A more efficient slice comparison implementation for T: !BytewiseEq

The previous implementation was not optimized properly by the compiler, which didn't leverage the fact that both length were equal.
rust-lang · Oct 17, 2023 · 9348c33 · 9348c33
1 parent 347452e
commit 9348c33
Show file tree

Hide file tree

Showing 4 changed files with 162 additions and 1 deletion.
diff --git a/library/core/src/slice/cmp.rs b/library/core/src/slice/cmp.rs
@@ -60,7 +60,24 @@ where
             return false;
         }
 
-        self.iter().zip(other.iter()).all(|(x, y)| x == y)
+        let mut i = self.len();
+        let mut ptr_self = self.as_ptr();
+        let mut ptr_other = other.as_ptr();
+        // SAFETY:
+        // This is sound because:
+        // - self.len == other.len
+        // - self.len <= isize::MAX
+        // so the two pointers will not overflow,
+        // will remain in bounds of the slice,
+        // and dereferencing them is sound.
+        unsafe {
+            while (i > 0) && (*ptr_self == *ptr_other) {
+                i -= 1;
+                ptr_self = ptr_self.add(1);
+                ptr_other = ptr_other.add(1);
+            }
+        }
+        i == 0
     }
 }
 

diff --git a/mre.ll b/mre.ll
@@ -0,0 +1,97 @@
+; ModuleID = 'mre.44c9d9ceb0d5fb53-cgu.0'
+source_filename = "mre.44c9d9ceb0d5fb53-cgu.0"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nofree nosync nounwind nonlazybind memory(read) uwtable
+define noundef zeroext i1 @test(ptr noalias nocapture noundef nonnull readonly align 8 %0, i64 noundef %1, ptr noalias nocapture noundef nonnull readonly align 8 %2, i64 noundef %3) unnamed_addr #0 {
+start:
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !3)
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !6)
+  %_3.not.i = icmp eq i64 %1, %3
+  br i1 %_3.not.i, label %bb3.preheader.i, label %"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E.exit"
+
+bb3.preheader.i:                                  ; preds = %start
+  %_9.not4.i = icmp eq i64 %1, 0
+  br i1 %_9.not4.i, label %"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E.exit", label %bb4.i
+
+bb4.i:                                            ; preds = %bb3.preheader.i, %bb6.i
+  %i.07.i = phi i64 [ %6, %bb6.i ], [ %1, %bb3.preheader.i ]
+  %ptr_self.06.i = phi ptr [ %_14.i, %bb6.i ], [ %0, %bb3.preheader.i ]
+  %ptr_other.05.i = phi ptr [ %_16.i, %bb6.i ], [ %2, %bb3.preheader.i ]
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !8)
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !11)
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !13)
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !16)
+  %_5.i.i.i = load i64, ptr %ptr_self.06.i, align 8, !range !18, !alias.scope !19, !noalias !20, !noundef !21
+  %trunc.not.i.i.i = icmp eq i64 %_5.i.i.i, 0
+  %_3.i.i.i = load i64, ptr %ptr_other.05.i, align 8, !range !18, !alias.scope !20, !noalias !19, !noundef !21
+  %4 = icmp eq i64 %_3.i.i.i, 0
+  %brmerge.i.i.i = or i1 %trunc.not.i.i.i, %4
+  br i1 %brmerge.i.i.i, label %"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E.exit.i", label %bb4.i.i.i
+
+bb4.i.i.i:                                        ; preds = %bb4.i
+  %_8.i.i.i = getelementptr inbounds { i64, i64 }, ptr %ptr_self.06.i, i64 0, i32 1
+  %_9.i.i.i = getelementptr inbounds { i64, i64 }, ptr %ptr_other.05.i, i64 0, i32 1
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !22)
+  tail call void @llvm.experimental.noalias.scope.decl(metadata !25)
+  %_3.i.i.i.i = load i64, ptr %_8.i.i.i, align 8, !alias.scope !27, !noalias !28, !noundef !21
+  %_4.i.i.i.i = load i64, ptr %_9.i.i.i, align 8, !alias.scope !28, !noalias !27, !noundef !21
+  %_0.i.i.i.i = icmp eq i64 %_3.i.i.i.i, %_4.i.i.i.i
+  br i1 %_0.i.i.i.i, label %bb6.i, label %"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E.exit"
+
+"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E.exit.i": ; preds = %bb4.i
+  %5 = or i64 %_3.i.i.i, %_5.i.i.i
+  %.mux.i.i.i = icmp eq i64 %5, 0
