]] : index
 // CHECK:                 }
-// CHECK:                 scf.yield %[[VAL_67:.*]] : index
+// CHECK:                 scf.yield %[[VAL_71:.*]], %[[VAL_45]] : index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:               } else {
-// CHECK:                 scf.yield %[[VAL_41]] : index
+// CHECK:                 scf.yield %[[VAL_44]], %[[VAL_45]] : index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:               }
-// CHECK:               %[[VAL_68:.*]] = arith.cmpi eq, %[[VAL_42]], %[[VAL_45]] : index
-// CHECK:               %[[VAL_69:.*]] = arith.addi %[[VAL_39]], %[[VAL_3]] : index
-// CHECK:               %[[VAL_70:.*]] = arith.select %[[VAL_68]], %[[VAL_69]], %[[VAL_39]] : index
-// CHECK:               %[[VAL_71:.*]] = arith.cmpi eq, %[[VAL_43]], %[[VAL_45]] : index
-// CHECK:               %[[VAL_72:.*]] = arith.addi %[[VAL_40]], %[[VAL_3]] : index
-// CHECK:               %[[VAL_73:.*]] = arith.select %[[VAL_71]], %[[VAL_72]], %[[VAL_40]] : index
-// CHECK:               scf.yield %[[VAL_70]], %[[VAL_73]], %[[VAL_74:.*]] : index, index, index
+// CHECK:               %[[VAL_72:.*]] = arith.cmpi eq, %[[VAL_46]], %[[VAL_49]] : index
+// CHECK:               %[[VAL_73:.*]] = arith.addi %[[VAL_42]], %[[VAL_3]] : index
+// CHECK:               %[[VAL_74:.*]] = arith.select %[[VAL_72]], %[[VAL_73]], %[[VAL_42]] : index
+// CHECK:               %[[VAL_75:.*]] = arith.cmpi eq, %[[VAL_47]], %[[VAL_49]] : index
+// CHECK:               %[[VAL_76:.*]] = arith.addi %[[VAL_43]], %[[VAL_3]] : index
+// CHECK:               %[[VAL_77:.*]] = arith.select %[[VAL_75]], %[[VAL_76]], %[[VAL_43]] : index
+// CHECK:               scf.yield %[[VAL_74]], %[[VAL_77]], %[[VAL_78:.*]]#0, %[[VAL_78]]#1 : index, index, index, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             }
-// CHECK:             sparse_tensor.compress %[[VAL_23]], %[[VAL_24]], %[[VAL_25]], %[[VAL_75:.*]]#2 into %[[VAL_8]]{{\[}}%[[VAL_22]]] : memref<?xf32>, memref<?xi1>, memref<?xindex>, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:             %[[VAL_79:.*]] = sparse_tensor.compress %[[VAL_25]], %[[VAL_26]], %[[VAL_27]], %[[VAL_80:.*]]#2 into %[[VAL_80]]#3{{\[}}%[[VAL_24]]] : memref<?xf32>, memref<?xi1>, memref<?xindex>, tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:             scf.yield %[[VAL_79]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           }
-// CHECK:           %[[VAL_76:.*]] = sparse_tensor.load %[[VAL_8]] hasInserts : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           return %[[VAL_76]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_81:.*]] = sparse_tensor.load %[[VAL_82:.*]] hasInserts : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_81]] : tensor<?x?xf32, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @matmat(%arga: tensor<?x?xf32, #DCSR>,
              %argb: tensor<?x?xf32, #DCSR>) -> tensor<?x?xf32, #DCSR> {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 1 : index
   %d0 = tensor.dim %arga, %c0 : tensor<?x?xf32, #DCSR>
   %d1 = tensor.dim %argb, %c1 : tensor<?x?xf32, #DCSR>
   %cinit = bufferization.alloc_tensor(%d0, %d1) : tensor<?x?xf32, #DCSR>

mlir/test/Dialect/SparseTensor/sparse_sddmm.mlir

 // CHECK-SAME:      %[[VAL_1:.*]]: tensor<8x8xf64>,
 // CHECK-SAME:      %[[VAL_2:.*]]: tensor<8x8xf64>) -> tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
 // CHECK-DAG:       %[[VAL_3:.*]] = arith.constant 8 : index
 // CHECK-DAG:       %[[VAL_4:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant false
 // CHECK-DAG:       %[[VAL_7:.*]] = arith.constant true
 // CHECK-DAG:       %[[VAL_8:.*]] = arith.constant dense<0.000000e+00> : tensor<8x8xf64>
-// CHECK:           %[[VAL_9:.*]] = bufferization.alloc_tensor() copy(%[[VAL_8]]) {bufferization.escape = [false]} : tensor<8x8xf64>
