Speed up connection time:

-Change my_rnd() slightly to make it safer if two threads use it at the same time.
-Avoid some sprintf and strmov in vio.
-Changed thread_count to be automically incremented (instead of under LOCK_thread_count).
-Thread cache now uses LOCK_thread_cache instead of LOCK_thread_count.
-Moved delete thd out from LOCK_thread_count.
-Save some mysql_cond_broadcast(&COND_thread_count) calls.
-Removed call to getsockname() during connect.
-Initialize random generator without locks.

Other things:
-Added thread_safe_decrement32() and thread_safe_increment32()
-Removed sql_rnd_with_mutex() and get_thread_running()

Thanks to Yoshinori Matsunobu for the benchmark of connection speed and to
Domas Mituzas for the inspiration for many of the fixes.
 

include/violite.h:
  Change desc to a string pointer
mysql-test/suite/perfschema/r/all_instances.result:
  Added new mutex
mysys/my_rnd.c:
  Change my_rnd() slightly to make it safer if two threads use it at the same time.
sql/event_scheduler.cc:
  Changed thread_count to be automically incremented
  Moved some safe things out from LOCK_thread_count.
  Simplify deleting of THD for running thread.
sql/mysqld.cc:
  Changed thread_count to be automically incremented
  Thread cache now uses LOCK_thread_cache instead of LOCK_thread_count
  Added delete_running_thd()
  Moved delete thd out from LOCK_thread_count
  More DBUG_ENTER
  Only call  mysql_cond_broadcast(&COND_thread_count) if thread_count is 0
  Removed call to getsockname() (old not anymore needed check)
sql/mysqld.h:
  Removed sql_rnd_with_mutex() (not needed anymore)
  Removed not used function get_thread_running()
  Added thread_safe_decrement32() and thread_safe_increment32()
  Simplified dec_thread_running() and inc_thread_running()
sql/sql_class.cc:
  Removed thd_lock_thread_count() and thd_unlock_thread_count()
  Initialize random generator without locks
sql/sql_insert.cc:
  Changed thread_count to be automically incremented
sql/sql_parse.cc:
  Changed thread_count to be automically incremented
vio/vio.c:
  Don't generate 'desc' with sprintf/strmov. Assign a pointer instead.
  (Good enough as this is just for debugging)

include/violite.h

@@ -216,7 +216,7 @@ struct st_vio
   struct sockaddr_storage	remote;		/* Remote internet address */
   int addrLen;                          /* Length of remote address */
   enum enum_vio_type	type;		/* Type of connection */
-  char			desc[30];	/* String description */
+  const char		*desc;		/* String description */
   char                  *read_buffer;   /* buffer for vio_read_buff */
   char                  *read_pos;      /* start of unfetched data in the
                                            read buffer */

mysql-test/suite/perfschema/r/all_instances.result

@@ -64,6 +64,7 @@ wait/synch/mutex/sql/LOCK_server_started
 wait/synch/mutex/sql/LOCK_slave_list
 wait/synch/mutex/sql/LOCK_stats
 wait/synch/mutex/sql/LOCK_status
+wait/synch/mutex/sql/LOCK_thread_cache
 wait/synch/mutex/sql/LOCK_thread_count
 wait/synch/mutex/sql/LOCK_user_conn
 wait/synch/mutex/sql/LOCK_user_locks

mysys/my_rnd.c

@@ -45,11 +45,20 @@ void my_rnd_init(struct my_rnd_struct *rand_st, ulong seed1, ulong seed2)
     
   RETURN VALUE
     generated pseudo random number
+
+  NOTE:
+    This is codes so that it can be called by two threads at the same time
+    with minimum impact.
+    (As the number is supposed to be random, it doesn't matter much if
+    rand->seed1 or rand->seed2 are updated with slightly wrong numbers or
+    if two threads gets the same number.
 */
 
 double my_rnd(struct my_rnd_struct *rand_st)
 {
-  rand_st->seed1=(rand_st->seed1*3+rand_st->seed2) % rand_st->max_value;
-  rand_st->seed2=(rand_st->seed1+rand_st->seed2+33) % rand_st->max_value;
-  return (((double) rand_st->seed1)/rand_st->max_value_dbl);
+  unsigned long seed1;
+  seed1= (rand_st->seed1*3+rand_st->seed2) % rand_st->max_value;
+  rand_st->seed2=(seed1+rand_st->seed2+33) % rand_st->max_value;
+  rand_st->seed1= seed1;
+  return (((double) seed1)/rand_st->max_value_dbl);
 }

sql/event_scheduler.cc

@@ -132,11 +132,11 @@ post_init_event_thread(THD *thd)
     return TRUE;
   }
 
+  thread_safe_increment32(&thread_count, &thread_count_lock);
   mysql_mutex_lock(&LOCK_thread_count);
   threads.append(thd);
-  thread_count++;
-  inc_thread_running();
   mysql_mutex_unlock(&LOCK_thread_count);
+  inc_thread_running();
   return FALSE;
 }
 
