diff --git a/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp b/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
--- a/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
@@ -385,8 +385,27 @@
         makeCanonicalStridedLayoutExpr(sizes, dimExprs, context);
     resultExprs.push_back(linearizedExpr);
   }
-  return AffineMap::get(sourceMap.getNumDims(), sourceMap.getNumSymbols(),
-                        resultExprs, context);
+  // Number of symbols may change as compared to what is present in the
+  // sourceMap as stride length may be constant or symbolic.
+  unsigned numSymbols = 0;
+  std::function<void(AffineExpr)> updateMaxSymbolPosition =
+      [&](AffineExpr expr) {
+        auto binOpExpr = expr.dyn_cast<AffineBinaryOpExpr>();
+        if (binOpExpr) {
+          updateMaxSymbolPosition(binOpExpr.getLHS());
+          updateMaxSymbolPosition(binOpExpr.getRHS());
+          return;
+        }
+        if (auto symbolExpr = expr.dyn_cast<AffineSymbolExpr>()) {
+          numSymbols = std::max(numSymbols, 1 + symbolExpr.getPosition());
+        }
+      };
+
+  for (AffineExpr expr : resultExprs)
+    updateMaxSymbolPosition(expr);
+
+  return AffineMap::get(sourceMap.getNumDims(), numSymbols, resultExprs,
+                        context);
 }
 
 // TensorExpandShapeOp is fusable with its consumer (i.e. reshape as a