-// CHECK:           %[[VAL_10:.*]] = bufferization.alloc_tensor() {bufferization.escape = [false]} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<8x8xf64>
-// CHECK:           %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<8x8xf64>
-// CHECK:           %[[VAL_13:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_14:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_15:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_16:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
-// CHECK:           %[[VAL_17:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>
+// CHECK-DAG:       %[[VAL_9:.*]] = bufferization.alloc_tensor() copy(%[[VAL_8]]) {bufferization.escape = [false]} : tensor<8x8xf64>
+// CHECK-DAG:       %[[VAL_10:.*]] = bufferization.alloc_tensor() {bufferization.escape = [false]} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK-DAG:       %[[VAL_11:.*]] = bufferization.to_memref %[[VAL_1]] : memref<8x8xf64>
+// CHECK-DAG:       %[[VAL_12:.*]] = bufferization.to_memref %[[VAL_2]] : memref<8x8xf64>
+// CHECK-DAG:       %[[VAL_13:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 0 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_14:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 0 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_15:.*]] = sparse_tensor.pointers %[[VAL_0]] {dimension = 1 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_16:.*]] = sparse_tensor.indices %[[VAL_0]] {dimension = 1 : index} : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_17:.*]] = sparse_tensor.values %[[VAL_0]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>
 // CHECK:           %[[VAL_18:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_4]]] : memref<?xindex>
 // CHECK:           %[[VAL_19:.*]] = memref.load %[[VAL_13]]{{\[}}%[[VAL_5]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_20:.*]] = %[[VAL_18]] to %[[VAL_19]] step %[[VAL_5]] {
-// CHECK:             %[[VAL_21:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_20]]] : memref<?xindex>
-// CHECK:             %[[VAL_22:.*]], %[[VAL_23:.*]], %[[VAL_24:.*]], %[[VAL_25:.*]] = sparse_tensor.expand %[[VAL_10]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>, memref<?xi1>, memref<?xindex>
-// CHECK:             %[[VAL_26:.*]] = scf.for %[[VAL_27:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] iter_args(%[[VAL_28:.*]] = %[[VAL_25]]) -> (index) {
-// CHECK:               %[[VAL_29:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_21]], %[[VAL_27]]] : memref<8x8xf64>
-// CHECK:               %[[VAL_30:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_20]]] : memref<?xindex>
-// CHECK:               %[[VAL_31:.*]] = arith.addi %[[VAL_20]], %[[VAL_5]] : index
-// CHECK:               %[[VAL_32:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_31]]] : memref<?xindex>
-// CHECK:               %[[VAL_33:.*]] = scf.for %[[VAL_34:.*]] = %[[VAL_30]] to %[[VAL_32]] step %[[VAL_5]] iter_args(%[[VAL_35:.*]] = %[[VAL_28]]) -> (index) {
-// CHECK:                 %[[VAL_36:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_34]]] : memref<?xindex>
-// CHECK:                 %[[VAL_37:.*]] = memref.load %[[VAL_22]]{{\[}}%[[VAL_36]]] : memref<?xf64>
-// CHECK:                 %[[VAL_38:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_27]], %[[VAL_36]]] : memref<8x8xf64>
-// CHECK:                 %[[VAL_39:.*]] = arith.mulf %[[VAL_29]], %[[VAL_38]] : f64
-// CHECK:                 %[[VAL_40:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_34]]] : memref<?xf64>
-// CHECK:                 %[[VAL_41:.*]] = arith.mulf %[[VAL_39]], %[[VAL_40]] : f64
-// CHECK:                 %[[VAL_42:.*]] = arith.addf %[[VAL_37]], %[[VAL_41]] : f64
