powerpc/vsx: implement vec_sld for vector_double

Author: lei137

crates/core_arch/src/powerpc/altivec.rs

@@ -411,7 +411,7 @@ unsafe extern "unadjusted" {
 }
 
 #[macro_use]
-mod sealed {
+pub(crate) mod sealed {
     use super::*;
 
     #[unstable(feature = "stdarch_powerpc", issue = "111145")]
@@ -3030,7 +3030,7 @@ mod sealed {
     #[inline]
     #[target_feature(enable = "altivec")]
     #[cfg_attr(test, assert_instr(vsldoi, UIMM4 = 1))]
-    unsafe fn vsldoi<const UIMM4: i32>(
+    pub(crate) unsafe fn vsldoi<const UIMM4: i32>(
         a: vector_unsigned_char,
         b: vector_unsigned_char,
     ) -> vector_unsigned_char {
@@ -3082,9 +3082,9 @@ mod sealed {
 
     // TODO: collapse the two once generic_const_exprs are usable.
     #[inline]
-    #[target_feature(enable = "altivec")]
+    #[target_feature(enable = "vsx")]
     #[cfg_attr(test, assert_instr(xxsldwi, UIMM2 = 1))]
-    unsafe fn xxsldwi<const UIMM2: i32>(
+    pub(crate) unsafe fn xxsldwi<const UIMM2: i32>(
         a: vector_unsigned_char,
         b: vector_unsigned_char,
     ) -> vector_unsigned_char {
@@ -3134,24 +3134,6 @@ mod sealed {
         }
     }
 
-    macro_rules! impl_vec_sld {
-        ($($ty:ident),+) => { $(
-            #[unstable(feature = "stdarch_powerpc", issue = "111145")]
-            impl VectorSld for $ty {
-                #[inline]
-                #[target_feature(enable = "altivec")]
-                unsafe fn vec_sld<const UIMM4: i32>(self, b: Self) -> Self {
-                    transmute(vsldoi::<UIMM4>(transmute(self), transmute(b)))
-                }
-                #[inline]
-                #[target_feature(enable = "altivec")]
-                unsafe fn vec_sldw<const UIMM2: i32>(self, b: Self) -> Self {
-                    transmute(xxsldwi::<UIMM2>(transmute(self), transmute(b)))
-                }
-           }
-        )+ };
-    }
-
     impl_vec_sld! { vector_bool_char, vector_signed_char, vector_unsigned_char }
     impl_vec_sld! { vector_bool_short, vector_signed_short, vector_unsigned_short }
     impl_vec_sld! { vector_bool_int, vector_signed_int, vector_unsigned_int }

crates/core_arch/src/powerpc/macros.rs

@@ -274,6 +274,25 @@ macro_rules! t_b {
     };
 }
 
+macro_rules! impl_vec_sld {
+    ($($ty:ident),+) => { $(
+        #[unstable(feature = "stdarch_powerpc", issue = "111145")]
+        impl VectorSld for $ty {
+            #[inline]
+            #[target_feature(enable = "altivec")]
+            unsafe fn vec_sld<const UIMM4: i32>(self, b: Self) -> Self {
+                transmute(vsldoi::<UIMM4>(transmute(self), transmute(b)))
+            }
+            #[inline]
+            #[target_feature(enable = "vsx")]
+            unsafe fn vec_sldw<const UIMM2: i32>(self, b: Self) -> Self {
+                transmute(xxsldwi::<UIMM2>(transmute(self), transmute(b)))
+            }
+       }
+    )+ };
+}
+
+pub(crate) use impl_vec_sld;
 pub(crate) use impl_vec_trait;
 pub(crate) use s_t_l;
 pub(crate) use t_b;

crates/core_arch/src/powerpc/vsx.rs

@@ -10,6 +10,7 @@
 
 use crate::core_arch::powerpc::*;
 use crate::core_arch::simd::*;
+use super::macros::*;
 
 #[cfg(test)]
 use stdarch_test::assert_instr;
@@ -173,6 +174,10 @@ mod sealed {
     vec_mergeeo! { vector_float, mergee, mergeo }
 }
 
+// Implement AltiVec's VectorSld trait for vector_double to enable vec_sld support
+use crate::core_arch::powerpc::altivec::sealed::{VectorSld, vsldoi, xxsldwi};
+impl_vec_sld! { vector_double }
+
 /// Vector permute.
 #[inline]
 #[target_feature(enable = "vsx")]
@@ -255,4 +260,24 @@ mod tests {
     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]}
+
+    #[simd_test(enable = "altivec")]
+    fn test_vec_sld_f64x2() {
+        use crate::core_arch::powerpc::altivec::sealed::VectorSld;
+
+        let a = vector_double::from(f64x2::from_array([1.0, 2.0]));
+        let b = vector_double::from(f64x2::from_array([3.0, 4.0]));
+
+        // Shift left by 8 bytes (1 f64 element)
+        // On little-endian: shifts right in memory, result is [b[1], a[0]] = [4.0, 1.0]
+        // On big-endian: shifts left in memory, result is [a[1], b[0]] = [2.0, 3.0]
+        unsafe {
+            let result: f64x2 = transmute(a.vec_sld::<8>(b));
+            #[cfg(target_endian = "little")]
+            let expected = f64x2::from_array([4.0, 1.0]);
+            #[cfg(target_endian = "big")]
+            let expected = f64x2::from_array([2.0, 3.0]);
+            assert_eq!(result, expected);
+        }
+    }
 }