diff --git a/kernels/portable/cpu/op_bitwise_and.cpp b/kernels/portable/cpu/op_bitwise_and.cpp
index b1078f780a4..fa840f7e5cc 100644
--- a/kernels/portable/cpu/op_bitwise_and.cpp
+++ b/kernels/portable/cpu/op_bitwise_and.cpp
@@ -6,8 +6,10 @@
  * LICENSE file in the root directory of this source tree.
  */
 
-#include <cmath>
+// patternlint-disable-next-line executorch-cpp-nostdinc
+#include <functional>
 
+#include <executorch/kernels/portable/cpu/pattern/bitwise_op.h>
 #include <executorch/kernels/portable/cpu/scalar_utils.h>
 #include <executorch/kernels/portable/cpu/util/broadcast_util.h>
 #include <executorch/kernels/portable/cpu/util/functional_util.h>
@@ -17,20 +19,6 @@ namespace torch {
 namespace executor {
 namespace native {
 
-namespace {
-
-template <typename CTYPE>
-CTYPE bitwise_and(CTYPE a, CTYPE b) {
-  return a & b;
-}
-
-template <>
-bool bitwise_and<bool>(bool a, bool b) {
-  return a && b;
-}
-
-} // namespace
-
 using Tensor = exec_aten::Tensor;
 
 Tensor& bitwise_and_Tensor_out(
@@ -38,60 +26,7 @@ Tensor& bitwise_and_Tensor_out(
     const Tensor& a,
     const Tensor& b,
     Tensor& out) {
-  ET_KERNEL_CHECK(
-      ctx,
-      resize_to_broadcast_target_size(a, b, out) == Error::Ok,
-      InvalidArgument,
-      out);
-
-  ScalarType a_type = a.scalar_type();
-  ScalarType b_type = b.scalar_type();
-  ScalarType common_type = promoteTypes(a_type, b_type);
-  ScalarType out_type = out.scalar_type();
-
-  ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);
-
-  ET_SWITCH_INT_TYPES_AND(
-      Bool, a_type, ctx, "bitwise_and.Tensor_out", CTYPE_A, [&]() {
-        ET_SWITCH_INT_TYPES_AND(
-            Bool, b_type, ctx, "bitwise_and.Tensor_out", CTYPE_B, [&]() {
-              ET_SWITCH_INT_TYPES_AND(
-                  Bool,
-                  common_type,
-                  ctx,
-                  "bitwise_and.Tensor_out",
-                  CTYPE_IN,
-                  [&]() {
-                    ET_SWITCH_REAL_TYPES_AND(
-                        Bool,
-                        out_type,
-                        ctx,
-                        "bitwise_and.Tensor_out",
-                        CTYPE_OUT,
-                        [&]() {
-                          apply_binary_elementwise_fn<
-                              CTYPE_A,
-                              CTYPE_B,
-                              CTYPE_OUT>(
-                              [](const CTYPE_A val_a, const CTYPE_B val_b) {
-                                CTYPE_IN a_casted =
-                                    static_cast<CTYPE_IN>(val_a);
-                                CTYPE_IN b_casted =
-                                    static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value =
-                                    bitwise_and(a_casted, b_casted);
-
-                                return static_cast<CTYPE_OUT>(value);
-                              },
-                              a,
-                              b,
-                              out);
-                        });
-                  });
-            });
-      });
-
-  return out;
+  return bitwise_op_out<std::bit_and>(ctx, a, b, out, "bitwise_and.Tensor_out");
 }
 
 Tensor& bitwise_and_Scalar_out(
@@ -142,8 +77,8 @@ Tensor& bitwise_and_Scalar_out(
                                     static_cast<CTYPE_IN>(val_a);
                                 CTYPE_IN b_casted =
                                     static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value =
-                                    bitwise_and(a_casted, b_casted);
+                                CTYPE_IN value = std::bit_and<CTYPE_IN>()(
+                                    a_casted, b_casted);
 
                                 return static_cast<CTYPE_OUT>(value);
                               },
diff --git a/kernels/portable/cpu/op_bitwise_or.cpp b/kernels/portable/cpu/op_bitwise_or.cpp
index c13c68d3db4..e42632b71f2 100644
--- a/kernels/portable/cpu/op_bitwise_or.cpp
+++ b/kernels/portable/cpu/op_bitwise_or.cpp
@@ -6,8 +6,10 @@
  * LICENSE file in the root directory of this source tree.
  */
 
-#include <cmath>
+// patternlint-disable-next-line executorch-cpp-nostdinc
+#include <functional>
 
+#include <executorch/kernels/portable/cpu/pattern/bitwise_op.h>
 #include <executorch/kernels/portable/cpu/scalar_utils.h>
 #include <executorch/kernels/portable/cpu/util/broadcast_util.h>
 #include <executorch/kernels/portable/cpu/util/functional_util.h>
@@ -17,20 +19,6 @@ namespace torch {
 namespace executor {
 namespace native {
 
-namespace {
-
-template <typename CTYPE>
-CTYPE bitwise_or(CTYPE a, CTYPE b) {
-  return a | b;
-}
-
-template <>
-bool bitwise_or<bool>(bool a, bool b) {
-  return a || b;
-}
-
-} // namespace
-
 using Tensor = exec_aten::Tensor;
 
