feat: Make XNode field `slots` slab-friendly and cacheline-aligned (#1)

- 原本的XNode大小为544字节，对采用slab分配器的内核而言不友好，会实际分配1K。优化后实际分配576字节
- 由于原本xarray的slots数组不是cacheline对齐的，因此速度慢。换了之后benchmark性能有提升

启用feature： `slab-friendly`即可开启本优化

Signed-off-by: longjin <longjin@DragonOS.org>

Cargo.toml

@@ -6,8 +6,10 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-smallvec = "*"
+smallvec = "1.13.1"
 
 [features]
 default = []
 std = []
+# 内存排布对slab分配器更友好，内存用量更低
+slab-friendly = []

src/node.rs

@@ -103,7 +103,10 @@ where
     offset_in_parent: u8,
     /// The slots storing `XEntry`s, which point to user-given items for leaf nodes and other
     /// `XNode`s for interior nodes.
+    #[cfg(not(feature = "slab-friendly"))]
     slots: [XEntry<I>; SLOT_SIZE],
+    #[cfg(feature = "slab-friendly")]
+    slots: alloc::boxed::Box<[XEntry<I>; SLOT_SIZE]>,
     /// The marks representing whether each slot is marked or not.
     ///
     /// Users can set mark or unset mark on user-given items, and a leaf node or an interior node
@@ -120,7 +123,10 @@ impl<I: ItemEntry> XNode<I> {
         Self {
             height,
             offset_in_parent: offset,
+            #[cfg(not(feature = "slab-friendly"))]
             slots: [XEntry::EMPTY; SLOT_SIZE],
+            #[cfg(feature = "slab-friendly")]
+            slots: alloc::boxed::Box::new([XEntry::EMPTY; SLOT_SIZE]),
             marks: [Mark::EMPTY; NUM_MARKS],
         }
     }
@@ -147,7 +153,10 @@ impl<I: ItemEntry> XNode<I> {
     }
 
     pub fn entries_mut(&mut self) -> &mut [XEntry<I>] {
-        &mut self.slots
+        #[cfg(not(feature = "slab-friendly"))]
+        return &mut self.slots;
+        #[cfg(feature = "slab-friendly")]
+        return self.slots.as_mut();
     }
 
     pub fn is_marked(&self, offset: u8, mark: usize) -> bool {