Optimise SmolStr::clone 4-5x speedup inline, 0.5x heap (slow down)

Author: alexheretic

lib/smol_str/CHANGELOG.md

@@ -1,6 +1,7 @@
 # Changelog
 
 ## Unreleased
+- Optimise `SmolStr::clone` 4-5x speedup inline, 0.5x heap (slow down).
 
 ## 0.3.4 - 2025-10-23
 

lib/smol_str/src/lib.rs

@@ -104,11 +104,19 @@ impl SmolStr {
 impl Clone for SmolStr {
     #[inline]
     fn clone(&self) -> Self {
+        // hint for faster inline / slower heap clones
+        #[cold]
+        #[inline(never)]
+        fn cold_clone(v: &SmolStr) -> SmolStr {
+            SmolStr(v.0.clone())
+        }
+
         if !self.is_heap_allocated() {
             // SAFETY: We verified that the payload of `Repr` is a POD
             return unsafe { core::ptr::read(self as *const SmolStr) };
         }
-        Self(self.0.clone())
+
+        cold_clone(self)
     }
 }