-// CHECK:                 %[[VAL_43:.*]] = memref.load %[[VAL_23]]{{\[}}%[[VAL_36]]] : memref<?xi1>
-// CHECK:                 %[[VAL_44:.*]] = arith.cmpi eq, %[[VAL_43]], %[[VAL_6]] : i1
-// CHECK:                 %[[VAL_45:.*]] = scf.if %[[VAL_44]] -> (index) {
-// CHECK:                   memref.store %[[VAL_7]], %[[VAL_23]]{{\[}}%[[VAL_36]]] : memref<?xi1>
-// CHECK:                   memref.store %[[VAL_36]], %[[VAL_24]]{{\[}}%[[VAL_35]]] : memref<?xindex>
-// CHECK:                   %[[VAL_46:.*]] = arith.addi %[[VAL_35]], %[[VAL_5]] : index
-// CHECK:                   scf.yield %[[VAL_46]] : index
+// CHECK:           %[[VAL_20:.*]] = scf.for %[[VAL_21:.*]] = %[[VAL_18]] to %[[VAL_19]] step %[[VAL_5]] iter_args(%[[VAL_22:.*]] = %[[VAL_10]]) -> (tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:             %[[VAL_23:.*]] = memref.load %[[VAL_14]]{{\[}}%[[VAL_21]]] : memref<?xindex>
+// CHECK:             %[[VAL_24:.*]], %[[VAL_25:.*]], %[[VAL_26:.*]], %[[VAL_27:.*]] = sparse_tensor.expand %[[VAL_10]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to memref<?xf64>, memref<?xi1>, memref<?xindex>
+// CHECK:             %[[VAL_28:.*]] = scf.for %[[VAL_29:.*]] = %[[VAL_4]] to %[[VAL_3]] step %[[VAL_5]] iter_args(%[[VAL_30:.*]] = %[[VAL_27]]) -> (index) {
+// CHECK:               %[[VAL_31:.*]] = memref.load %[[VAL_11]]{{\[}}%[[VAL_23]], %[[VAL_29]]] : memref<8x8xf64>
+// CHECK:               %[[VAL_32:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_21]]] : memref<?xindex>
+// CHECK:               %[[VAL_33:.*]] = arith.addi %[[VAL_21]], %[[VAL_5]] : index
+// CHECK:               %[[VAL_34:.*]] = memref.load %[[VAL_15]]{{\[}}%[[VAL_33]]] : memref<?xindex>
+// CHECK:               %[[VAL_35:.*]] = scf.for %[[VAL_36:.*]] = %[[VAL_32]] to %[[VAL_34]] step %[[VAL_5]] iter_args(%[[VAL_37:.*]] = %[[VAL_30]]) -> (index) {
+// CHECK:                 %[[VAL_38:.*]] = memref.load %[[VAL_16]]{{\[}}%[[VAL_36]]] : memref<?xindex>
+// CHECK:                 %[[VAL_39:.*]] = memref.load %[[VAL_24]]{{\[}}%[[VAL_38]]] : memref<?xf64>
+// CHECK:                 %[[VAL_40:.*]] = memref.load %[[VAL_12]]{{\[}}%[[VAL_29]], %[[VAL_38]]] : memref<8x8xf64>
+// CHECK:                 %[[VAL_41:.*]] = arith.mulf %[[VAL_31]], %[[VAL_40]] : f64
+// CHECK:                 %[[VAL_42:.*]] = memref.load %[[VAL_17]]{{\[}}%[[VAL_36]]] : memref<?xf64>
+// CHECK:                 %[[VAL_43:.*]] = arith.mulf %[[VAL_41]], %[[VAL_42]] : f64
+// CHECK:                 %[[VAL_44:.*]] = arith.addf %[[VAL_39]], %[[VAL_43]] : f64
+// CHECK:                 %[[VAL_45:.*]] = memref.load %[[VAL_25]]{{\[}}%[[VAL_38]]] : memref<?xi1>
+// CHECK:                 %[[VAL_46:.*]] = arith.cmpi eq, %[[VAL_45]], %[[VAL_6]] : i1
+// CHECK:                 %[[VAL_47:.*]] = scf.if %[[VAL_46]] -> (index) {
+// CHECK:                   memref.store %[[VAL_7]], %[[VAL_25]]{{\[}}%[[VAL_38]]] : memref<?xi1>
+// CHECK:                   memref.store %[[VAL_38]], %[[VAL_26]]{{\[}}%[[VAL_37]]] : memref<?xindex>
+// CHECK:                   %[[VAL_48:.*]] = arith.addi %[[VAL_37]], %[[VAL_5]] : index
+// CHECK:                   scf.yield %[[VAL_48]] : index
 // CHECK:                 } else {
-// CHECK:                   scf.yield %[[VAL_35]] : index
+// CHECK:                   scf.yield %[[VAL_37]] : index
 // CHECK:                 }
-// CHECK:                 memref.store %[[VAL_42]], %[[VAL_22]]{{\[}}%[[VAL_36]]] : memref<?xf64>
-// CHECK:                 scf.yield %[[VAL_47:.*]] : index
+// CHECK:                 memref.store %[[VAL_44]], %[[VAL_24]]{{\[}}%[[VAL_38]]] : memref<?xf64>
+// CHECK:                 scf.yield %[[VAL_49:.*]] : index
 // CHECK:               }
-// CHECK:               scf.yield %[[VAL_48:.*]] : index
+// CHECK:               scf.yield %[[VAL_50:.*]] : index
 // CHECK:             }
-// CHECK:             sparse_tensor.compress %[[VAL_22]], %[[VAL_23]], %[[VAL_24]], %[[VAL_49:.*]] into %[[VAL_10]]{{\[}}%[[VAL_21]]] : memref<?xf64>, memref<?xi1>, memref<?xindex>, tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:             %[[VAL_51:.*]] = sparse_tensor.compress %[[VAL_24]], %[[VAL_25]], %[[VAL_26]], %[[VAL_52:.*]] into %[[VAL_22]]{{\[}}%[[VAL_23]]] : memref<?xf64>, memref<?xi1>, memref<?xindex>, tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:             scf.yield %[[VAL_51]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           }
-// CHECK:           %[[VAL_50:.*]] = sparse_tensor.load %[[VAL_10]] hasInserts : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           return %[[VAL_50]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_53:.*]] = sparse_tensor.load %[[VAL_54:.*]] hasInserts : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_53]] : tensor<8x8xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @sparse_sampled_dd_unfused(%args: tensor<8x8xf64, #SM>,
                                      %arga: tensor<8x8xf64>,
                                      %argb: tensor<8x8xf64>) -> tensor<8x8xf64, #SM> {
   // Perform dense-dense matrix matrix multiplication.
   %1 = arith.constant dense<0.0> : tensor<8x8xf64>
   %2 = linalg.generic #trait_matmul
     ins(%arga, %argb : tensor<8x8xf64>, tensor<8x8xf64>)

mlir/test/Dialect/SparseTensor/sparse_transpose.mlir

 }
 
 // TODO: improve auto-conversion followed by yield
 
 // CHECK-LABEL:   func.func @sparse_transpose_auto(
 // CHECK-SAME:      %[[VAL_0:.*]]: tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) -> tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> {
 // CHECK-DAG:       %[[VAL_1:.*]] = arith.constant 0 : index
 // CHECK-DAG:       %[[VAL_2:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_3:.*]] = bufferization.alloc_tensor() : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
-// CHECK:           %[[VAL_4:.*]] = sparse_tensor.convert %[[VAL_0]] : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>>
-// CHECK:           %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_4]] {dimension = 0 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
-// CHECK:           %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_4]] {dimension = 0 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
-// CHECK:           %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_4]] {dimension = 1 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
-// CHECK:           %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_4]] {dimension = 1 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
-// CHECK:           %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_4]] : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xf64>
+// CHECK-DAG:       %[[VAL_3:.*]] = bufferization.alloc_tensor() : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK-DAG:       %[[VAL_4:.*]] = sparse_tensor.convert %[[VAL_0]] : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>> to tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>>
+// CHECK-DAG:       %[[VAL_5:.*]] = sparse_tensor.pointers %[[VAL_4]] {dimension = 0 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_6:.*]] = sparse_tensor.indices %[[VAL_4]] {dimension = 0 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_7:.*]] = sparse_tensor.pointers %[[VAL_4]] {dimension = 1 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_8:.*]] = sparse_tensor.indices %[[VAL_4]] {dimension = 1 : index} : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xindex>