 Tensor& bitwise_or_Tensor_out(
@@ -38,59 +26,7 @@ Tensor& bitwise_or_Tensor_out(
     const Tensor& a,
     const Tensor& b,
     Tensor& out) {
-  ET_KERNEL_CHECK(
-      ctx,
-      resize_to_broadcast_target_size(a, b, out) == Error::Ok,
-      InvalidArgument,
-      out);
-
-  ScalarType a_type = a.scalar_type();
-  ScalarType b_type = b.scalar_type();
-  ScalarType common_type = promoteTypes(a_type, b_type);
-  ScalarType out_type = out.scalar_type();
-
-  ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);
-
-  ET_SWITCH_INT_TYPES_AND(
-      Bool, a_type, ctx, "bitwise_or.Tensor_out", CTYPE_A, [&]() {
-        ET_SWITCH_INT_TYPES_AND(
-            Bool, b_type, ctx, "bitwise_or.Tensor_out", CTYPE_B, [&]() {
-              ET_SWITCH_INT_TYPES_AND(
-                  Bool,
-                  common_type,
-                  ctx,
-                  "bitwise_or.Tensor_out",
-                  CTYPE_IN,
-                  [&]() {
-                    ET_SWITCH_REAL_TYPES_AND(
-                        Bool,
-                        out_type,
-                        ctx,
-                        "bitwise_or.Tensor_out",
-                        CTYPE_OUT,
-                        [&]() {
-                          apply_binary_elementwise_fn<
-                              CTYPE_A,
-                              CTYPE_B,
-                              CTYPE_OUT>(
-                              [](const CTYPE_A val_a, const CTYPE_B val_b) {
-                                CTYPE_IN a_casted =
-                                    static_cast<CTYPE_IN>(val_a);
-                                CTYPE_IN b_casted =
-                                    static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value = bitwise_or(a_casted, b_casted);
-
-                                return static_cast<CTYPE_OUT>(value);
-                              },
-                              a,
-                              b,
-                              out);
-                        });
-                  });
-            });
-      });
-
-  return out;
+  return bitwise_op_out<std::bit_or>(ctx, a, b, out, "bitwise_or.Tensor_out");
 }
 
 Tensor& bitwise_or_Scalar_out(
@@ -141,7 +77,8 @@ Tensor& bitwise_or_Scalar_out(
                                     static_cast<CTYPE_IN>(val_a);
                                 CTYPE_IN b_casted =
                                     static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value = bitwise_or(a_casted, b_casted);
+                                CTYPE_IN value =
+                                    std::bit_or<CTYPE_IN>()(a_casted, b_casted);
 
                                 return static_cast<CTYPE_OUT>(value);
                               },
diff --git a/kernels/portable/cpu/op_bitwise_xor.cpp b/kernels/portable/cpu/op_bitwise_xor.cpp
index d2ea8a81cfb..c59cdcc692e 100644
--- a/kernels/portable/cpu/op_bitwise_xor.cpp
+++ b/kernels/portable/cpu/op_bitwise_xor.cpp
@@ -6,8 +6,10 @@
  * LICENSE file in the root directory of this source tree.
  */
 
-#include <cmath>
+// patternlint-disable-next-line executorch-cpp-nostdinc
+#include <functional>
 
+#include <executorch/kernels/portable/cpu/pattern/bitwise_op.h>
 #include <executorch/kernels/portable/cpu/scalar_utils.h>
 #include <executorch/kernels/portable/cpu/util/broadcast_util.h>
 #include <executorch/kernels/portable/cpu/util/functional_util.h>
@@ -17,20 +19,6 @@ namespace torch {
 namespace executor {
 namespace native {
 
-namespace {
-
-template <typename CTYPE>
-CTYPE bitwise_xor(CTYPE a, CTYPE b) {
-  return a ^ b;
-}
-
-template <>
-bool bitwise_xor<bool>(bool a, bool b) {
-  return a != b;
-}
-
-} // namespace
-
 using Tensor = exec_aten::Tensor;
 
 Tensor& bitwise_xor_Tensor_out(
@@ -38,61 +26,7 @@ Tensor& bitwise_xor_Tensor_out(
     const Tensor& a,
     const Tensor& b,
     Tensor& out) {
-  // Determine output size and resize for dynamic shapes
-  ET_KERNEL_CHECK(
-      ctx,
-      resize_to_broadcast_target_size(a, b, out) == Error::Ok,
-      InvalidArgument,
-      out);
-
-  ScalarType a_type = a.scalar_type();
-  ScalarType b_type = b.scalar_type();
-  ScalarType common_type = promoteTypes(a_type, b_type);
-  ScalarType out_type = out.scalar_type();
-
-  ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);
-
-  ET_SWITCH_INT_TYPES_AND(
-      Bool, a_type, ctx, "bitwise_xor.Tensor_out", CTYPE_A, [&]() {
-        ET_SWITCH_INT_TYPES_AND(
-            Bool, b_type, ctx, "bitwise_xor.Tensor_out", CTYPE_B, [&]() {
-              ET_SWITCH_INT_TYPES_AND(
-                  Bool,
-                  common_type,
-                  ctx,
-                  "bitwise_xor.Tensor_out",
-                  CTYPE_IN,
-                  [&]() {
-                    ET_SWITCH_REAL_TYPES_AND(
-                        Bool,
-                        out_type,
-                        ctx,
-                        "bitwise_xor.Tensor_out",
-                        CTYPE_OUT,
-                        [&]() {
-                          apply_binary_elementwise_fn<
-                              CTYPE_A,
-                              CTYPE_B,
-                              CTYPE_OUT>(
-                              [](const CTYPE_A val_a, const CTYPE_B val_b) {
-                                CTYPE_IN a_casted =
-                                    static_cast<CTYPE_IN>(val_a);
-                                CTYPE_IN b_casted =
-                                    static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value =
-                                    bitwise_xor(a_casted, b_casted);
-
-                                return static_cast<CTYPE_OUT>(value);
-                              },
-                              a,
-                              b,
-                              out);
-                        });
-                  });
-            });
-      });
-
-  return out;
+  return bitwise_op_out<std::bit_xor>(ctx, a, b, out, "bitwise_xor.Tensor_out");
 }
 
 Tensor& bitwise_xor_Scalar_out(
@@ -143,8 +77,8 @@ Tensor& bitwise_xor_Scalar_out(
                                     static_cast<CTYPE_IN>(val_a);
                                 CTYPE_IN b_casted =
                                     static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value =
-                                    bitwise_xor(a_casted, b_casted);
+                                CTYPE_IN value = std::bit_xor<CTYPE_IN>()(
+                                    a_casted, b_casted);
 
                                 return static_cast<CTYPE_OUT>(value);
                               },
diff --git a/kernels/portable/cpu/op_clamp.cpp b/kernels/portable/cpu/op_clamp.cpp
index 06c87d03f2d..50d7e8c374d 100644
--- a/kernels/portable/cpu/op_clamp.cpp
+++ b/kernels/portable/cpu/op_clamp.cpp
@@ -53,7 +53,7 @@ __ET_NODISCARD bool check_bounds(
         }
       });
     } else if (isFloatingType(out_type)) {
-      ET_SWITCH_FLOAT_TYPES(out_type, ctx, "clamp", CTYPE_OUT, [&]() {
+      ET_SWITCH_FLOATH_TYPES(out_type, ctx, "clamp", CTYPE_OUT, [&]() {
         if (std::isfinite(val) &&
             is_out_of_bounds<CTYPE_VAL, CTYPE_OUT, double>(val)) {
           ET_LOG(Error, "%s value out of bounds", val_name);
@@ -119,7 +119,7 @@ Tensor& clamp_out(
 
   ET_KERNEL_CHECK(ctx, common_type == out_type, InvalidArgument, out);
 
-  ET_SWITCH_REAL_TYPES(out_type, ctx, "clamp", CTYPE_OUT, [&]() {
+  ET_SWITCH_REALH_TYPES(out_type, ctx, "clamp", CTYPE_OUT, [&]() {
     // Extract optional min value
     CTYPE_OUT min = 0;
     if (has_min) {
@@ -140,7 +140,7 @@ Tensor& clamp_out(
       });
     }
 
-    ET_SWITCH_REAL_TYPES_AND(Bool, in_type, ctx, "clamp", CTYPE_IN, [&]() {
+    ET_SWITCH_REALHB_TYPES(in_type, ctx, "clamp", CTYPE_IN, [&]() {
       apply_unary_map_fn(
           [has_min, min, has_max, max](const CTYPE_IN val_in) {
             CTYPE_OUT val_out = static_cast<CTYPE_OUT>(val_in);
@@ -195,20 +195,20 @@ Tensor& clamp_tensor_out(
   ScalarType out_type = out.scalar_type();
 
   if (has_min) {
-    common_type = promoteTypes(common_type, min_type);
+    common_type = promoteTypes(common_type, min_type, /*half_to_float*/ true);
   }
   if (has_max) {
-    common_type = promoteTypes(common_type, max_type);
+    common_type = promoteTypes(common_type, max_type, /*half_to_float*/ true);
   }
 
   ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);
 
   constexpr auto name = "clamp.Tensor_out";
 
-  ET_SWITCH_REALB_TYPES(in_type, ctx, name, CTYPE_IN, [&]() {
-    ET_SWITCH_REALB_TYPES(min_type, ctx, name, CTYPE_MIN, [&]() {
-      ET_SWITCH_REALB_TYPES(max_type, ctx, name, CTYPE_MAX, [&]() {
-        ET_SWITCH_REALB_TYPES(out_type, ctx, name, CTYPE_OUT, [&]() {
+  ET_SWITCH_REALHB_TYPES(in_type, ctx, name, CTYPE_IN, [&]() {
+    ET_SWITCH_REALHB_TYPES(min_type, ctx, name, CTYPE_MIN, [&]() {
+      ET_SWITCH_REALHB_TYPES(max_type, ctx, name, CTYPE_MAX, [&]() {
+        ET_SWITCH_REALHB_TYPES(out_type, ctx, name, CTYPE_OUT, [&]() {
           apply_ternary_elementwise_fn<
               CTYPE_IN,
               CTYPE_MIN,
diff --git a/kernels/portable/cpu/op_floor_divide.cpp b/kernels/portable/cpu/op_floor_divide.cpp
index 261f77ce617..0514df0ca25 100644
--- a/kernels/portable/cpu/op_floor_divide.cpp
+++ b/kernels/portable/cpu/op_floor_divide.cpp
@@ -20,6 +20,60 @@ namespace native {
 using Tensor = exec_aten::Tensor;
 using ScalarType = exec_aten::ScalarType;
 
+namespace {
+template <
+    bool can_cast,
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct FloorDivideInner;
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct FloorDivideInner<true, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT> {
+  static void
+  run(const Tensor& a, const Tensor& b, Tensor& out, bool& div_by_zero_error) {
+    apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
+        // NOLINTNEXTLINE(facebook-hte-ConstantArgumentPassByValue)
+        [&div_by_zero_error](const CTYPE_A val_a, const CTYPE_B val_b) {
+          if (is_integral_type<CTYPE_IN, /*includeBool=*/true>::value) {
+            if (val_b == 0) {
+              div_by_zero_error = true;
+              return static_cast<CTYPE_OUT>(0);
+            }
+          }
+          CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
+          CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
+          CTYPE_IN value = utils::floor_divide<CTYPE_IN>(a_casted, b_casted);
+
+          return static_cast<CTYPE_OUT>(value);
+        },
+        a,
+        b,
+        out);
+  }
+};
+
+struct ReportCanCastBug {
+  static void run(const Tensor&, const Tensor&, Tensor&, bool&) {
+    ET_DCHECK_MSG(false, "BUG: canCast should have been checked above");
+  }
+};
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct FloorDivideInner<false, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT>
+    : public ReportCanCastBug {};
+
+} // namespace
+
 Tensor& floor_divide_out(
     RuntimeContext& ctx,
     const Tensor& a,
@@ -46,36 +100,17 @@ Tensor& floor_divide_out(
       Bool, a_type, ctx, "floor_divide.out", CTYPE_A, [&]() {
         ET_SWITCH_REAL_TYPES_AND(
             Bool, b_type, ctx, "floor_divide.out", CTYPE_B, [&]() {
+              using CTYPE_IN = typename torch::executor::
+                  promote_types<CTYPE_A, CTYPE_B>::type;
+              ET_DCHECK(CppTypeToScalarType<CTYPE_IN>::value == common_type);
               ET_SWITCH_REAL_TYPES(
-                  common_type, ctx, "floor_divide.out", CTYPE_IN, [&]() {
-                    ET_SWITCH_REAL_TYPES(
-                        out_type, ctx, "floor_divide.out", CTYPE_OUT, [&]() {
-                          apply_binary_elementwise_fn<
-                              CTYPE_A,
-                              CTYPE_B,
-                              CTYPE_OUT>(
-                              [common_type, &div_by_zero_error](
-                                  const CTYPE_A val_a, const CTYPE_B val_b) {
-                                if (isIntegralType(
-                                        common_type, /*includeBool=*/true)) {
-                                  if (val_b == 0) {
-                                    div_by_zero_error = true;
-                                    return static_cast<CTYPE_OUT>(0);
-                                  }
-                                }
-                                CTYPE_IN a_casted =
-                                    static_cast<CTYPE_IN>(val_a);
-                                CTYPE_IN b_casted =
-                                    static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value = utils::floor_divide<CTYPE_IN>(
-                                    a_casted, b_casted);
-
-                                return static_cast<CTYPE_OUT>(value);
-                              },
-                              a,
-                              b,
-                              out);
-                        });
+                  out_type, ctx, "floor_divide.out", CTYPE_OUT, [&]() {
+                    FloorDivideInner<
+                        can_cast<CTYPE_IN, CTYPE_OUT>::value,
+                        CTYPE_A,
+                        CTYPE_B,
+                        CTYPE_IN,
+                        CTYPE_OUT>::run(a, b, out, div_by_zero_error);
                   });
             });
       });
diff --git a/kernels/portable/cpu/op_fmod.cpp b/kernels/portable/cpu/op_fmod.cpp
index 0083c1379d5..42f83731199 100644
--- a/kernels/portable/cpu/op_fmod.cpp
+++ b/kernels/portable/cpu/op_fmod.cpp
@@ -19,6 +19,60 @@ namespace native {
 
 using Tensor = exec_aten::Tensor;
 
+namespace {
+template <
+    bool can_cast,
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct FmodInner;
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct FmodInner<true, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT> {
+  static void
+  run(const Tensor& a, const Tensor& b, Tensor& out, bool& div_by_zero_error) {
+    apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
+        // NOLINTNEXTLINE(facebook-hte-ConstantArgumentPassByValue)
+        [&div_by_zero_error](const CTYPE_A val_a, const CTYPE_B val_b) {
+          if (is_integral_type<CTYPE_IN, /*includeBool=*/true>::value) {
+            if (val_b == 0) {
+              div_by_zero_error = true;
+              return static_cast<CTYPE_OUT>(0);
+            }
+          }
+          CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
+          CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
+          CTYPE_IN value = std::fmod(a_casted, b_casted);
+
+          return static_cast<CTYPE_OUT>(value);
+        },
+        a,
+        b,
+        out);
+  }
+};
+
+struct ReportCanCastBug {
+  static void run(const Tensor&, const Tensor&, Tensor&, bool&) {
+    ET_DCHECK_MSG(false, "BUG: canCast should have been checked above");
+  }
+};
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct FmodInner<false, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT>
+    : public ReportCanCastBug {};
+
+} // namespace
+
 Tensor& fmod_Tensor_out(
     RuntimeContext& ctx,
     const Tensor& a,
@@ -44,35 +98,18 @@ Tensor& fmod_Tensor_out(
       Bool, a_type, ctx, "fmod.Tensor_out", CTYPE_A, [&]() {
         ET_SWITCH_REAL_TYPES_AND(
             Bool, b_type, ctx, "fmod.Tensor_out", CTYPE_B, [&]() {
+              using CTYPE_IN = typename torch::executor::
+                  promote_types<CTYPE_A, CTYPE_B>::type;
+              ET_DCHECK(CppTypeToScalarType<CTYPE_IN>::value == common_type);
               ET_SWITCH_REAL_TYPES(
-                  common_type, ctx, "fmod.Tensor_out", CTYPE_IN, [&]() {
-                    ET_SWITCH_REAL_TYPES(
-                        out_type, ctx, "fmod.Tensor_out", CTYPE_OUT, [&]() {
-                          apply_binary_elementwise_fn<
-                              CTYPE_A,
-                              CTYPE_B,
-                              CTYPE_OUT>(
-                              [common_type, &div_by_zero_error](
-                                  const CTYPE_A val_a, const CTYPE_B val_b) {
-                                if (isIntegralType(
-                                        common_type, /*includeBool=*/true)) {
-                                  if (val_b == 0) {
-                                    div_by_zero_error = true;
-                                    return static_cast<CTYPE_OUT>(0);
-                                  }
-                                }
-                                CTYPE_IN a_casted =
-                                    static_cast<CTYPE_IN>(val_a);
-                                CTYPE_IN b_casted =
-                                    static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value = std::fmod(a_casted, b_casted);
-
-                                return static_cast<CTYPE_OUT>(value);
-                              },
-                              a,
-                              b,
-                              out);
-                        });
+                  out_type, ctx, "fmod.Tensor_out", CTYPE_OUT, [&]() {
+                    FmodInner<
+                        !std::is_same<CTYPE_IN, bool>::value &&
+                            can_cast<CTYPE_IN, CTYPE_OUT>::value,
+                        CTYPE_A,
+                        CTYPE_B,
+                        CTYPE_IN,
+                        CTYPE_OUT>::run(a, b, out, div_by_zero_error);
                   });
             });
       });
diff --git a/kernels/portable/cpu/op_maximum.cpp b/kernels/portable/cpu/op_maximum.cpp
index 3e34035d5f6..1353479b294 100644
--- a/kernels/portable/cpu/op_maximum.cpp
+++ b/kernels/portable/cpu/op_maximum.cpp
@@ -8,6 +8,7 @@
 
 #include <executorch/kernels/portable/cpu/scalar_utils.h>
 #include <executorch/kernels/portable/cpu/util/broadcast_util.h>
+#include <executorch/kernels/portable/cpu/util/math_util.h>
 #include <executorch/runtime/kernel/kernel_includes.h>
 
 namespace torch {
@@ -15,10 +16,49 @@ namespace executor {
 namespace native {
 namespace {
 
-template <class T>
-const T& max(const T& a, const T& b) {
-  return (b > a) ? b : a;
-}
+template <
+    bool can_cast,
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct MaximumInner;
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct MaximumInner<true, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT> {
+  static void run(const Tensor& a, const Tensor& b, Tensor& out) {
+    apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
+        // NOLINTNEXTLINE(facebook-hte-ConstantArgumentPassByValue)
+        [](const CTYPE_A val_a, const CTYPE_B val_b) {
+          CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
+          CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
+          CTYPE_IN value = utils::max_override(a_casted, b_casted);
+
+          return static_cast<CTYPE_OUT>(value);
+        },
+        a,
+        b,
+        out);
+  }
+};
+
+struct ReportCanCastBug {
+  static void run(const Tensor&, const Tensor&, Tensor&) {
+    ET_DCHECK_MSG(false, "BUG: canCast should have been checked above");
+  }
+};
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct MaximumInner<false, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT>
+    : public ReportCanCastBug {};
 
 } // namespace
 
@@ -44,20 +84,16 @@ Tensor& maximum_out(
 
   ET_SWITCH_REALHB_TYPES(a_type, ctx, "maximum.out", CTYPE_A, [&]() {
     ET_SWITCH_REALHB_TYPES(b_type, ctx, "maximum.out", CTYPE_B, [&]() {
-      ET_SWITCH_REALB_TYPES(common_type, ctx, "maximum.out", CTYPE_IN, [&]() {
-        ET_SWITCH_REALHB_TYPES(out_type, ctx, "maximum.out", CTYPE_OUT, [&]() {
-          apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
-              [](const CTYPE_A val_a, const CTYPE_B val_b) {
-                CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
-                CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
-                CTYPE_IN value = max(a_casted, b_casted);
-
-                return static_cast<CTYPE_OUT>(value);
-              },
-              a,
-              b,
-              out);
-        });
+      using CTYPE_IN = typename torch::executor::
+          promote_types<CTYPE_A, CTYPE_B, /*half_to_float*/ true>::type;
+      ET_DCHECK(CppTypeToScalarType<CTYPE_IN>::value == common_type);
+      ET_SWITCH_REALHB_TYPES(out_type, ctx, "maximum.out", CTYPE_OUT, [&]() {
+        MaximumInner<
+            can_cast<CTYPE_IN, CTYPE_OUT>::value,
+            CTYPE_A,
+            CTYPE_B,
+            CTYPE_IN,
+            CTYPE_OUT>::run(a, b, out);
       });
     });
   });
diff --git a/kernels/portable/cpu/op_minimum.cpp b/kernels/portable/cpu/op_minimum.cpp
index 767a2c4ca59..f18d1a6d368 100644
--- a/kernels/portable/cpu/op_minimum.cpp
+++ b/kernels/portable/cpu/op_minimum.cpp
@@ -8,6 +8,7 @@
 
 #include <executorch/kernels/portable/cpu/scalar_utils.h>
 #include <executorch/kernels/portable/cpu/util/broadcast_util.h>
+#include <executorch/kernels/portable/cpu/util/math_util.h>
 #include <executorch/runtime/kernel/kernel_includes.h>
 
 namespace torch {
@@ -15,10 +16,49 @@ namespace executor {
 namespace native {
 namespace {
 
-template <class T>
-const T& min(const T& a, const T& b) {
-  return (b < a) ? b : a;
-}
+template <
+    bool can_cast,
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct MinimumInner;
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct MinimumInner<true, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT> {
+  static void run(const Tensor& a, const Tensor& b, Tensor& out) {
+    apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
+        // NOLINTNEXTLINE(facebook-hte-ConstantArgumentPassByValue)
+        [](const CTYPE_A val_a, const CTYPE_B val_b) {
+          CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
+          CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
+          CTYPE_IN value = utils::min_override(a_casted, b_casted);
+
+          return static_cast<CTYPE_OUT>(value);
+        },
+        a,
+        b,
+        out);
+  }
+};
+
+struct ReportCanCastBug {
+  static void run(const Tensor&, const Tensor&, Tensor&) {
+    ET_DCHECK_MSG(false, "BUG: canCast should have been checked above");
+  }
+};
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct MinimumInner<false, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT>
+    : public ReportCanCastBug {};
 
 } // namespace
 
@@ -37,30 +77,24 @@ Tensor& minimum_out(
 
   ScalarType a_type = a.scalar_type();
   ScalarType b_type = b.scalar_type();
-  ScalarType common_type = promoteTypes(a_type, b_type);
+  ScalarType common_type = promoteTypes(a_type, b_type, /*half_to_float*/ true);
   ScalarType out_type = out.scalar_type();
 
   ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);
 
-  ET_SWITCH_REAL_TYPES_AND(Bool, a_type, ctx, "minimum.out", CTYPE_A, [&]() {
-    ET_SWITCH_REAL_TYPES_AND(Bool, b_type, ctx, "minimum.out", CTYPE_B, [&]() {
-      ET_SWITCH_REAL_TYPES_AND(
-          Bool, common_type, ctx, "minimum.out", CTYPE_IN, [&]() {
-            ET_SWITCH_REAL_TYPES_AND(
-                Bool, out_type, ctx, "minimum.out", CTYPE_OUT, [&]() {
-                  apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
-                      [](const CTYPE_A val_a, const CTYPE_B val_b) {
-                        CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
-                        CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
-                        CTYPE_IN value = min(a_casted, b_casted);
-
-                        return static_cast<CTYPE_OUT>(value);
-                      },
-                      a,
-                      b,
-                      out);
-                });
-          });
+  ET_SWITCH_REALHB_TYPES(a_type, ctx, "minimum.out", CTYPE_A, [&]() {
+    ET_SWITCH_REALHB_TYPES(b_type, ctx, "minimum.out", CTYPE_B, [&]() {
+      using CTYPE_IN = typename torch::executor::
+          promote_types<CTYPE_A, CTYPE_B, /*half_to_float*/ true>::type;
+      ET_DCHECK(CppTypeToScalarType<CTYPE_IN>::value == common_type);
+      ET_SWITCH_REALHB_TYPES(out_type, ctx, "minimum.out", CTYPE_OUT, [&]() {
+        MinimumInner<
+            can_cast<CTYPE_IN, CTYPE_OUT>::value,
+            CTYPE_A,
+            CTYPE_B,
+            CTYPE_IN,
+            CTYPE_OUT>::run(a, b, out);
+      });
     });
   });
 
diff --git a/kernels/portable/cpu/op_remainder.cpp b/kernels/portable/cpu/op_remainder.cpp
index 9e48374a81a..7c858c1c08a 100644
--- a/kernels/portable/cpu/op_remainder.cpp
+++ b/kernels/portable/cpu/op_remainder.cpp
@@ -20,6 +20,52 @@ namespace native {
 
 using Tensor = exec_aten::Tensor;
 
+namespace {
+template <
+    bool can_cast,
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct RemainderInner;
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct RemainderInner<true, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT> {
+  static void run(const Tensor& a, const Tensor& b, Tensor& out) {
+    apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
+        // NOLINTNEXTLINE(facebook-hte-ConstantArgumentPassByValue)
+        [](const CTYPE_A val_a, const CTYPE_B val_b) {
+          CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
+          CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
+          CTYPE_IN value = utils::remainder_override(a_casted, b_casted);
+
+          return static_cast<CTYPE_OUT>(value);
+        },
+        a,
+        b,
+        out);
+  }
+};
+
+struct ReportCanCastBug {
+  static void run(const Tensor&, const Tensor&, Tensor&) {
+    ET_DCHECK_MSG(false, "BUG: canCast should have been checked above");
+  }
+};
+
+template <
+    typename CTYPE_A,
+    typename CTYPE_B,
+    typename CTYPE_IN,
+    typename CTYPE_OUT>
+struct RemainderInner<false, CTYPE_A, CTYPE_B, CTYPE_IN, CTYPE_OUT>
+    : public ReportCanCastBug {};
+
+} // namespace
 Tensor& remainder_Tensor_out(
     RuntimeContext& ctx,
     const Tensor& a,
@@ -45,32 +91,17 @@ Tensor& remainder_Tensor_out(
       Bool, a_type, ctx, "remainder.Tensor_out", CTYPE_A, [&]() {
         ET_SWITCH_REAL_TYPES_AND(
             Bool, b_type, ctx, "remainder.Tensor_out", CTYPE_B, [&]() {
+              using CTYPE_IN = typename torch::executor::
+                  promote_types<CTYPE_A, CTYPE_B>::type;
+              ET_DCHECK(CppTypeToScalarType<CTYPE_IN>::value == common_type);
               ET_SWITCH_REAL_TYPES(
-                  common_type, ctx, "remainder.Tensor_out", CTYPE_IN, [&]() {
-                    ET_SWITCH_REAL_TYPES(
-                        out_type,
-                        ctx,
-                        "remainder.Tensor_out",
-                        CTYPE_OUT,
-                        [&]() {
-                          apply_binary_elementwise_fn<
-                              CTYPE_A,
-                              CTYPE_B,
-                              CTYPE_OUT>(
-                              [](const CTYPE_A val_a, const CTYPE_B val_b) {
-                                CTYPE_IN a_casted =
-                                    static_cast<CTYPE_IN>(val_a);
-                                CTYPE_IN b_casted =
-                                    static_cast<CTYPE_IN>(val_b);
-                                CTYPE_IN value = utils::remainder_override(
-                                    a_casted, b_casted);
-
-                                return static_cast<CTYPE_OUT>(value);
-                              },
-                              a,
-                              b,
-                              out);
-                        });
+                  out_type, ctx, "remainder.Tensor_out", CTYPE_OUT, [&]() {
+                    RemainderInner<
+                        can_cast<CTYPE_IN, CTYPE_OUT>::value,
+                        CTYPE_A,
+                        CTYPE_B,
+                        CTYPE_IN,
+                        CTYPE_OUT>::run(a, b, out);
                   });
             });
       });
diff --git a/kernels/portable/cpu/pattern/bitwise_op.h b/kernels/portable/cpu/pattern/bitwise_op.h
new file mode 100644
index 00000000000..4a338bbf020
--- /dev/null
+++ b/kernels/portable/cpu/pattern/bitwise_op.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+
+#pragma once
+
+#include <executorch/kernels/portable/cpu/util/broadcast_util.h>
+#include <executorch/runtime/kernel/kernel_includes.h>
+
+namespace torch {
+namespace executor {
+namespace native {
+template <template <typename> typename OpFunc>
+Tensor& bitwise_op_out(
+    RuntimeContext& ctx,
+    const Tensor& a,
+    const Tensor& b,
+    Tensor& out,
+    const char* op_name) {
+  ET_KERNEL_CHECK(
+      ctx,
+      resize_to_broadcast_target_size(a, b, out) == Error::Ok,
+      InvalidArgument,
+      out);
+
+  ScalarType a_type = a.scalar_type();
+  ScalarType b_type = b.scalar_type();
+  ScalarType common_type = promoteTypes(a_type, b_type);
+  ScalarType out_type = out.scalar_type();
+
+  ET_KERNEL_CHECK(ctx, canCast(common_type, out_type), InvalidArgument, out);
+
+  ET_SWITCH_INT_TYPES_AND(Bool, a_type, ctx, op_name, CTYPE_A, [&]() {
+    ET_SWITCH_INT_TYPES_AND(Bool, b_type, ctx, op_name, CTYPE_B, [&]() {
+      ET_SWITCH_INT_TYPES_AND(Bool, common_type, ctx, op_name, CTYPE_IN, [&]() {
+        ET_SWITCH_REAL_TYPES_AND(
+            Bool, out_type, ctx, op_name, CTYPE_OUT, [&]() {
+              apply_binary_elementwise_fn<CTYPE_A, CTYPE_B, CTYPE_OUT>(
+                  [](const CTYPE_A val_a, const CTYPE_B val_b) {
+                    CTYPE_IN a_casted = static_cast<CTYPE_IN>(val_a);
+                    CTYPE_IN b_casted = static_cast<CTYPE_IN>(val_b);
+                    CTYPE_IN value = OpFunc<CTYPE_IN>()(a_casted, b_casted);
+
+                    return static_cast<CTYPE_OUT>(value);
+                  },
+                  a,
+                  b,
+                  out);
+            });
+      });
+    });
+  });
+
+  return out;
+}
+} // namespace native
+} // namespace executor
+} // namespace torch
diff --git a/kernels/portable/cpu/pattern/targets.bzl b/kernels/portable/cpu/pattern/targets.bzl
index 360d991767b..7e0b71ed950 100644
--- a/kernels/portable/cpu/pattern/targets.bzl
+++ b/kernels/portable/cpu/pattern/targets.bzl
@@ -6,6 +6,17 @@ def define_common_targets():
     The directory containing this targets.bzl file should also contain both
     TARGETS and BUCK files that call this function.
     """
+    runtime.cxx_library(
+        name = "bitwise_op",
+        exported_headers = [
+            "bitwise_op.h",
+        ],
+        compiler_flags = [],
+        deps = [
+            "//executorch/runtime/kernel:kernel_includes",
+        ],
+        visibility = ["//executorch/kernels/portable/cpu/...", "//executorch/kernels/optimized/cpu/..."],
+    )
 
     runtime.cxx_library(
         name = "pattern",
diff --git a/kernels/portable/cpu/targets.bzl b/kernels/portable/cpu/targets.bzl
index 77796c68526..7be1d94d2bf 100644
--- a/kernels/portable/cpu/targets.bzl
+++ b/kernels/portable/cpu/targets.bzl
@@ -142,6 +142,7 @@ _ATEN_OPS = (
         deps = [
             "//executorch/runtime/core/exec_aten/util:scalar_type_util",
             "//executorch/runtime/core/exec_aten/util:tensor_util",
+            "//executorch/kernels/portable/cpu/pattern:bitwise_op",
             "//executorch/kernels/portable/cpu/util:broadcast_util",
             "//executorch/kernels/portable/cpu/util:functional_util",
             ":scalar_utils",
@@ -160,6 +161,7 @@ _ATEN_OPS = (
         deps = [
             "//executorch/runtime/core/exec_aten/util:scalar_type_util",
             "//executorch/runtime/core/exec_aten/util:tensor_util",
+            "//executorch/kernels/portable/cpu/pattern:bitwise_op",
             "//executorch/kernels/portable/cpu/util:broadcast_util",
             "//executorch/kernels/portable/cpu/util:functional_util",
             ":scalar_utils",
@@ -170,6 +172,7 @@ _ATEN_OPS = (
         deps = [
             "//executorch/runtime/core/exec_aten/util:scalar_type_util",
             "//executorch/runtime/core/exec_aten/util:tensor_util",
+            "//executorch/kernels/portable/cpu/pattern:bitwise_op",
             "//executorch/kernels/portable/cpu/util:broadcast_util",
             "//executorch/kernels/portable/cpu/util:functional_util",
             ":scalar_utils",
@@ -560,6 +563,7 @@ _ATEN_OPS = (
         name = "op_maximum",
         deps = [
             "//executorch/kernels/portable/cpu/util:broadcast_util",
+            "//executorch/kernels/portable/cpu/util:math_util",
             ":scalar_utils",
         ],
     ),
@@ -591,6 +595,7 @@ _ATEN_OPS = (
         name = "op_minimum",
         deps = [
             "//executorch/kernels/portable/cpu/util:broadcast_util",
+            "//executorch/kernels/portable/cpu/util:math_util",
             ":scalar_utils",
         ],
     ),
diff --git a/kernels/portable/cpu/util/math_util.h b/kernels/portable/cpu/util/math_util.h
index 44cb47f8cba..df175147062 100644
--- a/kernels/portable/cpu/util/math_util.h
+++ b/kernels/portable/cpu/util/math_util.h
@@ -94,6 +94,48 @@ INT_T max_override(INT_T a, INT_T b) {
   return std::max(a, b);
 }
 
+template <
+    typename T,
+    typename std::enable_if<
+        std::is_same<T, torch::executor::Half>::value,
+        bool>::type = true>
+T min_override(T a, T b) {
+  const auto float_a = static_cast<float>(a);
+  if (std::isnan(float_a)) {
+    return a;
+  }
+  const auto float_b = static_cast<float>(b);
+  if (std::isnan(float_b)) {
+    return b;
+  }
+
+  if (float_a < float_b) {
+    return a;
+  }
+  return b;
+}
+
+template <
+    typename T,
+    typename std::enable_if<
+        std::is_same<T, torch::executor::Half>::value,
+        bool>::type = true>
+T max_override(T a, T b) {
+  const auto float_a = static_cast<float>(a);
+  if (std::isnan(float_a)) {
+    return a;
+  }
+  const auto float_b = static_cast<float>(b);
+  if (std::isnan(float_b)) {
+    return b;
+  }
+
+  if (float_a > float_b) {
+    return a;
+  }
+  return b;
+}
+
 /**
  * There is a slight difference in how std::fmod works compared to how ATen
  * determines remainders:
diff --git a/kernels/test/op_clamp_test.cpp b/kernels/test/op_clamp_test.cpp
index 871333482c8..0244fd55700 100644
--- a/kernels/test/op_clamp_test.cpp
+++ b/kernels/test/op_clamp_test.cpp
@@ -147,8 +147,16 @@ class OpClampOutTest : public OperatorTest {
   // Test cases that are compatible with float and double.
   template <ScalarType DTYPE>
   void run_floating_point_test_cases() {
-    constexpr auto kInfinity =
-        std::numeric_limits<typename TensorFactory<DTYPE>::ctype>::infinity();
+    using ctype = typename TensorFactory<DTYPE>::ctype;
+    using opt_infinity_type = std::conditional_t<
+        std::is_same<ctype, exec_aten::Half>::value,
+        float,
+        ctype>;
+    constexpr auto kInfinity = std::numeric_limits<ctype>::infinity();
+    const auto kOptInfinity =
+        OptScalar(static_cast<opt_infinity_type>(kInfinity));
+    const auto kOptMinusInfinity =
+        OptScalar(static_cast<opt_infinity_type>(-kInfinity));
     std::vector<ClampTestCase<DTYPE>> test_cases = {
         {
             std::string(__func__) + ": Simple negative/positive clamp",
@@ -178,7 +186,7 @@ class OpClampOutTest : public OperatorTest {
             std::string(__func__) + ": Infinite min",
             {2, 2}, // sizes
             {-10.1, -1.1, 1.1, 10.1}, // input_data
-            OptScalar(-kInfinity), // min
+            kOptMinusInfinity, // min
             OptScalar(5.5), // max
             {-10.1, -1.1, 1.1, 5.5}, // expected_data
         },
@@ -187,7 +195,7 @@ class OpClampOutTest : public OperatorTest {
             {2, 2}, // sizes
             {-10.1, -1.1, 1.1, 10.1}, // input_data
             OptScalar(-5.5), // min
-            OptScalar(kInfinity), // max
+            kOptInfinity, // max
             {-5.5, -1.1, 1.1, 10.1}, // expected_data
         },
         {
@@ -285,6 +293,15 @@ TEST_F(OpClampOutTest, LongTensors) {
   run_signed_integer_test_cases<ScalarType::Long>();
 }
 
+TEST_F(OpClampOutTest, HalfTensors) {
+  // Note that the integer test cases test the situation where the min/max value
+  // Scalars are integer types, demonstrating that floating point types can be
+  // clamped to integer values.
+  run_unsigned_integer_test_cases<ScalarType::Half>();
+  run_signed_integer_test_cases<ScalarType::Half>();
+  run_floating_point_test_cases<ScalarType::Half>();
+}
+
 TEST_F(OpClampOutTest, FloatTensors) {
   // Note that the integer test cases test the situation where the min/max value
   // Scalars are integer types, demonstrating that floating point types can be
diff --git a/kernels/test/op_fmod_test.cpp b/kernels/test/op_fmod_test.cpp
index 475d4ea5cb4..4ee4d84c1cc 100644
--- a/kernels/test/op_fmod_test.cpp
+++ b/kernels/test/op_fmod_test.cpp
@@ -32,3 +32,16 @@ class OpFmodTest : public OperatorTest {
     return torch::executor::aten::fmod_outf(context_, self, other, out);
   }
 };
+
+TEST_F(OpFmodTest, SmokeTest) {
+  TensorFactory<ScalarType::Long> tfDouble;
+  TensorFactory<ScalarType::Long> tfLong;
+  TensorFactory<ScalarType::Int> tfInt;
+
+  Tensor self = tfLong.full({2, 2}, 46);
+  Tensor other = tfInt.full({2, 2}, 4);
+  Tensor out = tfDouble.zeros({2, 2});
+  Tensor out_expected = tfDouble.full({2, 2}, 2.0);
+  op_fmod_tensor_out(self, other, out);
+  EXPECT_TENSOR_CLOSE(out, out_expected);
+}
diff --git a/kernels/test/op_minimum_test.cpp b/kernels/test/op_minimum_test.cpp
index be43e0af07d..7e12374b8d1 100644
--- a/kernels/test/op_minimum_test.cpp
+++ b/kernels/test/op_minimum_test.cpp
@@ -65,6 +65,10 @@ TEST_F(OpMinimumOutTest, LongTensors) {
   test_minimum_out_same_size<ScalarType::Long>();
 }
 
+TEST_F(OpMinimumOutTest, HalfTensors) {
+  test_minimum_out_same_size<ScalarType::Half>();
+}
+
 TEST_F(OpMinimumOutTest, FloatTensors) {
   test_minimum_out_same_size<ScalarType::Float>();
 }
diff --git a/kernels/test/op_remainder_test.cpp b/kernels/test/op_remainder_test.cpp
index 4a550958a1a..254e8122b61 100644
--- a/kernels/test/op_remainder_test.cpp
+++ b/kernels/test/op_remainder_test.cpp
@@ -21,6 +21,7 @@ using exec_aten::Tensor;
 using torch::executor::testing::TensorFactory;
 
 class OpRemainderOutTest : public OperatorTest {
+ protected:
   Tensor& op_remainder_tensor_out(
       const Tensor& self,
       const Tensor& other,
@@ -35,3 +36,16 @@ class OpRemainderOutTest : public OperatorTest {
     return torch::executor::aten::remainder_outf(context_, self, other, out);
   }
 };
+
+TEST_F(OpRemainderOutTest, SmokeTest) {
+  TensorFactory<ScalarType::Long> tfDouble;
+  TensorFactory<ScalarType::Long> tfLong;
+  TensorFactory<ScalarType::Int> tfInt;
+
+  Tensor self = tfLong.full({2, 2}, 46);
+  Tensor other = tfInt.full({2, 2}, 4);
+  Tensor out = tfDouble.zeros({2, 2});
+  Tensor out_expected = tfDouble.full({2, 2}, 2.0);
+  op_remainder_tensor_out(self, other, out);
+  EXPECT_TENSOR_CLOSE(out, out_expected);
+}
diff --git a/runtime/core/exec_aten/util/scalar_type_util.h b/runtime/core/exec_aten/util/scalar_type_util.h
index 595ed7a1c02..084289520aa 100644
--- a/runtime/core/exec_aten/util/scalar_type_util.h
+++ b/runtime/core/exec_aten/util/scalar_type_util.h
@@ -349,6 +349,12 @@ inline constexpr bool isIntegralType(
        t == exec_aten::ScalarType::Short);
 }
 
+template <typename T, bool includeBool>
+struct is_integral_type
+    : public std::integral_constant<
+          bool,
+          isIntegralType(CppTypeToScalarType<T>::value, includeBool)> {};
+
 inline constexpr bool isFloatingType(exec_aten::ScalarType t) {
   return (
       t == exec_aten::ScalarType::Double || t == exec_aten::ScalarType::Float ||