* [PATCH 0/2] Change sk_pacing_shift in ieee80211_hw for best tx throughput
@ 2018-07-26 7:16 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:16 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
From: Wen Gong <wgong@codeaurora.org>
Add a field for ath10k to adjust the sk_pacing_shift, mac80211 set
the default value to 8, and ath10k will change it to 6. Then mac80211
will use the changed value 6 as sk_pacing_shift since 6 is the best
value for tx throughput by test result.
Wen Gong (2):
mac80211: Change sk_pacing_shift saved to ieee80211_hw
ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
include/net/mac80211.h | 1 +
net/mac80211/main.c | 2 ++
net/mac80211/tx.c | 2 +-
4 files changed, 6 insertions(+), 1 deletion(-)
--
1.9.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [PATCH 0/2] Change sk_pacing_shift in ieee80211_hw for best tx throughput
@ 2018-07-26 7:16 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:16 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
From: Wen Gong <wgong@codeaurora.org>
Add a field for ath10k to adjust the sk_pacing_shift, mac80211 set
the default value to 8, and ath10k will change it to 6. Then mac80211
will use the changed value 6 as sk_pacing_shift since 6 is the best
value for tx throughput by test result.
Wen Gong (2):
mac80211: Change sk_pacing_shift saved to ieee80211_hw
ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
include/net/mac80211.h | 1 +
net/mac80211/main.c | 2 ++
net/mac80211/tx.c | 2 +-
4 files changed, 6 insertions(+), 1 deletion(-)
--
1.9.1
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
* [PATCH 1/2] mac80211: Save sk_pacing_shift to ieee80211_hw
2018-07-26 7:16 ` Wen Gong
@ 2018-07-26 7:16 ` Wen Gong
-1 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:16 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
From: Wen Gong <wgong@codeaurora.org>
Add the skb_pacing_shirt adjustment, but this change make it
configurable for other driver. If no other driver set it, then
mac8011 will use the default value.
Signed-off-by: Wen Gong <wgong@codeaurora.org>
---
include/net/mac80211.h | 1 +
net/mac80211/main.c | 2 ++
net/mac80211/tx.c | 2 +-
3 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/include/net/mac80211.h b/include/net/mac80211.h
index 851a5e1..f600d08 100644
--- a/include/net/mac80211.h
+++ b/include/net/mac80211.h
@@ -2279,6 +2279,7 @@ struct ieee80211_hw {
u8 n_cipher_schemes;
const struct ieee80211_cipher_scheme *cipher_schemes;
u8 max_nan_de_entries;
+ u8 tx_sk_pacing_shift;
};
static inline bool _ieee80211_hw_check(struct ieee80211_hw *hw,
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index fb73451..72dce49 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -584,6 +584,8 @@ struct ieee80211_hw *ieee80211_alloc_hw_nm(size_t priv_data_len,
local->ops = ops;
local->use_chanctx = use_chanctx;
+ local->hw.tx_sk_pacing_shift = 8;
+
/* set up some defaults */
local->hw.queues = 1;
local->hw.max_rates = 1;
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 44b5dfe..a2d61b9 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -3596,7 +3596,7 @@ void __ieee80211_subif_start_xmit(struct sk_buff *skb,
* second, so 8 is ~4ms of queued data. Only affects local TCP
* sockets.
*/
- sk_pacing_shift_update(skb->sk, 8);
+ sk_pacing_shift_update(skb->sk, sdata->local->hw.tx_sk_pacing_shift);
fast_tx = rcu_dereference(sta->fast_tx);
--
1.9.1
^ permalink raw reply related [flat|nested] 20+ messages in thread
* [PATCH 1/2] mac80211: Save sk_pacing_shift to ieee80211_hw
@ 2018-07-26 7:16 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:16 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
From: Wen Gong <wgong@codeaurora.org>
Add the skb_pacing_shirt adjustment, but this change make it
configurable for other driver. If no other driver set it, then
mac8011 will use the default value.
Signed-off-by: Wen Gong <wgong@codeaurora.org>
---
include/net/mac80211.h | 1 +
net/mac80211/main.c | 2 ++
net/mac80211/tx.c | 2 +-
3 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/include/net/mac80211.h b/include/net/mac80211.h
index 851a5e1..f600d08 100644
--- a/include/net/mac80211.h
+++ b/include/net/mac80211.h
@@ -2279,6 +2279,7 @@ struct ieee80211_hw {
u8 n_cipher_schemes;
const struct ieee80211_cipher_scheme *cipher_schemes;
u8 max_nan_de_entries;
+ u8 tx_sk_pacing_shift;
};
static inline bool _ieee80211_hw_check(struct ieee80211_hw *hw,
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index fb73451..72dce49 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -584,6 +584,8 @@ struct ieee80211_hw *ieee80211_alloc_hw_nm(size_t priv_data_len,
local->ops = ops;
local->use_chanctx = use_chanctx;
+ local->hw.tx_sk_pacing_shift = 8;
+
/* set up some defaults */
local->hw.queues = 1;
local->hw.max_rates = 1;
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 44b5dfe..a2d61b9 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -3596,7 +3596,7 @@ void __ieee80211_subif_start_xmit(struct sk_buff *skb,
* second, so 8 is ~4ms of queued data. Only affects local TCP
* sockets.
