Add vroundps, vceilps, vfloorps, vsqrtps, vsqrtpd (rust-lang#53)

* Add vroundps, vceilps, vfloorps, vsqrtps, vsqrtpd

* Uninhibit assert_instr on non-expanded intrinsics

Also use the new simd_test macro

* Use simd_test where possible

* Add automated tests for vround*

* Add target_feature guards to automated tests

* Move automated tests below their functions

Author: AdamNiederer

src/x86/avx.rs

@@ -110,6 +110,72 @@ pub unsafe fn _mm256_floor_pd(a: f64x4) -> f64x4 {
     roundpd256(a, 0x01)
 }
 
+/// Round packed single-precision (32-bit) floating point elements in `a`
+/// according to the flag `b`. The value of `b` may be as follows:
+/// 0x00: Round to the nearest whole number.
+/// 0x01: Round down, toward negative infinity.
+/// 0x02: Round up, toward positive infinity.
+/// 0x03: Truncate the values.
+/// For a few additional values options, check the LLVM docs:
+/// https://github.com/llvm-mirror/clang/blob/dcd8d797b20291f1a6b3e0ddda085aa2bbb382a8/lib/Headers/avxintrin.h#L382
+#[inline(always)]
+#[target_feature = "+avx"]
+// #[cfg_attr(test, assert_instr(vroundps))]
+// TODO: Replace with assert_expanded_instr https://github.com/rust-lang-nursery/stdsimd/issues/49
+pub fn _mm256_round_ps(a: f32x8, b: i32) -> f32x8 {
+    macro_rules! call {
+        ($imm8:expr) => {
+            unsafe { roundps256(a, $imm8) }
+        }
+    }
+    constify_imm8!(b, call)
+}
+
+// TODO: Remove once a macro is ipmlemented to automate these tests
+// https://github.com/rust-lang-nursery/stdsimd/issues/49
+#[cfg(test)]
+#[target_feature = "+avx"]
+#[cfg_attr(test, assert_instr(vroundps))]
+fn test_mm256_round_ps(a: f32x8) -> f32x8 {
+    _mm256_round_ps(a, 0x00)
+}
+
+/// Round packed single-precision (32-bit) floating point elements in `a` toward
+/// positive infinity.
+#[inline(always)]
+#[target_feature = "+avx"]
+#[cfg_attr(test, assert_instr(vroundps))]
+pub fn _mm256_ceil_ps(a: f32x8) -> f32x8 {
+    unsafe { roundps256(a, 0x02) }
+}
+
+/// Round packed single-precision (32-bit) floating point elements in `a` toward
+/// negative infinity.
+#[inline(always)]
+#[target_feature = "+avx"]
+#[cfg_attr(test, assert_instr(vroundps))]
+pub fn _mm256_floor_ps(a: f32x8) -> f32x8 {
+    unsafe { roundps256(a, 0x01) }
+}
+
+/// Return the square root of packed single-precision (32-bit) floating point
+/// elements in `a`.
+#[inline(always)]
+#[target_feature = "+avx"]
+#[cfg_attr(test, assert_instr(vsqrtps))]
+pub fn _mm256_sqrt_ps(a: f32x8) -> f32x8 {
+    unsafe { sqrtps256(a) }
+}
+
+/// Return the square root of packed double-precision (64-bit) floating point
+/// elements in `a`.
+#[inline(always)]
+#[target_feature = "+avx"]
+#[cfg_attr(test, assert_instr(vsqrtpd))]
+pub fn _mm256_sqrt_pd(a: f64x4) -> f64x4 {
+    unsafe { sqrtpd256(a) }
+}
+
 /// LLVM intrinsics used in the above functions
 #[allow(improper_ctypes)]
 extern "C" {
@@ -119,9 +185,15 @@ extern "C" {
     fn addsubps256(a: f32x8, b: f32x8) -> f32x8;
     #[link_name = "llvm.x86.avx.round.pd.256"]
     fn roundpd256(a: f64x4, b: i32) -> f64x4;
+    #[link_name = "llvm.x86.avx.round.ps.256"]
+    fn roundps256(a: f32x8, b: i32) -> f32x8;
+    #[link_name = "llvm.x86.avx.sqrt.pd.256"]
+    fn sqrtpd256(a: f64x4) -> f64x4;
+    #[link_name = "llvm.x86.avx.sqrt.ps.256"]
+    fn sqrtps256(a: f32x8) -> f32x8;
 }
 
-#[cfg(test)]
+#[cfg(all(test, target_feature = "avx", any(target_arch = "x86", target_arch = "x86_64")))]
 mod tests {
     use stdsimd_test::simd_test;
 
@@ -229,4 +301,51 @@ mod tests {
         let expected_up = f64x4::new(2.0, 3.0, 4.0, -1.0);
         assert_eq!(result_up, expected_up);
     }
+
+    #[simd_test = "avx"]
+    fn _mm256_round_ps() {
+        let a = f32x8::new(1.55, 2.2, 3.99, -1.2, 1.55, 2.2, 3.99, -1.2);
+        let result_closest = avx::_mm256_round_ps(a, 0b00000000);
+        let result_down = avx::_mm256_round_ps(a, 0b00000001);
+        let result_up = avx::_mm256_round_ps(a, 0b00000010);
+        let expected_closest = f32x8::new(2.0, 2.0, 4.0, -1.0, 2.0, 2.0, 4.0, -1.0);
+        let expected_down = f32x8::new(1.0, 2.0, 3.0, -2.0, 1.0, 2.0, 3.0, -2.0);
+        let expected_up = f32x8::new(2.0, 3.0, 4.0, -1.0, 2.0, 3.0, 4.0, -1.0);
+        assert_eq!(result_closest, expected_closest);
+        assert_eq!(result_down, expected_down);
+        assert_eq!(result_up, expected_up);
+    }
+
+    #[simd_test = "avx"]
+    fn _mm256_floor_ps() {
+        let a = f32x8::new(1.55, 2.2, 3.99, -1.2, 1.55, 2.2, 3.99, -1.2);
+        let result_down = avx::_mm256_floor_ps(a);
+        let expected_down = f32x8::new(1.0, 2.0, 3.0, -2.0, 1.0, 2.0, 3.0, -2.0);
+        assert_eq!(result_down, expected_down);
+    }
+
+    #[simd_test = "avx"]
+    fn _mm256_ceil_ps() {
+        let a = f32x8::new(1.55, 2.2, 3.99, -1.2, 1.55, 2.2, 3.99, -1.2);
+        let result_up = avx::_mm256_ceil_ps(a);
+        let expected_up = f32x8::new(2.0, 3.0, 4.0, -1.0, 2.0, 3.0, 4.0, -1.0);
+        assert_eq!(result_up, expected_up);
+    }
+
+    #[simd_test = "avx"]
+    fn _mm256_sqrt_pd() {
+        let a = f64x4::new(4.0, 9.0, 16.0, 25.0);
+        let r = avx::_mm256_sqrt_pd(a, );
+        let e = f64x4::new(2.0, 3.0, 4.0, 5.0);
+        assert_eq!(r, e);
+    }
+
+    #[simd_test = "avx"]
+    fn _mm256_sqrt_ps() {
+        let a = f32x8::new(4.0, 9.0, 16.0, 25.0, 4.0, 9.0, 16.0, 25.0);
+        let r = avx::_mm256_sqrt_ps(a);
+        let e = f32x8::new(2.0, 3.0, 4.0, 5.0, 2.0, 3.0, 4.0, 5.0);
+        assert_eq!(r, e);
+    }
+
 }