Auto merge of #137412 - scottmcm:redo-swap, r=cuviper

Ensure `swap_nonoverlapping` is really always untyped

This replaces #134954, which was arguably overcomplicated.

## Fixes #134713

Actually using the type passed to `ptr::swap_nonoverlapping` for anything other than its size + align turns out to not work, so this goes back to always erasing the types down to just bytes.

(Except in `const`, which keeps doing the same thing as before to preserve `@RalfJung's` fix from #134689)

## Fixes #134946

I'd previously moved the swapping to use auto-vectorization *on bytes*, but someone pointed out on Discord that the tail loop handling from that left a whole bunch of byte-by-byte swapping around.  This goes back to manual tail handling to avoid that, then still triggers auto-vectorization on pointer-width values.  (So you'll see `<4 x i64>` on `x86-64-v3` for example.)

1 files changed, 4 insertions, 4 deletions

diff --git a/tests/codegen/simd/swap-simd-types.rs b/tests/codegen/simd/swap-simd-types.rs
index 69767d0a755..c063cc683a6 100644
--- a/tests/codegen/simd/swap-simd-types.rs
+++ b/tests/codegen/simd/swap-simd-types.rs
@@ -23,8 +23,8 @@ pub fn swap_single_m256(x: &mut __m256, y: &mut __m256) {
 #[no_mangle]
 pub fn swap_m256_slice(x: &mut [__m256], y: &mut [__m256]) {
     // CHECK-NOT: alloca
-    // CHECK: load <8 x float>{{.+}}align 32
-    // CHECK: store <8 x float>{{.+}}align 32
+    // CHECK-COUNT-2: load <4 x i64>{{.+}}align 32
+    // CHECK-COUNT-2: store <4 x i64>{{.+}}align 32
     if x.len() == y.len() {
         x.swap_with_slice(y);
     }
@@ -34,7 +34,7 @@ pub fn swap_m256_slice(x: &mut [__m256], y: &mut [__m256]) {
 #[no_mangle]
 pub fn swap_bytes32(x: &mut [u8; 32], y: &mut [u8; 32]) {
     // CHECK-NOT: alloca
-    // CHECK: load <32 x i8>{{.+}}align 1
-    // CHECK: store <32 x i8>{{.+}}align 1
+    // CHECK-COUNT-2: load <4 x i64>{{.+}}align 1
+    // CHECK-COUNT-2: store <4 x i64>{{.+}}align 1
     swap(x, y)
 }