*/
- sk_pacing_shift_update(skb->sk, 8);
+ sk_pacing_shift_update(skb->sk, sdata->local->hw.tx_sk_pacing_shift);
fast_tx = rcu_dereference(sta->fast_tx);
--
1.9.1
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply related [flat|nested] 20+ messages in thread
* [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-26 7:16 ` Wen Gong
@ 2018-07-26 7:16 ` Wen Gong
-1 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:16 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
From: Wen Gong <wgong@codeaurora.org>
Upstream kernel has an interface to help adjust sk_pacing_shift to help
improve TCP UL throughput.
The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
VHT80 TCP UL throughput testing result shows 6 is the optimal.
Overwrite the sk_pacing_shift to 6 in ath10k driver.
Tested with QCA6174 PCI with firmware
WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
It's not a regression with new firmware releases.
There have 2 test result of different settings:
ARM CPU based device with QCA6174A PCI with different
sk_pacing_shift:
sk_pacing_shift throughput(Mbps) CPU utilization
6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
8 288 ~90% idle, Focus on CPU1: ~35%idle
9 ~200 ~92% idle, Focus on CPU1: ~50%idle
5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
set to 6:
tcp_limit_output_bytes throughput(Mbps)
default(262144)+1 Stream 336
default(262144)+2 Streams 558
default(262144)+3 Streams 584
default(262144)+4 Streams 602
default(262144)+5 Streams 598
changed(2621440)+1 Stream 598
changed(2621440)+2 Streams 601
Signed-off-by: Wen Gong <wgong@codeaurora.org>
---
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index f31ae3b..40d24c1 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -8348,6 +8348,8 @@ int ath10k_mac_register(struct ath10k *ar)
ar->hw->wiphy->max_scan_ssids = WLAN_SCAN_PARAMS_MAX_SSID;
ar->hw->wiphy->max_scan_ie_len = WLAN_SCAN_PARAMS_MAX_IE_LEN;
+ ar->hw->tx_sk_pacing_shift = 6;
+
ar->hw->vif_data_size = sizeof(struct ath10k_vif);
ar->hw->sta_data_size = sizeof(struct ath10k_sta);
ar->hw->txq_data_size = sizeof(struct ath10k_txq);
--
1.9.1
^ permalink raw reply related [flat|nested] 20+ messages in thread
* [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-26 7:16 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:16 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
From: Wen Gong <wgong@codeaurora.org>
Upstream kernel has an interface to help adjust sk_pacing_shift to help
improve TCP UL throughput.
The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
VHT80 TCP UL throughput testing result shows 6 is the optimal.
Overwrite the sk_pacing_shift to 6 in ath10k driver.
Tested with QCA6174 PCI with firmware
WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
It's not a regression with new firmware releases.
There have 2 test result of different settings:
ARM CPU based device with QCA6174A PCI with different
sk_pacing_shift:
sk_pacing_shift throughput(Mbps) CPU utilization
6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
8 288 ~90% idle, Focus on CPU1: ~35%idle
9 ~200 ~92% idle, Focus on CPU1: ~50%idle
5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
set to 6:
tcp_limit_output_bytes throughput(Mbps)
default(262144)+1 Stream 336
default(262144)+2 Streams 558
default(262144)+3 Streams 584
default(262144)+4 Streams 602
default(262144)+5 Streams 598
changed(2621440)+1 Stream 598
changed(2621440)+2 Streams 601
Signed-off-by: Wen Gong <wgong@codeaurora.org>
---
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index f31ae3b..40d24c1 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -8348,6 +8348,8 @@ int ath10k_mac_register(struct ath10k *ar)
ar->hw->wiphy->max_scan_ssids = WLAN_SCAN_PARAMS_MAX_SSID;
ar->hw->wiphy->max_scan_ie_len = WLAN_SCAN_PARAMS_MAX_IE_LEN;
+ ar->hw->tx_sk_pacing_shift = 6;
+
ar->hw->vif_data_size = sizeof(struct ath10k_vif);
ar->hw->sta_data_size = sizeof(struct ath10k_sta);
ar->hw->txq_data_size = sizeof(struct ath10k_txq);
--
1.9.1
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply related [flat|nested] 20+ messages in thread
* [PATCH 0/2] Change sk_pacing_shift in ieee80211_hw for best tx throughput
@ 2018-07-26 7:21 Wen Gong
2018-07-26 7:21 ` Wen Gong
0 siblings, 1 reply; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:21 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
Add a field for ath10k to adjust the sk_pacing_shift, mac80211 set
the default value to 8, and ath10k will change it to 6. Then mac80211
will use the changed value 6 as sk_pacing_shift since 6 is the best
value for tx throughput by test result.
