feat: add u8x16::unpack_low/high (#147)

This commit adds u8x16::unpack_low and u8x16::unpack_high which is basically SSE2's unpack_low_i8_m128i and unpack_high_i8_m128i, but also allows SIMD128 and NEON backends.

src/u8x16_.rs

@@ -459,6 +459,62 @@ impl u8x16 {
     }
   }
 
+  /// Unpack and interleave low lanes of two u8x16
+  pub fn unpack_low(lhs: u8x16, rhs: u8x16) -> u8x16 {
+    pick! {
+        if #[cfg(target_feature = "sse2")] {
+            u8x16 { sse: unpack_low_i8_m128i(lhs.sse, rhs.sse) }
+        } else if #[cfg(target_feature = "simd128")] {
+          u8x16 { simd: u8x16_shuffle::<0, 16, 1, 17, 2, 18, 3, 19, 4, 20, 5, 21, 6, 22, 7, 23>(lhs.simd, rhs.simd) }
+        } else if #[cfg(all(target_feature = "neon", target_arch = "aarch64"))] {
+            let lhs = unsafe { vget_low_u8(lhs.neon) };
+            let rhs = unsafe { vget_low_u8(rhs.neon) };
+
+            let zipped = unsafe { vzip_u8(lhs, rhs) };
+            u8x16 { neon: unsafe { vcombine_u8(zipped.0, zipped.1) } }
+        } else {
+            u8x16::new([
+                lhs.as_array_ref()[0], rhs.as_array_ref()[0],
+                lhs.as_array_ref()[1], rhs.as_array_ref()[1],
+                lhs.as_array_ref()[2], rhs.as_array_ref()[2],
+                lhs.as_array_ref()[3], rhs.as_array_ref()[3],
+                lhs.as_array_ref()[4], rhs.as_array_ref()[4],
+                lhs.as_array_ref()[5], rhs.as_array_ref()[5],
+                lhs.as_array_ref()[6], rhs.as_array_ref()[6],
+                lhs.as_array_ref()[7], rhs.as_array_ref()[7],
+            ])
+        }
+    }
+  }
+
+  /// Unpack and interleave high lanes of two u8x16
+  pub fn unpack_high(lhs: u8x16, rhs: u8x16) -> u8x16 {
+    pick! {
+        if #[cfg(target_feature = "sse2")] {
+            u8x16 { sse: unpack_high_i8_m128i(lhs.sse, rhs.sse) }
+        } else if #[cfg(target_feature = "simd128")] {
+            u8x16 { simd: u8x16_shuffle::<8, 24, 9, 25, 10, 26, 11, 27, 12, 28, 13, 29, 14, 30, 15, 31>(lhs.simd, rhs.simd) }
+        } else if #[cfg(all(target_feature = "neon", target_arch = "aarch64"))] {
+            let lhs = unsafe { vget_high_u8(lhs.neon) };
+            let rhs = unsafe { vget_high_u8(rhs.neon) };
+
+            let zipped = unsafe { vzip_u8(lhs, rhs) };
+            u8x16 { neon: unsafe { vcombine_u8(zipped.0, zipped.1) } }
+        } else {
+            u8x16::new([
+                lhs.as_array_ref()[8], rhs.as_array_ref()[8],
+                lhs.as_array_ref()[9], rhs.as_array_ref()[9],
+                lhs.as_array_ref()[10], rhs.as_array_ref()[10],
+                lhs.as_array_ref()[11], rhs.as_array_ref()[11],
+                lhs.as_array_ref()[12], rhs.as_array_ref()[12],
+                lhs.as_array_ref()[13], rhs.as_array_ref()[13],
+                lhs.as_array_ref()[14], rhs.as_array_ref()[14],
+                lhs.as_array_ref()[15], rhs.as_array_ref()[15],
+            ])
+        }
+    }
+  }
+
   #[inline]
   pub fn to_array(self) -> [u8; 16] {
     cast(self)

tests/all_tests/t_u8x16.rs

@@ -163,3 +163,19 @@ fn impl_u8x16_min() {
   let actual = a.min(b);
   assert_eq!(expected, actual);
 }
+
+#[test]
+fn impl_unpack_low_u8() {
+  let a = u8x16::from([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]);
+  let b = u8x16::from([12, 11, 22, 13, 99, 15, 16, 17, 8, 19, 2, 21, 22, 3, 24, 127]);
+  let c: [u8; 16] = u8x16::unpack_low(a, b).into();
+  assert_eq!(c, [0, 12, 1, 11, 2, 22, 3, 13, 4, 99, 5, 15, 6, 16, 7, 17]);
+}
+
+#[test]
+fn impl_unpack_high_u8() {
+  let a = u8x16::from([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]);
+  let b = u8x16::from([12, 11, 22, 13, 99, 15, 16, 17, 8, 19, 2, 21, 22, 3, 24, 127]);
+  let c: [u8; 16] = u8x16::unpack_high(a, b).into();
+  assert_eq!(c, [8, 8, 9, 19, 10, 2, 11, 21, 12, 22, 13, 3, 14, 24, 15, 127]);
+}