RepoMirrors/haproxy
1
0
Fork 0
mirror of http://git.haproxy.org/git/haproxy.git/ synced 2024-12-27 07:02:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

CLEANUP: ring: use only curr_cell and not next_cell in the main write loop

Browse Source 
It turns out that we can reduce by one variable in the loop and this
clobbers one less register, making it slightly faster on Cortex A72.
This commit is contained in:
Willy Tarreau 2024-03-17 16:54:36 +01:00
parent 0a0a64ef02
commit 4bc81ec985
1 changed files with 1 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
src/ring.c
View File

@ -266,8 +266,6 @@ ssize_t ring_write(struct ring *ring, size_t maxlen, const struct ist pfx[], siz
* comes in and becomes the leader in turn.
*/
next_cell = &cell;
/* Wait for another thread to take the lead or for the tail to
* be available again. It's critical to be read-only in this
* loop so as not to lose time synchronizing cache lines. Also,
@ -276,7 +274,7 @@ ssize_t ring_write(struct ring *ring, size_t maxlen, const struct ist pfx[], siz
*/
while (1) {
if ((next_cell = HA_ATOMIC_LOAD(ring_queue_ptr)) != &cell)
if ((curr_cell = HA_ATOMIC_LOAD(ring_queue_ptr)) != &cell)
goto wait_for_flush;
__ha_cpu_relax_for_read();
@ -296,7 +294,6 @@ ssize_t ring_write(struct ring *ring, size_t maxlen, const struct ist pfx[], siz
* which we'll confirm by trying to reset the queue. If we're
* still the leader, we're done.
*/
curr_cell = &cell;
if (HA_ATOMIC_CAS(ring_queue_ptr, &curr_cell, NULL))
break; // Won!