+  br i1 %.mux.i.i.i, label %bb6.i, label %"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E.exit"
+
+bb6.i:                                            ; preds = %"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E.exit.i", %bb4.i.i.i
+  %6 = add i64 %i.07.i, -1
+  %_14.i = getelementptr inbounds { i64, i64 }, ptr %ptr_self.06.i, i64 1
+  %_16.i = getelementptr inbounds { i64, i64 }, ptr %ptr_other.05.i, i64 1
+  %_9.not.i = icmp eq i64 %6, 0
+  br i1 %_9.not.i, label %"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E.exit", label %bb4.i
+
+"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E.exit": ; preds = %bb4.i.i.i, %"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E.exit.i", %bb6.i, %start, %bb3.preheader.i
+  %_0.0.i = phi i1 [ false, %start ], [ true, %bb3.preheader.i ], [ true, %bb6.i ], [ false, %"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E.exit.i" ], [ false, %bb4.i.i.i ]
+  ret i1 %_0.0.i
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite)
+declare void @llvm.experimental.noalias.scope.decl(metadata) #1
+
+attributes #0 = { nofree nosync nounwind nonlazybind memory(read) uwtable "probe-stack"="inline-asm" "target-cpu"="x86-64" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite) }
+
+!llvm.module.flags = !{!0, !1}
+!llvm.ident = !{!2}
+
+!0 = !{i32 8, !"PIC Level", i32 2}
+!1 = !{i32 2, !"RtLibUseGOT", i32 1}
+!2 = !{!"rustc version 1.74.0-beta.1 (b5c050feb 2023-10-03)"}
+!3 = !{!4}
+!4 = distinct !{!4, !5, !"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E: %self.0"}
+!5 = distinct !{!5, !"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E"}
+!6 = !{!7}
+!7 = distinct !{!7, !5, !"_ZN73_$LT$$u5b$A$u5d$$u20$as$u20$core..slice..cmp..SlicePartialEq$LT$B$GT$$GT$5equal17h92b3a432973c4ce3E: %other.0"}
+!8 = !{!9}
+!9 = distinct !{!9, !10, !"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E: %self"}
+!10 = distinct !{!10, !"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E"}
+!11 = !{!12}
+!12 = distinct !{!12, !10, !"_ZN70_$LT$core..option..Option$LT$T$GT$$u20$as$u20$core..cmp..PartialEq$GT$2eq17h839ffac60ba8c424E: %other"}
+!13 = !{!14}
+!14 = distinct !{!14, !15, !"_ZN55_$LT$T$u20$as$u20$core..option..SpecOptionPartialEq$GT$2eq17h42d3740d8c76b9d0E: %l"}
+!15 = distinct !{!15, !"_ZN55_$LT$T$u20$as$u20$core..option..SpecOptionPartialEq$GT$2eq17h42d3740d8c76b9d0E"}
+!16 = !{!17}
+!17 = distinct !{!17, !15, !"_ZN55_$LT$T$u20$as$u20$core..option..SpecOptionPartialEq$GT$2eq17h42d3740d8c76b9d0E: %r"}
+!18 = !{i64 0, i64 2}
+!19 = !{!14, !9, !4}
+!20 = !{!17, !12, !7}
+!21 = !{}
+!22 = !{!23}
+!23 = distinct !{!23, !24, !"_ZN4core3cmp5impls54_$LT$impl$u20$core..cmp..PartialEq$u20$for$u20$u64$GT$2eq17hcb36ad6f45b649e4E: %self"}
+!24 = distinct !{!24, !"_ZN4core3cmp5impls54_$LT$impl$u20$core..cmp..PartialEq$u20$for$u20$u64$GT$2eq17hcb36ad6f45b649e4E"}
+!25 = !{!26}
+!26 = distinct !{!26, !24, !"_ZN4core3cmp5impls54_$LT$impl$u20$core..cmp..PartialEq$u20$for$u20$u64$GT$2eq17hcb36ad6f45b649e4E: %other"}
+!27 = !{!23, !14, !9, !4}
+!28 = !{!26, !17, !12, !7}
diff --git a/mre.rs b/mre.rs
@@ -0,0 +1,4 @@
+#[no_mangle]
+pub fn test(x: &[Option<u64>], y: &[Option<u64>]) -> bool {
+    x == y
+}
diff --git a/mre.s b/mre.s
@@ -0,0 +1,43 @@
+	.text
+	.file	"mre.44c9d9ceb0d5fb53-cgu.0"
+	.section	.text.test,"ax",@progbits
+	.globl	test
+	.p2align	4, 0x90
+	.type	test,@function
+test:
+	.cfi_startproc
+	cmpq	%rcx, %rsi
+	jne	.LBB0_9
+	movb	$1, %al
+	testq	%rsi, %rsi
+	je	.LBB0_10
+	movl	$8, %ecx
+	jmp	.LBB0_5
+	.p2align	4, 0x90
+.LBB0_3:
+	orq	%r8, %r9
+	jne	.LBB0_9
+.LBB0_4:
+	addq	$16, %rcx
+	decq	%rsi
+	je	.LBB0_10
+.LBB0_5:
+	movq	-8(%rdi,%rcx), %r8
+	movq	-8(%rdx,%rcx), %r9
+	testq	%r8, %r8
+	je	.LBB0_3
+	testq	%r9, %r9
+	je	.LBB0_3
+	movq	(%rdi,%rcx), %r8
+	cmpq	(%rdx,%rcx), %r8
+	je	.LBB0_4
+.LBB0_9:
+	xorl	%eax, %eax
+.LBB0_10:
+	retq
+.Lfunc_end0:
+	.size	test, .Lfunc_end0-test
+	.cfi_endproc
+
+	.ident	"rustc version 1.74.0-beta.1 (b5c050feb 2023-10-03)"
+	.section	".note.GNU-stack","",@progbits