diff --git a/library/stdarch/crates/core_arch/src/simd_llvm.rs b/library/stdarch/crates/core_arch/src/simd_llvm.rs
index 5b5c6dd0a8ad..729953faeae5 100644
--- a/library/stdarch/crates/core_arch/src/simd_llvm.rs
+++ b/library/stdarch/crates/core_arch/src/simd_llvm.rs
@@ -61,5 +61,7 @@ extern "platform-intrinsic" {
     pub fn simd_fmax<T>(a: T, b: T) -> T;
 
     pub fn simd_fsqrt<T>(a: T) -> T;
+    pub fn simd_floor<T>(a: T) -> T;
+    pub fn simd_ceil<T>(a: T) -> T;
     pub fn simd_fma<T>(a: T, b: T, c: T) -> T;
 }
diff --git a/library/stdarch/crates/core_arch/src/x86/avx.rs b/library/stdarch/crates/core_arch/src/x86/avx.rs
index b4d8647d7358..76d648274310 100644
--- a/library/stdarch/crates/core_arch/src/x86/avx.rs
+++ b/library/stdarch/crates/core_arch/src/x86/avx.rs
@@ -426,7 +426,7 @@ pub unsafe fn _mm256_round_pd(a: __m256d, b: i32) -> __m256d {
 #[cfg_attr(test, assert_instr(vroundpd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_ceil_pd(a: __m256d) -> __m256d {
-    roundpd256(a, 0x02)
+    simd_ceil(a)
 }
 
 /// Rounds packed double-precision (64-bit) floating point elements in `a`
@@ -438,7 +438,7 @@ pub unsafe fn _mm256_ceil_pd(a: __m256d) -> __m256d {
 #[cfg_attr(test, assert_instr(vroundpd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_floor_pd(a: __m256d) -> __m256d {
-    roundpd256(a, 0x01)
+    simd_floor(a)
 }
 
 /// Rounds packed single-precision (32-bit) floating point elements in `a`
@@ -477,7 +477,7 @@ pub unsafe fn _mm256_round_ps(a: __m256, b: i32) -> __m256 {
 #[cfg_attr(test, assert_instr(vroundps))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_ceil_ps(a: __m256) -> __m256 {
-    roundps256(a, 0x02)
+    simd_ceil(a)
 }
 
 /// Rounds packed single-precision (32-bit) floating point elements in `a`
@@ -489,7 +489,7 @@ pub unsafe fn _mm256_ceil_ps(a: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vroundps))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_floor_ps(a: __m256) -> __m256 {
-    roundps256(a, 0x01)
+    simd_floor(a)
 }
 
 /// Returns the square root of packed single-precision (32-bit) floating point