Details
-
Task
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
None
-
None
Description
First set of optimisations pushed to 10.5 (gives ~25% better throughput in sysbench connect benchmark):
54b81cf mysql_socket_accept() microoptimisations
|
a61baa7 Maintain connection_count atomically
|
0bee021 Simplified away wake_thread
|
34dfcbe Move thread re-initialisation out of cache_thread
|
ebc55c8 Simplified away scheduler_functions::end_thread()
|
6900aaf Simplified away init_new_connection_thread()
|
8d9d4aa Signal COND_thread_cache out of mutex
|
5d183df Try accept a few times before falling back to poll
|
8777540 Improved ha_close_connection() scalability
|
5e13943 Adiue thd_ha_data(), you've broke many hearts
|
ec926b0 Fixed RocksDB to follow THD ha_data protocol
|
8c8d584 Fixed InnoDB to not use broken thd_ha_data()
|
5e1b3cc Fixed Aria to follow THD ha_data protocol
|
ba59cc0 Fixed Sphinx to follow THD ha_data protocol
|
00e533c Fixed Mroonga to follow THD ha_data protocol
|
5c18ba6 Fixed Spider to follow THD ha_data protocol
|
762d2b9 Fixed FederatedX to follow THD ha_data protocol
|
ce30c99 Moved vio allocation to connection thread
|
efb61c1 Simplified away CONNECT::real_id
|
7192d7b Simplified away CONNECT::extra_port
|
c90c769 Simplified away CONNECT::host
|
56b1cdd Removed duplicate thread cache check
|
701e2a7 Optimised fcntl() when accepting connections
|
218a68b Less shared variables loads under the mutex
|
8268fa8 Moved set_timespec out of LOCK_thread_cache mutex
|
ca84758 Cleanup redundant abort_loop checks
|
Attachments
Issue Links
- relates to
-
-
- Closed
-
-
MDEV-25389 Assertion `!is_thread_specific || (mysqld_server_initialized && thd)' failed in void my_malloc_size_cb_func(long long int, my_bool)
-
- Closed
-