Auto merge of #116422 - the8472:chunked-generic-slice-eq, r=<try>

Chunked generic slice eq

looks nice in a microbenchmark, let's see if perf agrees

```
OLD:
    slice::slice_cmp_generic 54.00ns/iter +/- 1.00ns
NEW:
    slice::slice_cmp_generic 20.00ns/iter +/- 2.00ns
```

Author: bors

library/core/benches/slice.rs

@@ -171,3 +171,17 @@ fn fold_to_last(b: &mut Bencher) {
     let slice: &[i32] = &[0; 1024];
     b.iter(|| black_box(slice).iter().fold(None, |_, r| Some(NonNull::from(r))));
 }
+
+#[bench]
+fn slice_cmp_generic(b: &mut Bencher) {
+    #[derive(PartialEq, Clone, Copy)]
+    struct Foo(u32, u32);
+
+    let left = [Foo(128, 128); 128];
+    let right = [Foo(128, 128); 128];
+
+    b.iter(|| {
+        let (left, right) = (black_box(&left), black_box(&right));
+        left.as_slice() == right.as_slice()
+    });
+}

library/core/src/slice/cmp.rs

@@ -60,7 +60,22 @@ where
             return false;
         }
 
-        self.iter().zip(other.iter()).all(|(x, y)| x == y)
+        let (chunks_a, residual_a) = self.as_chunks::<4>();
+        let (chunks_b, residual_b) = other.as_chunks::<4>();
+
+        // check the residual first to bail out fast if there's a mismatch and comparisons
+        // happen to be expensive
+        let mut result = residual_a.into_iter().zip(residual_b).all(|(a, b)| a == b);
+
+        // iter.all short-circuits which means the backend can't unroll the loop due to early exits.
+        // So we unroll it manually.
+        result = result
+            && chunks_a
+                .iter()
+                .zip(chunks_b.iter())
+                .all(|(a, b)| (a[0] == b[0]) & (a[1] == b[1]) & (a[2] == b[2]) & (a[3] == b[3]));
+
+        result
     }
 }