Wen Gong (2):
mac80211: Change sk_pacing_shift saved to ieee80211_hw
ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
include/net/mac80211.h | 1 +
net/mac80211/main.c | 2 ++
net/mac80211/tx.c | 2 +-
4 files changed, 6 insertions(+), 1 deletion(-)
--
1.9.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-26 7:21 [PATCH 0/2] Change sk_pacing_shift in ieee80211_hw for best tx throughput Wen Gong
@ 2018-07-26 7:21 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:21 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
Upstream kernel has an interface to help adjust sk_pacing_shift to help
improve TCP UL throughput.
The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
VHT80 TCP UL throughput testing result shows 6 is the optimal.
Overwrite the sk_pacing_shift to 6 in ath10k driver.
Tested with QCA6174 PCI with firmware
WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
It's not a regression with new firmware releases.
There have 2 test result of different settings:
ARM CPU based device with QCA6174A PCI with different
sk_pacing_shift:
sk_pacing_shift throughput(Mbps) CPU utilization
6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
8 288 ~90% idle, Focus on CPU1: ~35%idle
9 ~200 ~92% idle, Focus on CPU1: ~50%idle
5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
set to 6:
tcp_limit_output_bytes throughput(Mbps)
default(262144)+1 Stream 336
default(262144)+2 Streams 558
default(262144)+3 Streams 584
default(262144)+4 Streams 602
default(262144)+5 Streams 598
changed(2621440)+1 Stream 598
changed(2621440)+2 Streams 601
Signed-off-by: Wen Gong <wgong@codeaurora.org>
---
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index f31ae3b..40d24c1 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -8348,6 +8348,8 @@ int ath10k_mac_register(struct ath10k *ar)
ar->hw->wiphy->max_scan_ssids = WLAN_SCAN_PARAMS_MAX_SSID;
ar->hw->wiphy->max_scan_ie_len = WLAN_SCAN_PARAMS_MAX_IE_LEN;
+ ar->hw->tx_sk_pacing_shift = 6;
+
ar->hw->vif_data_size = sizeof(struct ath10k_vif);
ar->hw->sta_data_size = sizeof(struct ath10k_sta);
ar->hw->txq_data_size = sizeof(struct ath10k_txq);
--
1.9.1
^ permalink raw reply related [flat|nested] 20+ messages in thread
* [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-26 7:21 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-26 7:21 UTC (permalink / raw)
To: ath10k, johannes; +Cc: linux-wireless
Upstream kernel has an interface to help adjust sk_pacing_shift to help
improve TCP UL throughput.
The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
VHT80 TCP UL throughput testing result shows 6 is the optimal.
Overwrite the sk_pacing_shift to 6 in ath10k driver.
Tested with QCA6174 PCI with firmware
WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
It's not a regression with new firmware releases.
There have 2 test result of different settings:
ARM CPU based device with QCA6174A PCI with different
sk_pacing_shift:
sk_pacing_shift throughput(Mbps) CPU utilization
6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
8 288 ~90% idle, Focus on CPU1: ~35%idle
9 ~200 ~92% idle, Focus on CPU1: ~50%idle
5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
set to 6:
tcp_limit_output_bytes throughput(Mbps)
default(262144)+1 Stream 336
default(262144)+2 Streams 558
default(262144)+3 Streams 584
default(262144)+4 Streams 602
default(262144)+5 Streams 598
changed(2621440)+1 Stream 598
changed(2621440)+2 Streams 601
Signed-off-by: Wen Gong <wgong@codeaurora.org>
---
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index f31ae3b..40d24c1 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -8348,6 +8348,8 @@ int ath10k_mac_register(struct ath10k *ar)
ar->hw->wiphy->max_scan_ssids = WLAN_SCAN_PARAMS_MAX_SSID;
ar->hw->wiphy->max_scan_ie_len = WLAN_SCAN_PARAMS_MAX_IE_LEN;
+ ar->hw->tx_sk_pacing_shift = 6;
+
ar->hw->vif_data_size = sizeof(struct ath10k_vif);
ar->hw->sta_data_size = sizeof(struct ath10k_sta);
ar->hw->txq_data_size = sizeof(struct ath10k_txq);
--
1.9.1
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply related [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-26 7:21 ` Wen Gong
@ 2018-07-26 11:45 ` Toke Høiland-Jørgensen
-1 siblings, 0 replies; 20+ messages in thread
From: Toke Høiland-Jørgensen @ 2018-07-26 11:45 UTC (permalink / raw)
To: Wen Gong, ath10k, johannes; +Cc: linux-wireless
Wen Gong <wgong@codeaurora.org> writes:
> Upstream kernel has an interface to help adjust sk_pacing_shift to help
> improve TCP UL throughput.
> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
> VHT80 TCP UL throughput testing result shows 6 is the optimal.
> Overwrite the sk_pacing_shift to 6 in ath10k driver.
When I tested this, a pacing shift of 8 was quite close to optimal as
well for ath10k. Why are you getting different results?
> Tested with QCA6174 PCI with firmware
> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
> It's not a regression with new firmware releases.
>
> There have 2 test result of different settings:
>
> ARM CPU based device with QCA6174A PCI with different
> sk_pacing_shift:
>
> sk_pacing_shift throughput(Mbps) CPU utilization
> 6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
> 7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
> 8 288 ~90% idle, Focus on CPU1: ~35%idle
> 9 ~200 ~92% idle, Focus on CPU1: ~50%idle
Your tests do not include latency values; please try running a test that
also measures latency. The tcp_nup test in Flent (https://flent.org)
will do that, for instance. Also, is this a single TCP flow?
-Toke
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-26 11:45 ` Toke Høiland-Jørgensen
0 siblings, 0 replies; 20+ messages in thread
From: Toke Høiland-Jørgensen @ 2018-07-26 11:45 UTC (permalink / raw)
To: Wen Gong, ath10k, johannes; +Cc: linux-wireless
Wen Gong <wgong@codeaurora.org> writes:
> Upstream kernel has an interface to help adjust sk_pacing_shift to help
> improve TCP UL throughput.
> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
> VHT80 TCP UL throughput testing result shows 6 is the optimal.
> Overwrite the sk_pacing_shift to 6 in ath10k driver.
When I tested this, a pacing shift of 8 was quite close to optimal as
well for ath10k. Why are you getting different results?
> Tested with QCA6174 PCI with firmware
> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
> It's not a regression with new firmware releases.
>
> There have 2 test result of different settings:
>
> ARM CPU based device with QCA6174A PCI with different
> sk_pacing_shift:
>
> sk_pacing_shift throughput(Mbps) CPU utilization
> 6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
> 7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
> 8 288 ~90% idle, Focus on CPU1: ~35%idle
> 9 ~200 ~92% idle, Focus on CPU1: ~50%idle
Your tests do not include latency values; please try running a test that
also measures latency. The tcp_nup test in Flent (https://flent.org)
will do that, for instance. Also, is this a single TCP flow?
-Toke
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-26 11:45 ` Toke Høiland-Jørgensen
@ 2018-07-26 13:02 ` Michał Kazior
-1 siblings, 0 replies; 20+ messages in thread
From: Michał Kazior @ 2018-07-26 13:02 UTC (permalink / raw)
To: Toke Høiland-Jørgensen
Cc: Wen Gong, ath10k, Johannes Berg, linux-wireless
On 26 July 2018 at 13:45, Toke H=C3=B8iland-J=C3=B8rgensen <toke@toke.dk> w=
rote:
> Wen Gong <wgong@codeaurora.org> writes:
>
>> Upstream kernel has an interface to help adjust sk_pacing_shift to help
>> improve TCP UL throughput.
>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>
> When I tested this, a pacing shift of 8 was quite close to optimal as
> well for ath10k. Why are you getting different results?
>
>> Tested with QCA6174 PCI with firmware
>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
>> It's not a regression with new firmware releases.
>>
>> There have 2 test result of different settings:
>>
>> ARM CPU based device with QCA6174A PCI with different
>> sk_pacing_shift:
Different firmware releases have different tx buffering
characteristics. In some 10.2 firmware running on QCA9888 you can have
up to 5ms of delayed aggregation. Ideally sk_pacing_shift should be
adjusted per firmware release. Maybe this should become part of the
ath10k firmware wrapping "fw features" stuff?
Micha=C5=82
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-26 13:02 ` Michał Kazior
0 siblings, 0 replies; 20+ messages in thread
From: Michał Kazior @ 2018-07-26 13:02 UTC (permalink / raw)
To: Toke Høiland-Jørgensen
Cc: Johannes Berg, linux-wireless, ath10k, Wen Gong
On 26 July 2018 at 13:45, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Wen Gong <wgong@codeaurora.org> writes:
>
>> Upstream kernel has an interface to help adjust sk_pacing_shift to help
>> improve TCP UL throughput.
