[PATCH] malloc: Improve performance of __libc_malloc

[PATCH] malloc: Improve performance of __libc_malloc

[Wilco Dijkstra]

Improve performance of __libc_malloc by splitting it into 2 parts: first handle
the tcache fastpath, then do the rest in a separate tailcalled function.
This results in significant performance gains since __libc_malloc doesn't need
to setup a frame and we delay tcache initialization and setting of errno until
later.

On Neoverse V2, bench-malloc-simple improves by 6.7% overall (up to 8.5% for
ST case) and bench-malloc-thread improves by 19.4% for 1 thread and 14.2% for
32 threads.

Passes regress, OK for commit?

---

diff --git a/malloc/malloc.c b/malloc/malloc.c
index 06b508e22b14756d4aee8c6f62f0781557c7a96f..10526b58ed45d7733c8de9c70bc6ced300fedb05 100644
--- a/malloc/malloc.c
+++ b/malloc/malloc.c
@@ -1325,6 +1325,9 @@ nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 static __always_inline size_t
 checked_request2size (size_t req) __nonnull (1)
 {
+  _Static_assert (PTRDIFF_MAX <= SIZE_MAX / 2,
+                  "PTRDIFF_MAX is not more than half of SIZE_MAX");
+
   if (__glibc_unlikely (req > PTRDIFF_MAX))
     return 0;
 
@@ -3380,26 +3383,17 @@ tcache_thread_shutdown (void)
 #endif /* !USE_TCACHE  */
 
 #if IS_IN (libc)
-void *
-__libc_malloc (size_t bytes)
+
+static void * __attribute_noinline__
+__libc_malloc2 (size_t bytes)
 {
   mstate ar_ptr;
   void *victim;
 
-  _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);
-
-  if (err)
-      return NULL;
 
-  if (victim)
-      return tag_new_usable (victim);
-#endif
+  MAYBE_INIT_TCACHE ();
 
   if (SINGLE_THREAD_P)
     {
@@ -3430,6 +3424,19 @@ __libc_malloc (size_t bytes)
           ar_ptr == arena_for_chunk (mem2chunk (victim)));
   return victim;
 }
+
+void *
+__libc_malloc (size_t bytes)
+{
+#if USE_TCACHE
+  size_t tc_idx = csize2tidx (checked_request2size (bytes));
+
+  if (tcache_available (tc_idx))
+    return tcache_get (tc_idx);
+#endif
+
+  return __libc_malloc2 (bytes);
+}
 libc_hidden_def (__libc_malloc)
 
 void

Re: [PATCH] malloc: Improve performance of __libc_malloc

[Cupertino Miranda]

Hi Wilco,

Just wondering why you choose to set __attribute_noinline__ to 
__libc_malloc2 ?
I know it is tailcalled, but having a single caller, would it not make 
sense to inline?

Anything else and for what is worth, all seems Ok. ;-)

Cheers,
Cupertino

On 20-03-2025 21:11, Wilco Dijkstra wrote:
> 
> Improve performance of __libc_malloc by splitting it into 2 parts: first handle
> the tcache fastpath, then do the rest in a separate tailcalled function.
> This results in significant performance gains since __libc_malloc doesn't need
> to setup a frame and we delay tcache initialization and setting of errno until
> later.
> 
> On Neoverse V2, bench-malloc-simple improves by 6.7% overall (up to 8.5% for
> ST case) and bench-malloc-thread improves by 19.4% for 1 thread and 14.2% for
> 32 threads.
> 
> Passes regress, OK for commit?
> 
> ---
> 
> diff --git a/malloc/malloc.c b/malloc/malloc.c
> index 06b508e22b14756d4aee8c6f62f0781557c7a96f..10526b58ed45d7733c8de9c70bc6ced300fedb05 100644
> --- a/malloc/malloc.c
> +++ b/malloc/malloc.c
> @@ -1325,6 +1325,9 @@ nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>   static __always_inline size_t
>   checked_request2size (size_t req) __nonnull (1)
>   {
> +  _Static_assert (PTRDIFF_MAX <= SIZE_MAX / 2,
> +                  "PTRDIFF_MAX is not more than half of SIZE_MAX");
> +
>     if (__glibc_unlikely (req > PTRDIFF_MAX))
>       return 0;
>   
> @@ -3380,26 +3383,17 @@ tcache_thread_shutdown (void)
>   #endif /* !USE_TCACHE  */
>   
>   #if IS_IN (libc)
> -void *
> -__libc_malloc (size_t bytes)
> +
> +static void * __attribute_noinline__
> +__libc_malloc2 (size_t bytes)
>   {
>     mstate ar_ptr;
>     void *victim;
>   
> -  _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);
> -
> -  if (err)
> -      return NULL;
>   
> -  if (victim)
> -      return tag_new_usable (victim);
> -#endif
> +  MAYBE_INIT_TCACHE ();
>   
>     if (SINGLE_THREAD_P)
>       {
> @@ -3430,6 +3424,19 @@ __libc_malloc (size_t bytes)
>             ar_ptr == arena_for_chunk (mem2chunk (victim)));
>     return victim;
>   }
> +
> +void *
> +__libc_malloc (size_t bytes)
> +{
> +#if USE_TCACHE
> +  size_t tc_idx = csize2tidx (checked_request2size (bytes));
> +
> +  if (tcache_available (tc_idx))
> +    return tcache_get (tc_idx);
> +#endif
> +
> +  return __libc_malloc2 (bytes);
> +}
>   libc_hidden_def (__libc_malloc)
>   
>   void
> 
> 
> 

Re: [PATCH] malloc: Improve performance of __libc_malloc

[Wilco Dijkstra]

Hi Cupertino,

> Just wondering why you choose to set __attribute_noinline__ to
> __libc_malloc2 ?
> I know it is tailcalled, but having a single caller, would it not make
> sense to inline?

The tailcall means it ends up being a leaf function and thus doesn't
need a frame. If you inline it, you execute a prolog and epilog even
if you just use the tcache fastpath.

Cheers,
Wilco