Index: mlir/lib/Transforms/LoopFusion.cpp
===================================================================
--- mlir/lib/Transforms/LoopFusion.cpp
+++ mlir/lib/Transforms/LoopFusion.cpp
@@ -958,6 +958,63 @@
   return newMemRef;
 }
 
+// Walking from node 'srcId' to node 'dstId', if there is a use of 'memref' that
+// is a non-affine operation, return false.
+static bool hasNonAffineUsersOnThePath(unsigned srcId, unsigned dstId,
+                                       Value memref,
+                                       MemRefDependenceGraph *mdg) {
+  auto *srcNode = mdg->getNode(srcId);
+  auto *dstNode = mdg->getNode(dstId);
+  auto users = memref.getUsers();
+  for (auto &idAndNode : mdg->nodes) {
+    auto *op = idAndNode.second.op;
+    // Take care of operations between srcNode and dstNode.
+    if (srcNode->op->isBeforeInBlock(op) && op->isBeforeInBlock(dstNode->op)) {
+      // Walk inside the operation to find any use of the memref.
+      bool found = false;
+      op->walk([&](Operation *user) {
+        // Skip affine ops.
+        if (isMemRefDereferencingOp(*user))
+          return WalkResult::advance();
+        // Find a non-affine op that uses the memref.
+        if (std::find(users.begin(), users.end(), user) != users.end()) {
+          found = true;
+          return WalkResult::interrupt();
+        }
+        return WalkResult::advance();
+      });
+      if (found)
+        return true;
+    }
+  }
+  return false;
+}
+
+// Check whether a value in node 'srcId' has a non-affine consumer that is
+// between node 'srcId' and node 'dstId'.
+static bool hasNonAffineUsersOnThePath(unsigned srcId, unsigned dstId,
+                                       MemRefDependenceGraph *mdg) {
+  // Collect values in node 'srcId'.
+  auto *srcNode = mdg->getNode(srcId);
+  SmallVector<Value, 2> values;
+  srcNode->op->walk([&](Operation *op) {
+    // Skip affine ops
+    if (isa<AffineForOp>(op))
+      return WalkResult::advance();
+    for (Value v : op->getOperands()) {
+      if (std::find(values.begin(), values.end(), v) == values.end())
+        values.push_back(v);
+    }
+    return WalkResult::advance();
+  });
+  // Looking for users between node 'srcId' and node 'dstId'.
+  for (Value v : values) {
+    if (hasNonAffineUsersOnThePath(srcId, dstId, v, mdg))
+      return true;
+  }
+  return false;
+}
+
 // Checks if node 'srcId' can be safely fused into node 'dstId'. Node 'srcId'
 // may write to multiple memrefs but it is required that only one of them,
 // 'srcLiveOutStoreOp', has output edges.
@@ -1018,6 +1075,12 @@
   // TODO(andydavis) Check the shape and lower bounds here too.
   if (srcNumElements != dstNumElements)
     return false;
+
+  // Return false if 'memref' is used by a non-affine operation that is
+  // between node 'srcId' and node 'dstId'.
+  if (hasNonAffineUsersOnThePath(srcId, dstId, mdg))
+    return false;
+
   return true;
 }
 
@@ -1803,6 +1866,12 @@
       }
       if (storeMemrefs.size() != 1)
         return false;
+
+      // Skip if a value in one node is used by a non-affine operation that lies
+      // between 'dstNode' and 'sibNode'.
+      if (hasNonAffineUsersOnThePath(dstNode->id, sibNode->id, mdg) ||
+          hasNonAffineUsersOnThePath(sibNode->id, dstNode->id, mdg))
+        return false;
       return true;
     };
 
Index: mlir/test/Transforms/loop-fusion.mlir
===================================================================
--- mlir/test/Transforms/loop-fusion.mlir
+++ mlir/test/Transforms/loop-fusion.mlir
@@ -2535,3 +2535,67 @@
 // CHECK:        mulf
 // CHECK-NOT:  affine.for
 // CHECK:        divf
+
+// -----
+
+// CHECK-LABEL: func @should_not_fuse_since_non_affine_users
+func @should_not_fuse_since_non_affine_users(%in0 : memref<32xf32>,
+                      %in1 : memref<32xf32>) {
+  affine.for %d = 0 to 32 {
+    %lhs = affine.load %in0[%d] : memref<32xf32>
+    %rhs = affine.load %in1[%d] : memref<32xf32>
+    %add = addf %lhs, %rhs : f32
+    affine.store %add, %in0[%d] : memref<32xf32>
+  }
+  affine.for %d = 0 to 32 {
+    %lhs = load %in0[%d] : memref<32xf32>
+    %rhs = load %in1[%d] : memref<32xf32>
+    %add = subf %lhs, %rhs : f32
+    store %add, %in0[%d] : memref<32xf32>
+  }
+  affine.for %d = 0 to 32 {
+    %lhs = affine.load %in0[%d] : memref<32xf32>
+    %rhs = affine.load %in1[%d] : memref<32xf32>
+    %add = mulf %lhs, %rhs : f32
+    affine.store %add, %in0[%d] : memref<32xf32>
+  }
+  return
+}
+
+// CHECK:  affine.for
+// CHECK:    addf
+// CHECK:  affine.for
+// CHECK:    subf
+// CHECK:  affine.for
+// CHECK:    mulf
+
+// -----
+
+// CHECK-LABEL: func @should_not_fuse_since_top_level_non_affine_users
+func @should_not_fuse_since_top_level_non_affine_users(%in0 : memref<32xf32>,
+                      %in1 : memref<32xf32>) {
+  %sum = alloc() : memref<f32>
+  affine.for %d = 0 to 32 {
+    %lhs = affine.load %in0[%d] : memref<32xf32>
+    %rhs = affine.load %in1[%d] : memref<32xf32>
+    %add = addf %lhs, %rhs : f32
+    store %add, %sum[] : memref<f32>
+    affine.store %add, %in0[%d] : memref<32xf32>
+  }
+  %load_sum = load %sum[] : memref<f32>
+  affine.for %d = 0 to 32 {
+    %lhs = affine.load %in0[%d] : memref<32xf32>
+    %rhs = affine.load %in1[%d] : memref<32xf32>
+    %add = mulf %lhs, %rhs : f32
+    %sub = subf %add, %load_sum: f32
+    affine.store %sub, %in0[%d] : memref<32xf32>
+  }
+  dealloc %sum : memref<f32>
+  return
+}
+
+// CHECK:  affine.for
+// CHECK:    addf
+// CHECK:  affine.for
+// CHECK:    mulf
+// CHECK:    subf