>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>
> When I tested this, a pacing shift of 8 was quite close to optimal as
> well for ath10k. Why are you getting different results?
>
>> Tested with QCA6174 PCI with firmware
>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
>> It's not a regression with new firmware releases.
>>
>> There have 2 test result of different settings:
>>
>> ARM CPU based device with QCA6174A PCI with different
>> sk_pacing_shift:
Different firmware releases have different tx buffering
characteristics. In some 10.2 firmware running on QCA9888 you can have
up to 5ms of delayed aggregation. Ideally sk_pacing_shift should be
adjusted per firmware release. Maybe this should become part of the
ath10k firmware wrapping "fw features" stuff?
Michał
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-26 13:02 ` Michał Kazior
@ 2018-07-27 9:39 ` Wen Gong
-1 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-27 9:39 UTC (permalink / raw)
To: Michał Kazior
Cc: Toke Høiland-Jørgensen, ath10k, Johannes Berg, linux-wireless
On 2018-07-26 21:02, Michał Kazior wrote:
> On 26 July 2018 at 13:45, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>> Wen Gong <wgong@codeaurora.org> writes:
>>
>>> Upstream kernel has an interface to help adjust sk_pacing_shift to
>>> help
>>> improve TCP UL throughput.
>>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>>
>> When I tested this, a pacing shift of 8 was quite close to optimal as
>> well for ath10k. Why are you getting different results?
>>
>>> Tested with QCA6174 PCI with firmware
>>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377
>>> PCI.
>>> It's not a regression with new firmware releases.
>>>
>>> There have 2 test result of different settings:
>>>
>>> ARM CPU based device with QCA6174A PCI with different
>>> sk_pacing_shift:
>
> Different firmware releases have different tx buffering
> characteristics. In some 10.2 firmware running on QCA9888 you can have
> up to 5ms of delayed aggregation. Ideally sk_pacing_shift should be
> adjusted per firmware release. Maybe this should become part of the
> ath10k firmware wrapping "fw features" stuff?
>
recently we do not want to do like this since no test data for each
firmware.
>
> Michał
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-27 9:39 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-27 9:39 UTC (permalink / raw)
To: Michał Kazior
Cc: Johannes Berg, Toke Høiland-Jørgensen, linux-wireless, ath10k
On 2018-07-26 21:02, Michał Kazior wrote:
> On 26 July 2018 at 13:45, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>> Wen Gong <wgong@codeaurora.org> writes:
>>
>>> Upstream kernel has an interface to help adjust sk_pacing_shift to
>>> help
>>> improve TCP UL throughput.
>>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>>
>> When I tested this, a pacing shift of 8 was quite close to optimal as
>> well for ath10k. Why are you getting different results?
>>
>>> Tested with QCA6174 PCI with firmware
>>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377
>>> PCI.
>>> It's not a regression with new firmware releases.
>>>
>>> There have 2 test result of different settings:
>>>
>>> ARM CPU based device with QCA6174A PCI with different
>>> sk_pacing_shift:
>
> Different firmware releases have different tx buffering
> characteristics. In some 10.2 firmware running on QCA9888 you can have
> up to 5ms of delayed aggregation. Ideally sk_pacing_shift should be
> adjusted per firmware release. Maybe this should become part of the
> ath10k firmware wrapping "fw features" stuff?
>
recently we do not want to do like this since no test data for each
firmware.
>
> Michał
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-27 9:39 ` Wen Gong
@ 2018-07-27 12:33 ` Michał Kazior
-1 siblings, 0 replies; 20+ messages in thread
From: Michał Kazior @ 2018-07-27 12:33 UTC (permalink / raw)
To: Wen Gong
Cc: Toke Høiland-Jørgensen, ath10k, Johannes Berg, linux-wireless
On 27 July 2018 at 11:39, Wen Gong <wgong@codeaurora.org> wrote:
> On 2018-07-26 21:02, Micha=C5=82 Kazior wrote:
>>
>> On 26 July 2018 at 13:45, Toke H=C3=B8iland-J=C3=B8rgensen <toke@toke.dk=
> wrote:
>>>
>>> Wen Gong <wgong@codeaurora.org> writes:
>>>
>>>> Upstream kernel has an interface to help adjust sk_pacing_shift to hel=
p
>>>> improve TCP UL throughput.
>>>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>>>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>>>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>>>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>>>
>>>
>>> When I tested this, a pacing shift of 8 was quite close to optimal as
>>> well for ath10k. Why are you getting different results?
>>>
>>>> Tested with QCA6174 PCI with firmware
>>>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI=
.
>>>> It's not a regression with new firmware releases.
