Auto merge of #85828 - scottmcm:raw-eq, r=oli-obk

Stop generating `alloca`s & `memcmp` for simple short array equality

Example:
```rust
pub fn demo(x: [u16; 6], y: [u16; 6]) -> bool { x == y }
```

Before:
```llvm
define zeroext i1 `@_ZN10playground4demo17h48537f7eac23948fE(i96` %0, i96 %1) unnamed_addr #0 {
start:
  %y = alloca [6 x i16], align 8
  %x = alloca [6 x i16], align 8
  %.0..sroa_cast = bitcast [6 x i16]* %x to i96*
  store i96 %0, i96* %.0..sroa_cast, align 8
  %.0..sroa_cast3 = bitcast [6 x i16]* %y to i96*
  store i96 %1, i96* %.0..sroa_cast3, align 8
  %_11.i.i.i = bitcast [6 x i16]* %x to i8*
  %_14.i.i.i = bitcast [6 x i16]* %y to i8*
  %bcmp.i.i.i = call i32 `@bcmp(i8*` nonnull dereferenceable(12) %_11.i.i.i, i8* nonnull dereferenceable(12) %_14.i.i.i, i64 12) #2, !alias.scope !2
  %2 = icmp eq i32 %bcmp.i.i.i, 0
  ret i1 %2
}
```
```x86
playground::demo: # `@playground::demo`
	sub	rsp, 32
	mov	qword ptr [rsp], rdi
	mov	dword ptr [rsp + 8], esi
	mov	qword ptr [rsp + 16], rdx
	mov	dword ptr [rsp + 24], ecx
	xor	rdi, rdx
	xor	esi, ecx
	or	rsi, rdi
	sete	al
	add	rsp, 32
	ret
```

After:
```llvm
define zeroext i1 `@_ZN4mini4demo17h7a8994aaa314c981E(i96` %0, i96 %1) unnamed_addr #0 {
start:
  %2 = icmp eq i96 %0, %1
  ret i1 %2
}
```
```x86
_ZN4mini4demo17h7a8994aaa314c981E:
	xor	rcx, r8
	xor	edx, r9d
	or	rdx, rcx
	sete	al
	ret
```

2 files changed, 74 insertions, 2 deletions

diff --git a/src/test/codegen/array-equality.rs b/src/test/codegen/array-equality.rs
new file mode 100644
index 00000000000..4b60fa4b0bf
--- /dev/null
+++ b/src/test/codegen/array-equality.rs
@@ -0,0 +1,57 @@
+// compile-flags: -O
+// only-x86_64
+
+#![crate_type = "lib"]
+
+// CHECK-LABEL: @array_eq_value
+#[no_mangle]
+pub fn array_eq_value(a: [u16; 6], b: [u16; 6]) -> bool {
+    // CHECK-NEXT: start:
+    // CHECK-NEXT: %2 = icmp eq i96 %0, %1
+    // CHECK-NEXT: ret i1 %2
+    a == b
+}
+
+// CHECK-LABEL: @array_eq_ref
+#[no_mangle]
+pub fn array_eq_ref(a: &[u16; 6], b: &[u16; 6]) -> bool {
+    // CHECK: start:
+    // CHECK: load i96, i96* %{{.+}}, align 2
+    // CHECK: load i96, i96* %{{.+}}, align 2
+    // CHECK: icmp eq i96
+    // CHECK-NEXT: ret
+    a == b
+}
+
+// CHECK-LABEL: @array_eq_value_still_passed_by_pointer
+#[no_mangle]
+pub fn array_eq_value_still_passed_by_pointer(a: [u16; 9], b: [u16; 9]) -> bool {
+    // CHECK-NEXT: start:
+    // CHECK-NEXT: bitcast
+    // CHECK-NEXT: bitcast
+    // CHECK-NEXT: %[[CMP:.+]] = tail call i32 @{{bcmp|memcmp}}(i8* nonnull dereferenceable(18) %{{.+}}, i8* nonnull dereferenceable(18) %{{.+}}, i64 18)
+    // CHECK-NEXT: %[[EQ:.+]] = icmp eq i32 %[[CMP]], 0
+    // CHECK-NEXT: ret i1 %[[EQ]]
+    a == b
+}
+
+// CHECK-LABEL: @array_eq_long
+#[no_mangle]
+pub fn array_eq_long(a: &[u16; 1234], b: &[u16; 1234]) -> bool {
+    // CHECK-NEXT: start:
+    // CHECK-NEXT: bitcast
+    // CHECK-NEXT: bitcast
+    // CHECK-NEXT: %[[CMP:.+]] = tail call i32 @{{bcmp|memcmp}}(i8* nonnull dereferenceable(2468) %{{.+}}, i8* nonnull dereferenceable(2468) %{{.+}}, i64 2468)
+    // CHECK-NEXT: %[[EQ:.+]] = icmp eq i32 %[[CMP]], 0
+    // CHECK-NEXT: ret i1 %[[EQ]]
+    a == b
+}
+
+// CHECK-LABEL: @array_eq_zero(i128 %0)
+#[no_mangle]
+pub fn array_eq_zero(x: [u16; 8]) -> bool {
+    // CHECK-NEXT: start:
+    // CHECK-NEXT: %[[EQ:.+]] = icmp eq i128 %0, 0
+    // CHECK-NEXT: ret i1 %[[EQ]]
+    x == [0; 8]
+}
diff --git a/src/test/codegen/slice-ref-equality.rs b/src/test/codegen/slice-ref-equality.rs
index acc7879e7b1..1f99ac7342b 100644
--- a/src/test/codegen/slice-ref-equality.rs
+++ b/src/test/codegen/slice-ref-equality.rs
@@ -2,15 +2,30 @@
 
 #![crate_type = "lib"]
 
-// #71602: check that slice equality just generates a single bcmp
+// #71602 reported a simple array comparison just generating a loop.
+// This was originally fixed by ensuring it generates a single bcmp,
+// but we now generate it as a load instead. `is_zero_slice` was
+// tweaked to still test the case of comparison against a slice,
+// and `is_zero_array` tests the new array-specific behaviour.
 
 // CHECK-LABEL: @is_zero_slice
 #[no_mangle]
 pub fn is_zero_slice(data: &[u8; 4]) -> bool {
-    // CHECK: start:
+    // CHECK: :
     // CHECK-NEXT: %{{.+}} = getelementptr {{.+}}
     // CHECK-NEXT: %[[BCMP:.+]] = tail call i32 @{{bcmp|memcmp}}({{.+}})
     // CHECK-NEXT: %[[EQ:.+]] = icmp eq i32 %[[BCMP]], 0
     // CHECK-NEXT: ret i1 %[[EQ]]
+    &data[..] == [0; 4]
+}
+
+// CHECK-LABEL: @is_zero_array
+#[no_mangle]
+pub fn is_zero_array(data: &[u8; 4]) -> bool {
+    // CHECK: start:
+    // CHECK-NEXT: %[[PTR:.+]] = bitcast [4 x i8]* {{.+}} to i32*
+    // CHECK-NEXT: %[[LOAD:.+]] = load i32, i32* %[[PTR]], align 1
+    // CHECK-NEXT: %[[EQ:.+]] = icmp eq i32 %[[LOAD]], 0
+    // CHECK-NEXT: ret i1 %[[EQ]]
     *data == [0; 4]
 }