+// CHECK-DAG:       %[[VAL_9:.*]] = sparse_tensor.values %[[VAL_4]] : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>> to memref<?xf64>
 // CHECK:           %[[VAL_10:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_1]]] : memref<?xindex>
 // CHECK:           %[[VAL_11:.*]] = memref.load %[[VAL_5]]{{\[}}%[[VAL_2]]] : memref<?xindex>
-// CHECK:           scf.for %[[VAL_12:.*]] = %[[VAL_10]] to %[[VAL_11]] step %[[VAL_2]] {
-// CHECK:             %[[VAL_13:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_12]]] : memref<?xindex>
-// CHECK:             %[[VAL_14:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_12]]] : memref<?xindex>
-// CHECK:             %[[VAL_15:.*]] = arith.addi %[[VAL_12]], %[[VAL_2]] : index
-// CHECK:             %[[VAL_16:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_15]]] : memref<?xindex>
-// CHECK:             scf.for %[[VAL_17:.*]] = %[[VAL_14]] to %[[VAL_16]] step %[[VAL_2]] {
-// CHECK:               %[[VAL_18:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_17]]] : memref<?xindex>
-// CHECK:               %[[VAL_19:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_17]]] : memref<?xf64>
-// CHECK:               sparse_tensor.insert %[[VAL_19]] into %[[VAL_3]]{{\[}}%[[VAL_13]], %[[VAL_18]]] : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_12:.*]] = scf.for %[[VAL_13:.*]] = %[[VAL_10]] to %[[VAL_11]] step %[[VAL_2]] iter_args(%[[VAL_14:.*]] = %[[VAL_3]]) -> (tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:             %[[VAL_15:.*]] = memref.load %[[VAL_6]]{{\[}}%[[VAL_13]]] : memref<?xindex>
+// CHECK:             %[[VAL_16:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_13]]] : memref<?xindex>
+// CHECK:             %[[VAL_17:.*]] = arith.addi %[[VAL_13]], %[[VAL_2]] : index
+// CHECK:             %[[VAL_18:.*]] = memref.load %[[VAL_7]]{{\[}}%[[VAL_17]]] : memref<?xindex>
+// CHECK:             %[[VAL_19:.*]] = scf.for %[[VAL_20:.*]] = %[[VAL_16]] to %[[VAL_18]] step %[[VAL_2]] iter_args(%[[VAL_21:.*]] = %[[VAL_14]]) -> (tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>) {
+// CHECK:               %[[VAL_22:.*]] = memref.load %[[VAL_8]]{{\[}}%[[VAL_20]]] : memref<?xindex>
+// CHECK:               %[[VAL_23:.*]] = memref.load %[[VAL_9]]{{\[}}%[[VAL_20]]] : memref<?xf64>
+// CHECK:               %[[VAL_24:.*]] = sparse_tensor.insert %[[VAL_23]] into %[[VAL_21]]{{\[}}%[[VAL_15]], %[[VAL_22]]] : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:               scf.yield %[[VAL_24]] : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:             }
+// CHECK:             scf.yield %[[VAL_25:.*]] : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           }
-// CHECK:           %[[VAL_20:.*]] = sparse_tensor.load %[[VAL_3]] hasInserts : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           %[[VAL_26:.*]] = sparse_tensor.load %[[VAL_27:.*]] hasInserts : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:           bufferization.dealloc_tensor %[[VAL_4]] : tensor<3x4xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ], dimOrdering = affine_map<(d0, d1) -> (d1, d0)> }>>
-// CHECK:           return %[[VAL_20]] : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
+// CHECK:           return %[[VAL_26]] : tensor<4x3xf64, #sparse_tensor.encoding<{ dimLevelType = [ "compressed", "compressed" ] }>>
 // CHECK:         }
 func.func @sparse_transpose_auto(%arga: tensor<3x4xf64, #DCSR>)
                                      -> tensor<4x3xf64, #DCSR> {
   %i = bufferization.alloc_tensor() : tensor<4x3xf64, #DCSR>
   %0 = linalg.generic #transpose_trait
      ins(%arga: tensor<3x4xf64, #DCSR>)
      outs(%i: tensor<4x3xf64, #DCSR>) {
      ^bb(%a: f64, %x: f64):
        linalg.yield %a : f64
   } -> tensor<4x3xf64, #DCSR>
   return %0 : tensor<4x3xf64, #DCSR>
 }