Current ring selector logic follows round-robin scheduling,
when a process switch from one processor to another each time,
the data in the cache would have to be invalidated
and re-loaded from RAM leads to poor cache utilization.
Therefore smp_processor_id() attempt to keep processes
on the same processor via processor affinity which improves
cache utilization. Also, ring selection could be chosen round
robin when the processor_id based ring selection does not
have the resources.
Tested-on : IPQ8074 hw2.0 AHB WLAN.HK.2.4.0.1-00874-QCAHKSWPL_SILICONZ-1
Signed-off-by: P Praneesh <[email protected]>
---
drivers/net/wireless/ath/ath11k/core.h | 3 ---
drivers/net/wireless/ath/ath11k/dp_tx.c | 8 ++++----
2 files changed, 4 insertions(+), 7 deletions(-)
---
v3: corrected commit log.
v2: corrected tested tags.
---
diff --git a/drivers/net/wireless/ath/ath11k/core.h b/drivers/net/wireless/ath/ath11k/core.h
index 1cd9ef3..c2a4a7d 100644
--- a/drivers/net/wireless/ath/ath11k/core.h
+++ b/drivers/net/wireless/ath/ath11k/core.h
@@ -714,9 +714,6 @@ struct ath11k_base {
} stats;
u32 pktlog_defs_checksum;
- /* Round robbin based TCL ring selector */
- atomic_t tcl_ring_selector;
-
struct ath11k_dbring_cap *db_caps;
u32 num_db_cap;
diff --git a/drivers/net/wireless/ath/ath11k/dp_tx.c b/drivers/net/wireless/ath/ath11k/dp_tx.c
index a9632b7..897c368 100644
--- a/drivers/net/wireless/ath/ath11k/dp_tx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_tx.c
@@ -99,14 +99,14 @@ int ath11k_dp_tx(struct ath11k *ar, struct ath11k_vif *arvif,
pool_id = skb_get_queue_mapping(skb) & (ATH11K_HW_MAX_QUEUES - 1);
- /* Let the default ring selection be based on a round robin
- * fashion where one of the 3 tcl rings are selected based on
- * the tcl_ring_selector counter. In case that ring
+ /* Let the default ring selection be based on current processor
+ * number, where one of the 3 tcl rings are selected based on
+ * the smp_processor_id(). In case that ring
* is full/busy, we resort to other available rings.
* If all rings are full, we drop the packet.
* //TODO Add throttling logic when all rings are full
*/
- ring_selector = atomic_inc_return(&ab->tcl_ring_selector);
+ ring_selector = smp_processor_id();
tcl_ring_sel:
tcl_ring_retry = false;
--
2.7.4
P Praneesh <[email protected]> wrote:
> Current ring selector logic follows round-robin scheduling,
> when a process switch from one processor to another each time,
> the data in the cache would have to be invalidated
> and re-loaded from RAM leads to poor cache utilization.
>
> Therefore smp_processor_id() attempt to keep processes
> on the same processor via processor affinity which improves
> cache utilization. Also, ring selection could be chosen round
> robin when the processor_id based ring selection does not
> have the resources.
>
> Tested-on : IPQ8074 hw2.0 AHB WLAN.HK.2.4.0.1-00874-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: P Praneesh <[email protected]>
> Signed-off-by: Kalle Valo <[email protected]>
Patch applied to ath-next branch of ath.git, thanks.
0eaa4c1d34c5 ath11k: add processor_id based ring_selector logic
--
https://patchwork.kernel.org/project/linux-wireless/patch/[email protected]/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches