Start on AVX (#3)

start on avx

Author: jackmott

TODO.md

@@ -540,9 +540,9 @@ sse4.2
 
 avx
 ---
-* [ ] `_mm256_add_pd`
-* [ ] `_mm256_add_ps`
-* [ ] `_mm256_addsub_pd`
+* [x] `_mm256_add_pd`
+* [x] `_mm256_add_ps`
+* [x] `_mm256_addsub_pd`
 * [ ] `_mm256_addsub_ps`
 * [ ] `_mm256_and_pd`
 * [ ] `_mm256_and_ps`

src/x86/avx.rs

@@ -0,0 +1,71 @@
+use v256::*;
+
+/// Add packed double-precision (64-bit) floating-point elements
+/// in `a` and `b`.
+#[inline(always)]
+#[target_feature = "+avx"]
+pub fn _mm256_add_pd(a: f64x4, b: f64x4) -> f64x4 {
+    a + b
+}
+
+/// Add packed single-precision (32-bit) floating-point elements in `a` and `b`.
+#[inline(always)]
+#[target_feature = "+avx"]
+pub fn _mm256_add_ps(a: f32x8, b: f32x8) -> f32x8 {
+    a + b
+}
+
+/// Alternatively add and subtract packed double-precision (64-bit)
+/// floating-point elements in `a` to/from packed elements in `b`.
+#[inline(always)]
+#[target_feature = "+avx"]
+pub fn _mm256_addsub_pd(a: f64x4, b: f64x4) -> f64x4 {
+    unsafe { addsubpd256(a, b) }
+}
+
+
+#[allow(improper_ctypes)]
+extern "C" {
+    #[link_name = "llvm.x86.avx.addsub.pd.256"]
+    fn addsubpd256(a: f64x4, b:f64x4) -> f64x4;
+}
+
+
+#[cfg(test)]
+mod tests {
+    use v256::*;
+    use x86::avx;
+
+    #[test]
+    #[target_feature = "+avx"]
+    fn _mm256_add_pd() {
+        let a = f64x4::new(1.0, 2.0, 3.0, 4.0);
+        let b = f64x4::new(5.0, 6.0, 7.0, 8.0);
+        let r = avx::_mm256_add_pd(a, b);
+        let e = f64x4::new(6.0, 8.0, 10.0, 12.0);
+        assert_eq!(r, e);
+    }
+
+    #[test]
+    #[target_feature = "+avx"]
+    fn _mm256_add_ps() {
+        let a = f32x8::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
+        let b = f32x8::new(9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0);
+        let r = avx::_mm256_add_ps(a, b);
+        let e = f32x8::new(10.0, 12.0, 14.0, 16.0, 18.0, 20.0, 22.0, 24.0);
+        assert_eq!(r, e);
+    }
+
+    #[test]
+    #[target_feature = "+avx"]
+    fn _mm256_addsub_pd() {
+        let a = f64x4::new(1.0, 2.0, 3.0, 4.0);
+        let b = f64x4::new(5.0, 6.0, 7.0, 8.0);
+        let r = avx::_mm256_addsub_pd(a, b);
+        let e = f64x4::new(-4.0,8.0,-4.0,12.0);
+        assert_eq!(r, e);
+    }
+
+
+
+}

src/x86/avx2.rs

@@ -568,7 +568,7 @@ mod tests {
             0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
             16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31);
         let b = i8x32::new(
-            31, 30, 2, 28, 27, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, 
+            31, 30, 2, 28, 27, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 16,
             15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
         let r = avx2::_mm256_cmpeq_epi8(a, b);
         assert_eq!(r, i8x32::splat(0).replace(2,0xFFu8 as i8));
@@ -641,7 +641,7 @@ mod tests {
         let b = i16x16::splat(4);
         let r = avx2::_mm256_hadd_epi16(a, b);
         let e = i16x16::new(4, 4, 4, 4, 8, 8, 8, 8, 4, 4, 4, 4, 8, 8, 8, 8);
-        assert_eq!(r,e);
+        assert_eq!(r, e);
     }
 
     #[test]
@@ -651,7 +651,7 @@ mod tests {
         let b = i32x8::splat(4);
         let r = avx2::_mm256_hadd_epi32(a, b);
         let e = i32x8::new(4, 4, 8, 8, 4, 4, 8, 8);
-        assert_eq!(r,e);
+        assert_eq!(r, e);
     }
 
     #[test]
@@ -662,7 +662,7 @@ mod tests {
         let r = avx2::_mm256_hadds_epi16(a, b);
         let e = i16x16::new(
             0x7FFF, 4, 4, 4, 8, 8, 8, 8, 4, 4, 4, 4, 8, 8, 8, 8);
-        assert_eq!(r,e);
+        assert_eq!(r, e);
     }
 
     #[test]
@@ -672,7 +672,7 @@ mod tests {
         let b = i16x16::splat(4);
         let r = avx2::_mm256_hsub_epi16(a, b);
         let e = i16x16::splat(0);
-        assert_eq!(r,e);
+        assert_eq!(r, e);
     }
 
     #[test]
@@ -682,7 +682,7 @@ mod tests {
         let b = i32x8::splat(4);
         let r = avx2::_mm256_hsub_epi32(a, b);
         let e = i32x8::splat(0);
-        assert_eq!(r,e);
+        assert_eq!(r, e);
     }
 
     #[test]
@@ -692,8 +692,7 @@ mod tests {
         let b = i16x16::splat(4);
         let r = avx2::_mm256_hsubs_epi16(a, b);
         let e = i16x16::splat(0).replace(0,0x7FFF);
-        assert_eq!(r,e);
+        assert_eq!(r, e);
     }
 
-
 }

src/x86/mod.rs

@@ -2,6 +2,7 @@ pub use self::sse::*;
 pub use self::sse2::*;
 pub use self::ssse3::*;
 pub use self::sse42::*;
+pub use self::avx::*;
 pub use self::avx2::*;
 
 #[allow(non_camel_case_types)]
@@ -13,4 +14,5 @@ mod sse;
 mod sse2;
 mod ssse3;
 mod sse42;
+mod avx;
 mod avx2;