>>>>
>>>> There have 2 test result of different settings:
>>>>
>>>> ARM CPU based device with QCA6174A PCI with different
>>>> sk_pacing_shift:
>>
>>
>> Different firmware releases have different tx buffering
>> characteristics. In some 10.2 firmware running on QCA9888 you can have
>> up to 5ms of delayed aggregation. Ideally sk_pacing_shift should be
>> adjusted per firmware release. Maybe this should become part of the
>> ath10k firmware wrapping "fw features" stuff?
>>
> recently we do not want to do like this since no test data for each
> firmware.
All the more reason to *not* change the pacing shift from 8 to 6 for
entire ath10k because you have no idea what impact that is going to
have on other chips/firmwares, e.g. QCA4019, QCA9888X, QCA9984.
Micha=C5=82
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-27 12:33 ` Michał Kazior
0 siblings, 0 replies; 20+ messages in thread
From: Michał Kazior @ 2018-07-27 12:33 UTC (permalink / raw)
To: Wen Gong
Cc: Johannes Berg, Toke Høiland-Jørgensen, linux-wireless, ath10k
On 27 July 2018 at 11:39, Wen Gong <wgong@codeaurora.org> wrote:
> On 2018-07-26 21:02, Michał Kazior wrote:
>>
>> On 26 July 2018 at 13:45, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
>>>
>>> Wen Gong <wgong@codeaurora.org> writes:
>>>
>>>> Upstream kernel has an interface to help adjust sk_pacing_shift to help
>>>> improve TCP UL throughput.
>>>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>>>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>>>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>>>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>>>
>>>
>>> When I tested this, a pacing shift of 8 was quite close to optimal as
>>> well for ath10k. Why are you getting different results?
>>>
>>>> Tested with QCA6174 PCI with firmware
>>>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377 PCI.
>>>> It's not a regression with new firmware releases.
>>>>
>>>> There have 2 test result of different settings:
>>>>
>>>> ARM CPU based device with QCA6174A PCI with different
>>>> sk_pacing_shift:
>>
>>
>> Different firmware releases have different tx buffering
>> characteristics. In some 10.2 firmware running on QCA9888 you can have
>> up to 5ms of delayed aggregation. Ideally sk_pacing_shift should be
>> adjusted per firmware release. Maybe this should become part of the
>> ath10k firmware wrapping "fw features" stuff?
>>
> recently we do not want to do like this since no test data for each
> firmware.
All the more reason to *not* change the pacing shift from 8 to 6 for
entire ath10k because you have no idea what impact that is going to
have on other chips/firmwares, e.g. QCA4019, QCA9888X, QCA9984.
Michał
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-26 11:45 ` Toke Høiland-Jørgensen
@ 2018-07-27 9:29 ` Wen Gong
-1 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-27 9:29 UTC (permalink / raw)
To: Toke Høiland-Jørgensen; +Cc: ath10k, johannes, linux-wireless
On 2018-07-26 19:45, Toke Høiland-Jørgensen wrote:
> Wen Gong <wgong@codeaurora.org> writes:
>
>> Upstream kernel has an interface to help adjust sk_pacing_shift to
>> help
>> improve TCP UL throughput.
>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>
> When I tested this, a pacing shift of 8 was quite close to optimal as
> well for ath10k. Why are you getting different results?
the default value is still 8 in the patch:
https://patchwork.kernel.org/patch/10545361/
In my test, pacing shift 6 is better than 8.
The test is for ath10k/11AC WiFi chips.
Test result is show in the commit logs before.
>
>> Tested with QCA6174 PCI with firmware
>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377
>> PCI.
>> It's not a regression with new firmware releases.
>>
>> There have 2 test result of different settings:
>>
>> ARM CPU based device with QCA6174A PCI with different
>> sk_pacing_shift:
>>
>> sk_pacing_shift throughput(Mbps) CPU utilization
>> 6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
>> 7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
>> 8 288 ~90% idle, Focus on CPU1: ~35%idle
>> 9 ~200 ~92% idle, Focus on CPU1: ~50%idle
>
> Your tests do not include latency values; please try running a test
> that
> also measures latency. The tcp_nup test in Flent (https://flent.org)
> will do that, for instance. Also, is this a single TCP flow?
>
It is not a single TCP flow, it is 500Mbps with 5 flows.
below is result show in commit log before:
5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
set to 6:
tcp_limit_output_bytes throughput(Mbps)
default(262144)+1 Stream 336
default(262144)+2 Streams 558
default(262144)+3 Streams 584
default(262144)+4 Streams 602
default(262144)+5 Streams 598
changed(2621440)+1 Stream 598
changed(2621440)+2 Streams 601
> -Toke
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-27 9:29 ` Wen Gong
0 siblings, 0 replies; 20+ messages in thread
From: Wen Gong @ 2018-07-27 9:29 UTC (permalink / raw)
To: Toke Høiland-Jørgensen; +Cc: johannes, linux-wireless, ath10k
On 2018-07-26 19:45, Toke Høiland-Jørgensen wrote:
> Wen Gong <wgong@codeaurora.org> writes:
>
>> Upstream kernel has an interface to help adjust sk_pacing_shift to
>> help
>> improve TCP UL throughput.
