Use llvm.ppc.altivec.lvx intrinsic for vec_ld

The `vec_ld` intrinsic is expected to automatically round the given
address down to a 16-byte boundary.

Author: Amanieu

crates/core_arch/src/powerpc/altivec.rs

@@ -51,6 +51,10 @@ types! {
 
 #[allow(improper_ctypes)]
 extern "C" {
+    #[link_name = "llvm.ppc.altivec.lvx"]
+    fn lvx(
+        p: *const i8,
+    ) -> vector_unsigned_int;
     #[link_name = "llvm.ppc.altivec.vperm"]
     fn vperm(
         a: vector_signed_int,
@@ -442,8 +446,7 @@ mod sealed {
     #[inline(always)]
     unsafe fn load(off: i32, p: *const i8) -> u32x4 {
         let addr = p.offset(off as isize);
-
-        *(addr as *const u32x4)
+        transmute(lvx(addr))
     }
 
     pub trait VectorLd {