malloc: Ensure mtag code path in checked_request2size is cold

This is a workaround (hack) for a gcc optimization issue (PR 99551).
Without this the generated code may evaluate the expression in the
cold path which causes performance regression for small allocations
in the memory tagging disabled (common) case.

diff --git a/malloc/malloc.c b/malloc/malloc.c
index cb1837d0d7818d028ff132fd88fecae4a60d933b..ef1f5b16210e80bfe624d41f28401efe8993230f 100644 (file)
--- a/malloc/malloc.c
+++ b/malloc/malloc.c
@@ -1357,8 +1357,13 @@ checked_request2size (size_t req, size_t *sz) __nonnull (1)
      must be a macro that produces a compile time constant if passed
      a constant literal.  */
   if (__glibc_unlikely (mtag_enabled))
-    req = (req + (__MTAG_GRANULE_SIZE - 1)) &
-         ~(size_t)(__MTAG_GRANULE_SIZE - 1);
+    {
+      /* Ensure this is not evaluated if !mtag_enabled, see gcc PR 99551.  */
+      asm ("");
+
+      req = (req + (__MTAG_GRANULE_SIZE - 1)) &
+           ~(size_t)(__MTAG_GRANULE_SIZE - 1);
+    }
 
   *sz = request2size (req);
   return true;