Move the vsx code to powerpc (rust-lang#538)

Even if it is unlikely to happen the VSX instructions can be executed in
32bit mode just as well.

Author: lu-zero

coresimd/powerpc/vsx.rs

@@ -1,12 +1,120 @@
-//! PowerPC Vectir Scalar eXtensions (VSX) intrinsics.
+//! PowerPC Vector Scalar eXtensions (VSX) intrinsics.
 //!
 //! The references are: [POWER ISA v2.07B (for POWER8 & POWER8 with NVIDIA
 //! NVlink)] and [POWER ISA v3.0B (for POWER9)].
 //!
 //! [POWER ISA v2.07B (for POWER8 & POWER8 with NVIDIA NVlink)]: https://ibm.box.com/s/jd5w15gz301s5b5dt375mshpq9c3lh4u
 //! [POWER ISA v3.0B (for POWER9)]: https://ibm.box.com/s/1hzcwkwf8rbju5h9iyf44wm94amnlcrv
 
-//#[cfg(test)]
-//use stdsimd_test::assert_instr;
-//use coresimd::simd_llvm::simd_add;
-//use coresimd::simd::*;
+#![allow(non_camel_case_types)]
+
+use coresimd::simd_llvm::*;
+
+#[cfg(test)]
+use stdsimd_test::assert_instr;
+
+use mem;
+
+types! {
+    // pub struct vector_Float16 = f16x8;
+    /// PowerPC-specific 128-bit wide vector of two packed `i64`
+    pub struct vector_signed_long(i64, i64);
+    /// PowerPC-specific 128-bit wide vector of two packed `u64`
+    pub struct vector_unsigned_long(u64, u64);
+    /// PowerPC-specific 128-bit wide vector mask of two elements
+    pub struct vector_bool_long(i64, i64);
+    /// PowerPC-specific 128-bit wide vector of two packed `f64`
+    pub struct vector_double(f64, f64);
+    // pub struct vector_signed_long_long = vector_signed_long;
+    // pub struct vector_unsigned_long_long = vector_unsigned_long;
+    // pub struct vector_bool_long_long = vector_bool_long;
+    // pub struct vector_signed___int128 = i128x1;
+    // pub struct vector_unsigned___int128 = i128x1;
+}
+
+mod sealed {
+    use coresimd::simd::*;
+    use super::*;
+
+    pub trait VectorPermDI {
+        unsafe fn vec_xxpermdi(self, b: Self, dm: u8) -> Self;
+    }
+
+    // xxpermdi has an big-endian bias and extended mnemonics
+    #[inline]
+    #[target_feature(enable = "vsx")]
+    #[cfg_attr(
+        all(test, target_endian = "little"), assert_instr(xxmrgld, dm = 0x0)
+    )]
+    #[cfg_attr(
+        all(test, target_endian = "big"), assert_instr(xxspltd, dm = 0x0)
+    )]
+    unsafe fn xxpermdi(a: i64x2, b: i64x2, dm: u8) -> i64x2 {
+        match dm & 0b11 {
+            0 => simd_shuffle2(a, b, [0b00, 0b10]),
+            1 => simd_shuffle2(a, b, [0b01, 0b10]),
+            2 => simd_shuffle2(a, b, [0b00, 0b11]),
+            _ => simd_shuffle2(a, b, [0b01, 0b11]),
+        }
+    }
+
+    macro_rules! vec_xxpermdi {
+        {$impl: ident} => {
+            impl VectorPermDI for $impl {
+                #[inline]
+                #[target_feature(enable = "vsx")]
+                unsafe fn vec_xxpermdi(self, b: Self, dm: u8) -> Self {
+                    mem::transmute(xxpermdi(mem::transmute(self), mem::transmute(b), dm))
+                }
+            }
+        }
+    }
+
+    vec_xxpermdi! { vector_unsigned_long }
+    vec_xxpermdi! { vector_signed_long }
+    vec_xxpermdi! { vector_bool_long }
+    vec_xxpermdi! { vector_double }
+}
+
+/// Vector permute.
+#[inline]
+#[target_feature(enable = "vsx")]
+#[rustc_args_required_const(2)]
+pub unsafe fn vec_xxpermdi<T>(a: T, b: T, dm: u8) -> T
+where
+    T: sealed::VectorPermDI,
+{
+    a.vec_xxpermdi(b, dm)
+}
+
+#[cfg(test)]
+mod tests {
+    #[cfg(target_arch = "powerpc")]
+    use coresimd::arch::powerpc::*;
+
+    #[cfg(target_arch = "powerpc64")]
+    use coresimd::arch::powerpc64::*;
+
+    use coresimd::simd::*;
+    use stdsimd_test::simd_test;
+
+    macro_rules! test_vec_xxpermdi {
+        {$name:ident, $shorttype:ident, $longtype:ident, [$($a:expr),+], [$($b:expr),+], [$($c:expr),+], [$($d:expr),+]} => {
+            #[simd_test(enable = "vsx")]
+            unsafe fn $name() {
+                let a: $longtype = ::mem::transmute($shorttype::new($($a),+, $($b),+));
+                let b = ::mem::transmute($shorttype::new($($c),+, $($d),+));
+
+                assert_eq!($shorttype::new($($a),+, $($c),+), ::mem::transmute(vec_xxpermdi(a, b, 0)));
+                assert_eq!($shorttype::new($($b),+, $($c),+), ::mem::transmute(vec_xxpermdi(a, b, 1)));
+                assert_eq!($shorttype::new($($a),+, $($d),+), ::mem::transmute(vec_xxpermdi(a, b, 2)));
+                assert_eq!($shorttype::new($($b),+, $($d),+), ::mem::transmute(vec_xxpermdi(a, b, 3)));
+            }
+        }
+    }
+
+    test_vec_xxpermdi!{test_vec_xxpermdi_u64x2, u64x2, vector_unsigned_long, [0], [1], [2], [3]}
+    test_vec_xxpermdi!{test_vec_xxpermdi_i64x2, i64x2, vector_signed_long, [0], [-1], [2], [-3]}
+    test_vec_xxpermdi!{test_vec_xxpermdi_m64x2, m64x2, vector_bool_long, [false], [true], [false], [true]}
+    test_vec_xxpermdi!{test_vec_xxpermdi_f64x2, f64x2, vector_double, [0.0], [1.0], [2.0], [3.0]}
+}

coresimd/powerpc64/mod.rs

@@ -6,6 +6,3 @@
 //! [64-Bit ELF V2 ABI Specification - Power Architecture]: http://openpowerfoundation.org/wp-content/uploads/resources/leabi/leabi-20170510.pdf
 
 pub use coresimd::powerpc::*;
-
-mod vsx;
-pub use self::vsx::*;