@@ -154,12 +154,8 @@ deinit_event_thread(THD *thd)
 {
   thd->proc_info= "Clearing";
   DBUG_PRINT("exit", ("Event thread finishing"));
-  mysql_mutex_lock(&LOCK_thread_count);
-  thread_count--;
-  dec_thread_running();
-  delete thd;
-  mysql_cond_broadcast(&COND_thread_count);
-  mysql_mutex_unlock(&LOCK_thread_count);
+
+  delete_running_thd(thd);
 }
 
 
@@ -436,12 +432,7 @@ Event_scheduler::start()
     ret= TRUE;
 
     new_thd->proc_info= "Clearing";
-    mysql_mutex_lock(&LOCK_thread_count);
-    thread_count--;
-    dec_thread_running();
-    delete new_thd;
-    mysql_cond_broadcast(&COND_thread_count);
-    mysql_mutex_unlock(&LOCK_thread_count);
+    delete_running_thd(new_thd);
   }
 end:
   UNLOCK_DATA();
@@ -570,12 +561,7 @@ Event_scheduler::execute_top(Event_queue_element_for_exec *event_name)
   if (new_thd)
   {
     new_thd->proc_info= "Clearing";
-    mysql_mutex_lock(&LOCK_thread_count);
-    thread_count--;
-    dec_thread_running();
-    delete new_thd;
-    mysql_cond_broadcast(&COND_thread_count);
-    mysql_mutex_unlock(&LOCK_thread_count);
+    delete_running_thd(new_thd);
   }
   delete event_name;
   DBUG_RETURN(TRUE);

sql/mysqld.h

@@ -56,6 +56,7 @@ void kill_mysql(void);
 void close_connection(THD *thd, uint sql_errno= 0);
 void handle_connection_in_main_thread(THD *thd);
 void create_thread_to_handle_connection(THD *thd);
+void delete_running_thd(THD *thd);
 void unlink_thd(THD *thd);
 bool one_thread_per_connection_end(THD *thd, bool put_in_cache);
 void flush_thread_cache();
@@ -89,7 +90,6 @@ extern bool opt_ignore_builtin_innodb;
 extern my_bool opt_character_set_client_handshake;
 extern bool volatile abort_loop;
 extern bool in_bootstrap;
-extern uint volatile thread_count;
 extern uint connection_count;
 extern my_bool opt_safe_user_create;
 extern my_bool opt_safe_show_db, opt_local_infile, opt_myisam_use_mmap;
@@ -352,7 +352,8 @@ extern mysql_rwlock_t LOCK_system_variables_hash;
 extern mysql_cond_t COND_thread_count;
 extern mysql_cond_t COND_manager;
 extern int32 thread_running;
-extern my_atomic_rwlock_t thread_running_lock;
+extern int32 thread_count;
+extern my_atomic_rwlock_t thread_running_lock, thread_count_lock;
 
 extern char *opt_ssl_ca, *opt_ssl_capath, *opt_ssl_cert, *opt_ssl_cipher,
             *opt_ssl_key;
@@ -479,42 +480,30 @@ inline void table_case_convert(char * name, uint length)
                                      name, length, name, length);
 }
 
-inline ulong sql_rnd_with_mutex()
+inline void thread_safe_increment32(int32 *value, my_atomic_rwlock_t *lock)
 {
-  mysql_mutex_lock(&LOCK_thread_count);
-  ulong tmp=(ulong) (my_rnd(&sql_rand) * 0xffffffff); /* make all bits random */
-  mysql_mutex_unlock(&LOCK_thread_count);
-  return tmp;
+  my_atomic_rwlock_wrlock(lock);
+  (void) my_atomic_add32(value, 1);
+  my_atomic_rwlock_wrunlock(lock);
 }
 
-inline int32
-inc_thread_running()
+inline void thread_safe_decrement32(int32 *value, my_atomic_rwlock_t *lock)
 {
-  int32 num_thread_running;
-  my_atomic_rwlock_wrlock(&thread_running_lock);
-  num_thread_running= my_atomic_add32(&thread_running, 1);
-  my_atomic_rwlock_wrunlock(&thread_running_lock);
-  return (num_thread_running+1);
+  my_atomic_rwlock_wrlock(lock);
+  (void) my_atomic_add32(value, -1);
+  my_atomic_rwlock_wrunlock(lock);
 }
 
-inline int32
-dec_thread_running()
+inline void
+inc_thread_running()
 {
-  int32 num_thread_running;
-  my_atomic_rwlock_wrlock(&thread_running_lock);
-  num_thread_running= my_atomic_add32(&thread_running, -1);
-  my_atomic_rwlock_wrunlock(&thread_running_lock);
-  return (num_thread_running-1);
+  thread_safe_increment32(&thread_running, &thread_running_lock);
 }
 
