diff --git a/mlir/lib/Dialect/Arith/Utils/CMakeLists.txt b/mlir/lib/Dialect/Arith/Utils/CMakeLists.txt
--- a/mlir/lib/Dialect/Arith/Utils/CMakeLists.txt
+++ b/mlir/lib/Dialect/Arith/Utils/CMakeLists.txt
@@ -6,6 +6,7 @@
 
   LINK_LIBS PUBLIC
   MLIRArithDialect
+  MLIRComplexDialect
   MLIRDialect
   MLIRIR
   )
diff --git a/mlir/lib/Dialect/Arith/Utils/Utils.cpp b/mlir/lib/Dialect/Arith/Utils/Utils.cpp
--- a/mlir/lib/Dialect/Arith/Utils/Utils.cpp
+++ b/mlir/lib/Dialect/Arith/Utils/Utils.cpp
@@ -12,6 +12,7 @@
 
 #include "mlir/Dialect/Arith/Utils/Utils.h"
 #include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Complex/IR/Complex.h"
 #include "llvm/ADT/SmallBitVector.h"
 
 using namespace mlir;
@@ -122,7 +123,57 @@
       if (toFloatType.getWidth() < fromFloatType.getWidth())
         return b.create<arith::TruncFOp>(loc, toFloatType, operand);
     }
+  } else if (auto toComplexType = dyn_cast<ComplexType>(toType)) {
+    if (auto fromComplexType = dyn_cast<ComplexType>(operand.getType())) {
+      if (isa<FloatType>(toComplexType.getElementType()) &&
+          isa<FloatType>(fromComplexType.getElementType())) {
+        Value real = b.create<complex::ReOp>(loc, operand);
+        Value imag = b.create<complex::ImOp>(loc, operand);
+        if (toComplexType.getElementType().getIntOrFloatBitWidth() <
+            fromComplexType.getElementType().getIntOrFloatBitWidth()) {
+          real = b.create<arith::TruncFOp>(loc, toComplexType.getElementType(),
+                                           real);
+          imag = b.create<arith::TruncFOp>(loc, toComplexType.getElementType(),
+                                           imag);
+        } else {
+          real = b.create<arith::ExtFOp>(loc, toComplexType.getElementType(),
+                                         real);
+          imag = b.create<arith::ExtFOp>(loc, toComplexType.getElementType(),
+                                         imag);
+        }
+        return b.create<complex::CreateOp>(loc, toComplexType, real, imag);
+      }
+    }
+
+    if (auto fromFpType = dyn_cast<FloatType>(operand.getType())) {
+      FloatType toFloatTy = toComplexType.getElementType().cast<FloatType>();
+      int32_t toBitwidth = toFloatTy.getIntOrFloatBitWidth();
+      Value from = operand;
+      if (from.getType().getIntOrFloatBitWidth() < toBitwidth) {
+        from = b.create<arith::ExtFOp>(loc, toFloatTy, from);
+      }
+      if (from.getType().getIntOrFloatBitWidth() > toBitwidth) {
+        from = b.create<arith::TruncFOp>(loc, toFloatTy, from);
+      }
+      Value zero = b.create<mlir::arith::ConstantFloatOp>(
+          loc, mlir::APFloat(toFloatTy.getFloatSemantics(), 0), toFloatTy);
+      return b.create<complex::CreateOp>(loc, toComplexType, from, zero);
+    }
+
+    if (auto fromIntType = dyn_cast<IntegerType>(operand.getType())) {
+      FloatType toFloatTy = toComplexType.getElementType().cast<FloatType>();
+      Value from = operand;
+      if (isUnsignedCast) {
+        from = b.create<arith::UIToFPOp>(loc, toFloatTy, from);
+      } else {
+        from = b.create<arith::SIToFPOp>(loc, toFloatTy, from);
+      }
+      Value zero = b.create<mlir::arith::ConstantFloatOp>(
+          loc, mlir::APFloat(toFloatTy.getFloatSemantics(), 0), toFloatTy);
+      return b.create<complex::CreateOp>(loc, toComplexType, from, zero);
+    }
   }
+
   emitWarning(loc) << "could not cast operand of type " << operand.getType()
                    << " to " << toType;
   return operand;