added support for _mm_cvtpd_ps / cvtpd2ps

Author: krampenschiesser

src/x86/sse2.rs

@@ -1718,6 +1718,14 @@ pub unsafe fn _mm_ucomineq_sd(a: f64x2, b: f64x2) -> bool {
     mem::transmute(ucomineqsd(a, b) as u8)
 }
 
+/// Convert packed double-precision (64-bit) floating-point elements in "a" to packed single-precision (32-bit) floating-point elements
+#[inline(always)]
+#[target_feature = "+sse2"]
+#[cfg_attr(test, assert_instr(cvtpd2ps))]
+pub unsafe fn _mm_cvtpd_ps(a: f64x2) -> f32x4 {
+    cvtpd2ps(a)
+}
+
 /// Return a mask of the most significant bit of each element in `a`.
 ///
 /// The mask is stored in the 2 least significant bits of the return value.
@@ -1882,6 +1890,8 @@ extern {
     fn ucomineqsd(a: f64x2, b: f64x2) -> i32;
     #[link_name = "llvm.x86.sse2.movmsk.pd"]
     fn movmskpd(a: f64x2) -> i32;
+    #[link_name = "llvm.x86.sse2.cvtpd2ps"]
+    fn cvtpd2ps(a: f64x2) -> f32x4;
 }
 
 #[cfg(test)]
@@ -3406,4 +3416,21 @@ mod tests {
         let r = sse2::_mm_movemask_pd(f64x2::new(-1.0, -5.0));
         assert_eq!(r, 0b11);
     }
+
+    #[simd_test = "sse2"]
+    unsafe fn _mm_cvtpd_ps() {
+        use std::{f64,f32};
+
+        let r = sse2::_mm_cvtpd_ps(f64x2::new(-1.0, 5.0));
+        assert_eq!(r, f32x4::new(-1.0, 5.0, 0.0, 0.0));
+
+        let r = sse2::_mm_cvtpd_ps(f64x2::new(-1.0, -5.0));
+        assert_eq!(r, f32x4::new(-1.0, -5.0, 0.0, 0.0));
+
+        let r = sse2::_mm_cvtpd_ps(f64x2::new(f64::MAX, f64::MIN));
+        assert_eq!(r, f32x4::new(f32::INFINITY, f32::NEG_INFINITY, 0.0,0.0));
+
+        let r = sse2::_mm_cvtpd_ps(f64x2::new(f32::MAX as f64, f32::MIN as f64));
+        assert_eq!(r, f32x4::new(f32::MAX, f32::MIN, 0.0,0.0));
+    }
 }