>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>
> When I tested this, a pacing shift of 8 was quite close to optimal as
> well for ath10k. Why are you getting different results?
the default value is still 8 in the patch:
https://patchwork.kernel.org/patch/10545361/
In my test, pacing shift 6 is better than 8.
The test is for ath10k/11AC WiFi chips.
Test result is show in the commit logs before.
>
>> Tested with QCA6174 PCI with firmware
>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377
>> PCI.
>> It's not a regression with new firmware releases.
>>
>> There have 2 test result of different settings:
>>
>> ARM CPU based device with QCA6174A PCI with different
>> sk_pacing_shift:
>>
>> sk_pacing_shift throughput(Mbps) CPU utilization
>> 6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
>> 7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
>> 8 288 ~90% idle, Focus on CPU1: ~35%idle
>> 9 ~200 ~92% idle, Focus on CPU1: ~50%idle
>
> Your tests do not include latency values; please try running a test
> that
> also measures latency. The tcp_nup test in Flent (https://flent.org)
> will do that, for instance. Also, is this a single TCP flow?
>
It is not a single TCP flow, it is 500Mbps with 5 flows.
below is result show in commit log before:
5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
set to 6:
tcp_limit_output_bytes throughput(Mbps)
default(262144)+1 Stream 336
default(262144)+2 Streams 558
default(262144)+3 Streams 584
default(262144)+4 Streams 602
default(262144)+5 Streams 598
changed(2621440)+1 Stream 598
changed(2621440)+2 Streams 601
> -Toke
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
2018-07-27 9:29 ` Wen Gong
@ 2018-07-27 20:06 ` Toke Høiland-Jørgensen
-1 siblings, 0 replies; 20+ messages in thread
From: Toke Høiland-Jørgensen @ 2018-07-27 20:06 UTC (permalink / raw)
To: Wen Gong; +Cc: ath10k, johannes, linux-wireless
Wen Gong <wgong@codeaurora.org> writes:
> On 2018-07-26 19:45, Toke H=C3=B8iland-J=C3=B8rgensen wrote:
>> Wen Gong <wgong@codeaurora.org> writes:
>>=20
>>> Upstream kernel has an interface to help adjust sk_pacing_shift to=20
>>> help
>>> improve TCP UL throughput.
>>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>>=20
>> When I tested this, a pacing shift of 8 was quite close to optimal as
>> well for ath10k. Why are you getting different results?
>
> the default value is still 8 in the patch:
> https://patchwork.kernel.org/patch/10545361/
>
> In my test, pacing shift 6 is better than 8.
> The test is for ath10k/11AC WiFi chips.
> Test result is show in the commit logs before.
>>=20
>>> Tested with QCA6174 PCI with firmware
>>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377=20
>>> PCI.
>>> It's not a regression with new firmware releases.
>>>=20
>>> There have 2 test result of different settings:
>>>=20
>>> ARM CPU based device with QCA6174A PCI with different
>>> sk_pacing_shift:
>>>=20
>>> sk_pacing_shift throughput(Mbps) CPU utilization
>>> 6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
>>> 7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
>>> 8 288 ~90% idle, Focus on CPU1: ~35%idle
>>> 9 ~200 ~92% idle, Focus on CPU1: ~50%idle
>>=20
>> Your tests do not include latency values; please try running a test=20
>> that
>> also measures latency. The tcp_nup test in Flent (https://flent.org)
>> will do that, for instance. Also, is this a single TCP flow?
>>=20
>
> It is not a single TCP flow, it is 500Mbps with 5 flows.
>
> below is result show in commit log before:
> 5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
> set to 6:
>
> tcp_limit_output_bytes throughput(Mbps)
> default(262144)+1 Stream 336
> default(262144)+2 Streams 558
> default(262144)+3 Streams 584
> default(262144)+4 Streams 602
> default(262144)+5 Streams 598
> changed(2621440)+1 Stream 598
> changed(2621440)+2 Streams 601
This is useless without latency numbers. The whole point of
sk_pacing_shift is to control the tradeoff between latency and
throughput. You're only showing the throughput, so it's impossible to
judge if setting the pacing shift to 6 is right (and from your results I
suspect the sweet spot is actually 7).
