Rollup merge of rust-lang#107022 - scottmcm:ordering-option-eq, r=m-ou-se

Implement `SpecOptionPartialEq` for `cmp::Ordering`

Noticed as I continue to explore options for having code using `partial_cmp` optimize better.

Before:
```llvm
; Function Attrs: mustprogress nofree nosync nounwind willreturn uwtable
define noundef zeroext i1 `@ordering_eq(i8` noundef %0, i8 noundef %1) unnamed_addr #0 {
start:
  %2 = icmp eq i8 %0, 2
  br i1 %2, label %bb1.i, label %bb3.i

bb1.i:                                            ; preds = %start
  %3 = icmp eq i8 %1, 2
  br label %"_ZN55_$LT$T$u20$as$u20$core..option..SpecOptionPartialEq$GT$2eq17hb7e7beacecde585fE.exit"

bb3.i:                                            ; preds = %start
  %.not.i = icmp ne i8 %1, 2
  %4 = icmp eq i8 %0, %1
  %spec.select.i = and i1 %.not.i, %4
  br label %"_ZN55_$LT$T$u20$as$u20$core..option..SpecOptionPartialEq$GT$2eq17hb7e7beacecde585fE.exit"

"_ZN55_$LT$T$u20$as$u20$core..option..SpecOptionPartialEq$GT$2eq17hb7e7beacecde585fE.exit": ; preds = %bb1.i, %bb3.i
  %.0.i = phi i1 [ %3, %bb1.i ], [ %spec.select.i, %bb3.i ]
  ret i1 %.0.i
}
```

After:
```llvm
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn uwtable
define noundef zeroext i1 `@ordering_eq(i8` noundef %0, i8 noundef %1) unnamed_addr #1 {
start:
  %2 = icmp eq i8 %0, %1
  ret i1 %2
}
```

(Which <https://alive2.llvm.org/ce/z/-rop5r> says LLVM *could* just do itself, but there's probably an issue already open for that problem from when this was originally looked at for `Option<NonZeroU8>` and friends.)

Author: matthiaskrgr

library/core/src/option.rs

@@ -551,7 +551,7 @@ use crate::marker::Destruct;
 use crate::panicking::{panic, panic_str};
 use crate::pin::Pin;
 use crate::{
-    convert, hint, mem,
+    cmp, convert, hint, mem,
     ops::{self, ControlFlow, Deref, DerefMut},
 };
 
@@ -2090,6 +2090,12 @@ impl<T: PartialEq> PartialEq for Option<T> {
     }
 }
 
+/// This specialization trait is a workaround for LLVM not currently (2023-01)
+/// being able to optimize this itself, even though Alive confirms that it would
+/// be legal to do so: <https://github.com/llvm/llvm-project/issues/52622>
+///
+/// Once that's fixed, `Option` should go back to deriving `PartialEq`, as
+/// it used to do before <https://github.com/rust-lang/rust/pull/103556>.
 #[unstable(feature = "spec_option_partial_eq", issue = "none", reason = "exposed only for rustc")]
 #[doc(hidden)]
 pub trait SpecOptionPartialEq: Sized {
@@ -2146,6 +2152,14 @@ impl<T> SpecOptionPartialEq for crate::ptr::NonNull<T> {
     }
 }
 
+#[stable(feature = "rust1", since = "1.0.0")]
+impl SpecOptionPartialEq for cmp::Ordering {
+    #[inline]
+    fn eq(l: &Option<Self>, r: &Option<Self>) -> bool {
+        l.map_or(2, |x| x as i8) == r.map_or(2, |x| x as i8)
+    }
+}
+
 /////////////////////////////////////////////////////////////////////////////
 // The Option Iterators
 /////////////////////////////////////////////////////////////////////////////

tests/codegen/option-nonzero-eq.rs

@@ -3,6 +3,7 @@
 #![crate_type = "lib"]
 
 extern crate core;
+use core::cmp::Ordering;
 use core::num::{NonZeroU32, NonZeroI64};
 use core::ptr::NonNull;
 
@@ -32,3 +33,12 @@ pub fn non_null_eq(l: Option<NonNull<u8>>, r: Option<NonNull<u8>>) -> bool {
     // CHECK-NEXT: ret i1
     l == r
 }
+
+// CHECK-lABEL: @ordering_eq
+#[no_mangle]
+pub fn ordering_eq(l: Option<Ordering>, r: Option<Ordering>) -> bool {
+    // CHECK: start:
+    // CHECK-NEXT: icmp eq i8
+    // CHECK-NEXT: ret i1
+    l == r
+}