2020-04-09 08:45:50

by Manikanta Pubbisetty

[permalink] [raw]
Subject: [PATCH] ath11k: rx path optimizations

During RX, accessing the reo dest ring descriptor directly is consuming
a lot of CPU cycles. Accessing the descriptor after copying it locally
has improved CPU usage by around ~10-15% while measuring throughput
in RX DBTC test cases(all radios are involved in the throughput
measurement).

HW tested: IPQ8074

Signed-off-by: Manikanta Pubbisetty <[email protected]>
---
drivers/net/wireless/ath/ath11k/dp_rx.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
index a3f2c76..203fd44 100644
--- a/drivers/net/wireless/ath/ath11k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
@@ -2403,12 +2403,12 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,

try_again:
while ((rx_desc = ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
- struct hal_reo_dest_ring *desc = (struct hal_reo_dest_ring *)rx_desc;
+ struct hal_reo_dest_ring desc = *(struct hal_reo_dest_ring *)rx_desc;
enum hal_reo_dest_ring_push_reason push_reason;
u32 cookie;

cookie = FIELD_GET(BUFFER_ADDR_INFO1_SW_COOKIE,
- desc->buf_addr_info.info1);
+ desc.buf_addr_info.info1);
buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
cookie);
mac_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_PDEV_ID, cookie);
@@ -2436,7 +2436,7 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
total_msdu_reaped++;

push_reason = FIELD_GET(HAL_REO_DEST_RING_INFO0_PUSH_REASON,
- desc->info0);
+ desc.info0);
if (push_reason !=
HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION) {
dev_kfree_skb_any(msdu);
@@ -2444,15 +2444,15 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
continue;
}

- rxcb->is_first_msdu = !!(desc->rx_msdu_info.info0 &
+ rxcb->is_first_msdu = !!(desc.rx_msdu_info.info0 &
RX_MSDU_DESC_INFO0_FIRST_MSDU_IN_MPDU);
- rxcb->is_last_msdu = !!(desc->rx_msdu_info.info0 &
+ rxcb->is_last_msdu = !!(desc.rx_msdu_info.info0 &
RX_MSDU_DESC_INFO0_LAST_MSDU_IN_MPDU);
- rxcb->is_continuation = !!(desc->rx_msdu_info.info0 &
+ rxcb->is_continuation = !!(desc.rx_msdu_info.info0 &
RX_MSDU_DESC_INFO0_MSDU_CONTINUATION);
rxcb->mac_id = mac_id;
rxcb->tid = FIELD_GET(HAL_REO_DEST_RING_INFO0_RX_QUEUE_NUM,
- desc->info0);
+ desc.info0);

__skb_queue_tail(&msdu_list, msdu);

--
2.7.4


2020-04-15 17:38:37

by Kalle Valo

[permalink] [raw]
Subject: Re: [PATCH] ath11k: rx path optimizations

Manikanta Pubbisetty <[email protected]> wrote:

> During RX, accessing the reo dest ring descriptor directly is consuming
> a lot of CPU cycles. Accessing the descriptor after copying it locally
> has improved CPU usage by around ~10-15% while measuring throughput
> in RX DBTC test cases(all radios are involved in the throughput
> measurement).
>
> HW tested: IPQ8074
>
> Signed-off-by: Manikanta Pubbisetty <[email protected]>
> Signed-off-by: Kalle Valo <[email protected]>

Patch applied to ath-next branch of ath.git, thanks.

7395fb496577 ath11k: rx path optimizations

--
https://patchwork.kernel.org/patch/11481425/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches