diff --git a/neon2rvv.h b/neon2rvv.h
index 272f9fd8..8cd0f9ec 100644
--- a/neon2rvv.h
+++ b/neon2rvv.h
@@ -10587,13 +10587,25 @@ FORCE_INLINE uint64x2_t vmull_n_u32(uint32x2_t a, uint32_t b) {
   return __riscv_vlmul_trunc_v_u64m2_u64m1(__riscv_vwmulu_vx_u64m2(a, b, 4));
 }
 
-// FORCE_INLINE int32x4_t vmull_high_n_s16(int16x8_t a, int16_t b);
+FORCE_INLINE int32x4_t vmull_high_n_s16(int16x8_t a, int16_t b) {
+  vint16m1_t a_high = __riscv_vslidedown_vx_i16m1(a, 4, 8);
+  return __riscv_vlmul_trunc_v_i32m2_i32m1(__riscv_vwmul_vx_i32m2(a_high, b, 4));
+}
 
-// FORCE_INLINE int64x2_t vmull_high_n_s32(int32x4_t a, int32_t b);
+FORCE_INLINE int64x2_t vmull_high_n_s32(int32x4_t a, int32_t b) {
+  vint32m1_t a_high = __riscv_vslidedown_vx_i32m1(a, 2, 4);
+  return __riscv_vlmul_trunc_v_i64m2_i64m1(__riscv_vwmul_vx_i64m2(a_high, b, 2));
+}
 
-// FORCE_INLINE uint32x4_t vmull_high_n_u16(uint16x8_t a, uint16_t b);
+FORCE_INLINE uint32x4_t vmull_high_n_u16(uint16x8_t a, uint16_t b) {
+  vuint16m1_t a_high = __riscv_vslidedown_vx_u16m1(a, 4, 8);
+  return __riscv_vlmul_trunc_v_u32m2_u32m1(__riscv_vwmulu_vx_u32m2(a_high, b, 4));
+}
 
-// FORCE_INLINE uint64x2_t vmull_high_n_u32(uint32x4_t a, uint32_t b);
+FORCE_INLINE uint64x2_t vmull_high_n_u32(uint32x4_t a, uint32_t b) {
+  vuint32m1_t a_high = __riscv_vslidedown_vx_u32m1(a, 2, 4);
+  return __riscv_vlmul_trunc_v_u64m2_u64m1(__riscv_vwmulu_vx_u64m2(a_high, b, 4));
+}
 
 FORCE_INLINE int32x4_t vqdmull_n_s16(int16x4_t a, int16_t b) {
   vint32m2_t ab_mul = __riscv_vwmul_vx_i32m2(a, b, 4);
diff --git a/tests/impl.cpp b/tests/impl.cpp
index e7390d8d..bad19c1b 100644
--- a/tests/impl.cpp
+++ b/tests/impl.cpp
@@ -33858,13 +33858,73 @@ result_t test_vmull_n_u32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #endif  // ENABLE_TEST_ALL
 }
 
-result_t test_vmull_high_n_s16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+result_t test_vmull_high_n_s16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const int16_t *_a = (int16_t *)impl.test_cases_int_pointer1;
+  const int16_t *_b = (int16_t *)impl.test_cases_int_pointer2;
+  int32_t _c[4];
+  for (int i = 0; i < 4; i++) {
+    _c[i] = (int32_t)_a[i + 4] * (int32_t)_b[0];
+  }
+
+  int16x8_t a = vld1q_s16(_a);
+  int32x4_t c = vmull_high_n_s16(a, _b[0]);
+  return validate_int32(c, _c[0], _c[1], _c[2], _c[3]);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
+
+result_t test_vmull_high_n_s32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const int32_t *_a = (int32_t *)impl.test_cases_int_pointer1;
+  const int32_t *_b = (int32_t *)impl.test_cases_int_pointer2;
+  int64_t _c[2];
+  for (int i = 0; i < 2; i++) {
+    _c[i] = (int64_t)_a[i + 2] * (int64_t)_b[0];
+  }
+
+  int32x4_t a = vld1q_s32(_a);
+  int64x2_t c = vmull_high_n_s32(a, _b[0]);
+  return validate_int64(c, _c[0], _c[1]);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
 
-result_t test_vmull_high_n_s32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+result_t test_vmull_high_n_u16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const uint16_t *_a = (uint16_t *)impl.test_cases_int_pointer1;
+  const uint16_t *_b = (uint16_t *)impl.test_cases_int_pointer2;
+  uint32_t _c[4];
+  for (int i = 0; i < 4; i++) {
+    _c[i] = (uint32_t)_a[i + 4] * (uint32_t)_b[0];
+  }
+
+  uint16x8_t a = vld1q_u16(_a);
+  uint32x4_t c = vmull_high_n_u16(a, _b[0]);
+  return validate_uint32(c, _c[0], _c[1], _c[2], _c[3]);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
 
-result_t test_vmull_high_n_u16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+result_t test_vmull_high_n_u32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const uint32_t *_a = (uint32_t *)impl.test_cases_int_pointer1;
+  const uint32_t *_b = (uint32_t *)impl.test_cases_int_pointer2;
+  uint64_t _c[2];
+  for (int i = 0; i < 2; i++) {
+    _c[i] = (uint64_t)_a[i + 2] * (uint64_t)_b[0];
+  }
 
-result_t test_vmull_high_n_u32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+  uint32x4_t a = vld1q_u32(_a);
+  uint64x2_t c = vmull_high_n_u32(a, _b[0]);
+  return validate_uint64(c, _c[0], _c[1]);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
 
 result_t test_vqdmull_n_s16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #ifdef ENABLE_TEST_ALL
diff --git a/tests/impl.h b/tests/impl.h
index 071b28a7..4d0cad98 100644
--- a/tests/impl.h
+++ b/tests/impl.h
@@ -2191,10 +2191,10 @@
   _(vmull_n_s32)                                                                 \
   _(vmull_n_u16)                                                                 \
   _(vmull_n_u32)                                                                 \
-  /*_(vmull_high_n_s16)                                                       */ \
-  /*_(vmull_high_n_s32)                                                       */ \
-  /*_(vmull_high_n_u16)                                                       */ \
-  /*_(vmull_high_n_u32)                                                       */ \
+  _(vmull_high_n_s16)                                                            \
+  _(vmull_high_n_s32)                                                            \
+  _(vmull_high_n_u16)                                                            \
+  _(vmull_high_n_u32)                                                            \
   _(vqdmull_n_s16)                                                               \
   _(vqdmull_n_s32)                                                               \
   /*_(vqdmull_high_n_s16)                                                     */ \