-Toke
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips
@ 2018-07-27 20:06 ` Toke Høiland-Jørgensen
0 siblings, 0 replies; 20+ messages in thread
From: Toke Høiland-Jørgensen @ 2018-07-27 20:06 UTC (permalink / raw)
To: Wen Gong; +Cc: johannes, linux-wireless, ath10k
Wen Gong <wgong@codeaurora.org> writes:
> On 2018-07-26 19:45, Toke Høiland-Jørgensen wrote:
>> Wen Gong <wgong@codeaurora.org> writes:
>>
>>> Upstream kernel has an interface to help adjust sk_pacing_shift to
>>> help
>>> improve TCP UL throughput.
>>> The sk_pacing_shift is 8 in mac80211, this is based on test with 11N
>>> WiFi chips with ath9k. For QCA6174/QCA9377 PCI 11AC chips, the 11AC
>>> VHT80 TCP UL throughput testing result shows 6 is the optimal.
>>> Overwrite the sk_pacing_shift to 6 in ath10k driver.
>>
>> When I tested this, a pacing shift of 8 was quite close to optimal as
>> well for ath10k. Why are you getting different results?
>
> the default value is still 8 in the patch:
> https://patchwork.kernel.org/patch/10545361/
>
> In my test, pacing shift 6 is better than 8.
> The test is for ath10k/11AC WiFi chips.
> Test result is show in the commit logs before.
>>
>>> Tested with QCA6174 PCI with firmware
>>> WLAN.RM.4.4.1-00109-QCARMSWPZ-1, but this will also affect QCA9377
>>> PCI.
>>> It's not a regression with new firmware releases.
>>>
>>> There have 2 test result of different settings:
>>>
>>> ARM CPU based device with QCA6174A PCI with different
>>> sk_pacing_shift:
>>>
>>> sk_pacing_shift throughput(Mbps) CPU utilization
>>> 6 500(-P5) ~75% idle, Focus on CPU1: ~14%idle
>>> 7 454(-P5) ~80% idle, Focus on CPU1: ~4%idle
>>> 8 288 ~90% idle, Focus on CPU1: ~35%idle
>>> 9 ~200 ~92% idle, Focus on CPU1: ~50%idle
>>
>> Your tests do not include latency values; please try running a test
>> that
>> also measures latency. The tcp_nup test in Flent (https://flent.org)
>> will do that, for instance. Also, is this a single TCP flow?
>>
>
> It is not a single TCP flow, it is 500Mbps with 5 flows.
>
> below is result show in commit log before:
> 5G TCP UL VTH80 on X86 platform with QCA6174A PCI with sk_packing_shift
> set to 6:
>
> tcp_limit_output_bytes throughput(Mbps)
> default(262144)+1 Stream 336
> default(262144)+2 Streams 558
> default(262144)+3 Streams 584
> default(262144)+4 Streams 602
> default(262144)+5 Streams 598
> changed(2621440)+1 Stream 598
> changed(2621440)+2 Streams 601
This is useless without latency numbers. The whole point of
sk_pacing_shift is to control the tradeoff between latency and
throughput. You're only showing the throughput, so it's impossible to
judge if setting the pacing shift to 6 is right (and from your results I
suspect the sweet spot is actually 7).
-Toke
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 20+ messages in thread
end of thread, other threads:[~2018-07-27 21:29 UTC | newest]
Thread overview: 20+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-26 7:16 [PATCH 0/2] Change sk_pacing_shift in ieee80211_hw for best tx throughput Wen Gong
2018-07-26 7:16 ` Wen Gong
2018-07-26 7:16 ` [PATCH 1/2] mac80211: Save sk_pacing_shift to ieee80211_hw Wen Gong
2018-07-26 7:16 ` Wen Gong
2018-07-26 7:16 ` [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips Wen Gong
2018-07-26 7:16 ` Wen Gong
2018-07-26 7:21 [PATCH 0/2] Change sk_pacing_shift in ieee80211_hw for best tx throughput Wen Gong
2018-07-26 7:21 ` [PATCH 2/2] ath10k: Set sk_pacing_shift to 6 for 11AC WiFi chips Wen Gong
2018-07-26 7:21 ` Wen Gong
2018-07-26 11:45 ` Toke Høiland-Jørgensen
2018-07-26 11:45 ` Toke Høiland-Jørgensen
2018-07-26 13:02 ` Michał Kazior
2018-07-26 13:02 ` Michał Kazior
2018-07-27 9:39 ` Wen Gong
2018-07-27 9:39 ` Wen Gong
2018-07-27 12:33 ` Michał Kazior
2018-07-27 12:33 ` Michał Kazior
2018-07-27 9:29 ` Wen Gong
2018-07-27 9:29 ` Wen Gong
2018-07-27 20:06 ` Toke Høiland-Jørgensen
2018-07-27 20:06 ` Toke Høiland-Jørgensen
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.