[PATCH] malloc: Improve malloc initialization

[Wilco Dijkstra]

Move malloc initialization out of hot paths - it is only required after checking
fastbins in _int_malloc or before tcache initialization. Bench-malloc-thread
improves by 1.8% for 1 thread and 1.3% for 32 threads on Neoverse V2.

Passes regress, OK for commit?

---

diff --git a/malloc/malloc.c b/malloc/malloc.c
index b73ddbf554461da34d99258fae87c6ece6d175ba..b8eb4180570d20223109b1f0084f4a01bb97b9d6 100644
--- a/malloc/malloc.c
+++ b/malloc/malloc.c
@@ -3313,6 +3313,10 @@ tcache_init(void)
   if (tcache_shutting_down)
     return;
 
+  /* Ensure malloc is initialized before tcache.  */
+  if (!__malloc_initialized)
+    ptmalloc_init ();
+
   arena_get (ar_ptr, bytes);
   victim = _int_malloc (ar_ptr, bytes);
   if (!victim && ar_ptr != NULL)
@@ -3389,8 +3393,6 @@ __libc_malloc (size_t bytes)
   _Static_assert (PTRDIFF_MAX <= SIZE_MAX / 2,
                   "PTRDIFF_MAX is not more than half of SIZE_MAX");
 
-  if (!__malloc_initialized)
-    ptmalloc_init ();
 #if USE_TCACHE
   bool err = tcache_try_malloc (bytes, &victim);
 
@@ -3488,9 +3490,6 @@ __libc_realloc (void *oldmem, size_t bytes)
 
   void *newp;             /* chunk to return */
 
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
 #if REALLOC_ZERO_BYTES_FREES
   if (bytes == 0 && oldmem != NULL)
     {
@@ -3619,9 +3618,6 @@ libc_hidden_def (__libc_realloc)
 void *
 __libc_memalign (size_t alignment, size_t bytes)
 {
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
   void *address = RETURN_ADDRESS (0);
   return _mid_memalign (alignment, bytes, address);
 }
@@ -3632,9 +3628,6 @@ void *
 weak_function
 aligned_alloc (size_t alignment, size_t bytes)
 {
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
 /* Similar to memalign, but starting with ISO C17 the standard
    requires an error for alignments that are not supported by the
    implementation.  Valid alignments for the current implementation
@@ -3742,9 +3735,6 @@ _mid_memalign (size_t alignment, size_t bytes, void *address)
 void *
 __libc_valloc (size_t bytes)
 {
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
   void *address = RETURN_ADDRESS (0);
   size_t pagesize = GLRO (dl_pagesize);
   return _mid_memalign (pagesize, bytes, address);
@@ -3753,9 +3743,6 @@ __libc_valloc (size_t bytes)
 void *
 __libc_pvalloc (size_t bytes)
 {
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
   void *address = RETURN_ADDRESS (0);
   size_t pagesize = GLRO (dl_pagesize);
   size_t rounded_bytes;
@@ -3790,9 +3777,6 @@ __libc_calloc (size_t n, size_t elem_size)
 
   sz = bytes;
 
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
 #if USE_TCACHE
   bool err = tcache_try_malloc (bytes, &mem);
 
@@ -3816,7 +3800,7 @@ __libc_calloc (size_t n, size_t elem_size)
   else
     arena_get (av, sz);
 
-  if (av)
+  if (av && top (av) != NULL)
     {
       /* Check if we hand out the top chunk, in which case there may be no
 	 need to clear. */
@@ -4029,6 +4013,10 @@ _int_malloc (mstate av, size_t bytes)
 	}
     }
 
+  /* Ensure malloc is initialized before accessing small/large bins.  */
+  if (!__malloc_initialized)
+    ptmalloc_init ();
+
   /*
      If a small request, check regular bin.  Since these "smallbins"
      hold one size each, no searching within bins is necessary.
@@ -5848,9 +5836,6 @@ __posix_memalign (void **memptr, size_t alignment, size_t size)
 {
   void *mem;
 
-  if (!__malloc_initialized)
-    ptmalloc_init ();
-
   /* Test whether the SIZE argument is valid.  It must be a power of
      two multiple of sizeof (void *).  */
   if (alignment % sizeof (void *) != 0