-inline int32
-get_thread_running()
+inline void
+dec_thread_running()
 {
-  int32 num_thread_running;
-  my_atomic_rwlock_wrlock(&thread_running_lock);
-  num_thread_running= my_atomic_load32(&thread_running);
-  my_atomic_rwlock_wrunlock(&thread_running_lock);
-  return num_thread_running;
+  thread_safe_decrement32(&thread_running, &thread_running_lock);
 }
 
 void set_server_version(void);

sql/sql_class.cc

@@ -349,29 +349,6 @@ void thd_set_thread_stack(THD *thd, char *stack_start)
   thd->thread_stack= stack_start;
 }
 
-/**
-  Lock connection data for the set of connections this connection
-  belongs to
-
-  @param thd                       THD object
-*/
-void thd_lock_thread_count(THD *)
-{
-  mysql_mutex_lock(&LOCK_thread_count);
-}
-
-/**
-  Lock connection data for the set of connections this connection
-  belongs to
-
-  @param thd                       THD object
-*/
-void thd_unlock_thread_count(THD *)
-{
-  mysql_cond_broadcast(&COND_thread_count);
-  mysql_mutex_unlock(&LOCK_thread_count);
-}
-
 /**
   Close the socket used by this connection
 
@@ -947,7 +924,14 @@ THD::THD()
   protocol_binary.init(this);
 
   tablespace_op=FALSE;
-  tmp= sql_rnd_with_mutex();
+
+  /*
+    Initialize the random generator. We call my_rnd() without a lock as
+    it's not really critical if two threads modifies the structure at the
+    same time.  We ensure that we have an unique number foreach thread
+    by adding the address of the stack.
+  */
+  tmp= (ulong) (my_rnd(&sql_rand) * 0xffffffff);
   my_rnd_init(&rand, tmp + (ulong) &rand, tmp + (ulong) ::global_query_id);
   substitute_null_with_insert_id = FALSE;
   thr_lock_info_init(&lock_info); /* safety: will be reset after start */

sql/sql_insert.cc

@@ -2038,9 +2038,9 @@ class Delayed_insert :public ilink {
     thd.unlink();				// Must be unlinked under lock
     my_free(thd.query());
     thd.security_ctx->user= thd.security_ctx->host=0;
-    thread_count--;
     delayed_insert_threads--;
     mysql_mutex_unlock(&LOCK_thread_count);
+    thread_safe_decrement32(&thread_count, &thread_count_lock);
     mysql_cond_broadcast(&COND_thread_count); /* Tell main we are ready */
   }
 
@@ -2175,9 +2175,9 @@ bool delayed_get_table(THD *thd, MDL_request *grl_protection_request,
     {
       if (!(di= new Delayed_insert()))
         goto end_create;
-      mysql_mutex_lock(&LOCK_thread_count);
-      thread_count++;
-      mysql_mutex_unlock(&LOCK_thread_count);
+
+      thread_safe_increment32(&thread_count, &thread_count_lock);
+
       /*
         Annotating delayed inserts is not supported.
       */

sql/sql_parse.cc

@@ -645,9 +645,10 @@ void do_handle_bootstrap(THD *thd)
   delete thd;
 
 #ifndef EMBEDDED_LIBRARY
-  mysql_mutex_lock(&LOCK_thread_count);
-  thread_count--;
+  thread_safe_decrement32(&thread_count, &thread_count_lock);
   in_bootstrap= FALSE;
+
+  mysql_mutex_lock(&LOCK_thread_count);
   mysql_cond_broadcast(&COND_thread_count);
   mysql_mutex_unlock(&LOCK_thread_count);
   my_thread_end();

vio/vio.c

@@ -213,9 +213,7 @@ Vio *vio_new(my_socket sd, enum enum_vio_type type, uint flags)
   if ((vio = (Vio*) my_malloc(sizeof(*vio),MYF(MY_WME))))
   {
     vio_init(vio, type, sd, 0, flags);
-    sprintf(vio->desc,
-	    (vio->type == VIO_TYPE_SOCKET ? "socket (%d)" : "TCP/IP (%d)"),
-	    vio->sd);
+    vio->desc= (vio->type == VIO_TYPE_SOCKET ? "socket" : "TCP/IP");
 #if !defined(__WIN__)
 #if !defined(NO_FCNTL_NONBLOCK)
     /*
@@ -257,7 +255,7 @@ Vio *vio_new_win32pipe(HANDLE hPipe)
   if ((vio = (Vio*) my_malloc(sizeof(Vio),MYF(MY_WME))))
   {
     vio_init(vio, VIO_TYPE_NAMEDPIPE, 0, hPipe, VIO_LOCALHOST);
-    strmov(vio->desc, "named pipe");
+    vio->desc= "named pipe";
   }
   DBUG_RETURN(vio);
 }
@@ -282,7 +280,7 @@ Vio *vio_new_win32shared_memory(HANDLE handle_file_map, HANDLE handle_map,
     vio->event_conn_closed= event_conn_closed;
     vio->shared_memory_remain= 0;
     vio->shared_memory_pos= handle_map;
-    strmov(vio->desc, "shared memory");
+    vio->desc= "shared memory";
   }
   DBUG_RETURN(vio);
 }