* [PATCH] net/gve: Refill all descriptors that were consumed
@ 2023-09-14 20:52 Rushil Gupta
2023-09-15 2:28 ` Guo, Junfeng
0 siblings, 1 reply; 3+ messages in thread
From: Rushil Gupta @ 2023-09-14 20:52 UTC (permalink / raw)
To: qi.z.zhang, ferruh.yigit
Cc: junfeng.guo, dev, Rushil Gupta, Joshua Washington
Refill all consumed descriptors instead of refilling a set amount of
descriptors everytime. Also, lower the rx-free-thresh to refill more
aggressively. This yields in fewer packet drops for DQO queue format.
Signed-off-by: Rushil Gupta <rushilg@google.com>
Reviewed-by: Joshua Washington <joshwash@google.com>
---
drivers/net/gve/gve_ethdev.h | 2 +-
drivers/net/gve/gve_rx_dqo.c | 4 ++--
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/net/gve/gve_ethdev.h b/drivers/net/gve/gve_ethdev.h
index c9bcfa553c..fe7095ed76 100644
--- a/drivers/net/gve/gve_ethdev.h
+++ b/drivers/net/gve/gve_ethdev.h
@@ -27,7 +27,7 @@
#define PCI_MSIX_FLAGS 2 /* Message Control */
#define PCI_MSIX_FLAGS_QSIZE 0x07FF /* Table size */
-#define GVE_DEFAULT_RX_FREE_THRESH 512
+#define GVE_DEFAULT_RX_FREE_THRESH 64
#define GVE_DEFAULT_TX_FREE_THRESH 32
#define GVE_DEFAULT_TX_RS_THRESH 32
#define GVE_TX_MAX_FREE_SZ 512
diff --git a/drivers/net/gve/gve_rx_dqo.c b/drivers/net/gve/gve_rx_dqo.c
index 236aefd2a8..7e7ddac48e 100644
--- a/drivers/net/gve/gve_rx_dqo.c
+++ b/drivers/net/gve/gve_rx_dqo.c
@@ -12,8 +12,8 @@ gve_rx_refill_dqo(struct gve_rx_queue *rxq)
{
volatile struct gve_rx_desc_dqo *rx_buf_ring;
volatile struct gve_rx_desc_dqo *rx_buf_desc;
- struct rte_mbuf *nmb[rxq->free_thresh];
- uint16_t nb_refill = rxq->free_thresh;
+ struct rte_mbuf *nmb[rxq->nb_rx_hold];
+ uint16_t nb_refill = rxq->nb_rx_hold;
uint16_t nb_desc = rxq->nb_rx_desc;
uint16_t next_avail = rxq->bufq_tail;
struct rte_eth_dev *dev;
--
2.42.0.459.ge4e396fd5e-goog
^ permalink raw reply [flat|nested] 3+ messages in thread* RE: [PATCH] net/gve: Refill all descriptors that were consumed
2023-09-14 20:52 [PATCH] net/gve: Refill all descriptors that were consumed Rushil Gupta
@ 2023-09-15 2:28 ` Guo, Junfeng
2023-09-21 16:37 ` Ferruh Yigit
0 siblings, 1 reply; 3+ messages in thread
From: Guo, Junfeng @ 2023-09-15 2:28 UTC (permalink / raw)
To: Rushil Gupta, Zhang, Qi Z, ferruh.yigit; +Cc: dev, Joshua Washington
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
Regards,
Junfeng
> -----Original Message-----
> From: Rushil Gupta <rushilg@google.com>
> Sent: Friday, September 15, 2023 04:52
> To: Zhang, Qi Z <qi.z.zhang@intel.com>; ferruh.yigit@amd.com
> Cc: Guo, Junfeng <junfeng.guo@intel.com>; dev@dpdk.org; Rushil Gupta
> <rushilg@google.com>; Joshua Washington <joshwash@google.com>
> Subject: [PATCH] net/gve: Refill all descriptors that were consumed
>
> Refill all consumed descriptors instead of refilling a set amount of
> descriptors everytime. Also, lower the rx-free-thresh to refill more
> aggressively. This yields in fewer packet drops for DQO queue format.
>
> Signed-off-by: Rushil Gupta <rushilg@google.com>
> Reviewed-by: Joshua Washington <joshwash@google.com>
> ---
> drivers/net/gve/gve_ethdev.h | 2 +-
> drivers/net/gve/gve_rx_dqo.c | 4 ++--
> 2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/gve/gve_ethdev.h b/drivers/net/gve/gve_ethdev.h
> index c9bcfa553c..fe7095ed76 100644
> --- a/drivers/net/gve/gve_ethdev.h
> +++ b/drivers/net/gve/gve_ethdev.h
> @@ -27,7 +27,7 @@
> #define PCI_MSIX_FLAGS 2 /* Message Control */
> #define PCI_MSIX_FLAGS_QSIZE 0x07FF /* Table size */
>
> -#define GVE_DEFAULT_RX_FREE_THRESH 512
> +#define GVE_DEFAULT_RX_FREE_THRESH 64
> #define GVE_DEFAULT_TX_FREE_THRESH 32
> #define GVE_DEFAULT_TX_RS_THRESH 32
> #define GVE_TX_MAX_FREE_SZ 512
> diff --git a/drivers/net/gve/gve_rx_dqo.c b/drivers/net/gve/gve_rx_dqo.c
> index 236aefd2a8..7e7ddac48e 100644
> --- a/drivers/net/gve/gve_rx_dqo.c
> +++ b/drivers/net/gve/gve_rx_dqo.c
> @@ -12,8 +12,8 @@ gve_rx_refill_dqo(struct gve_rx_queue *rxq)
> {
> volatile struct gve_rx_desc_dqo *rx_buf_ring;
> volatile struct gve_rx_desc_dqo *rx_buf_desc;
> - struct rte_mbuf *nmb[rxq->free_thresh];
> - uint16_t nb_refill = rxq->free_thresh;
> + struct rte_mbuf *nmb[rxq->nb_rx_hold];
> + uint16_t nb_refill = rxq->nb_rx_hold;
> uint16_t nb_desc = rxq->nb_rx_desc;
> uint16_t next_avail = rxq->bufq_tail;
> struct rte_eth_dev *dev;
> --
> 2.42.0.459.ge4e396fd5e-goog
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] net/gve: Refill all descriptors that were consumed
2023-09-15 2:28 ` Guo, Junfeng
@ 2023-09-21 16:37 ` Ferruh Yigit
0 siblings, 0 replies; 3+ messages in thread
From: Ferruh Yigit @ 2023-09-21 16:37 UTC (permalink / raw)
To: Guo, Junfeng, Rushil Gupta, Zhang, Qi Z; +Cc: dev, Joshua Washington
On 9/15/2023 3:28 AM, Guo, Junfeng wrote:
>> -----Original Message-----
>> From: Rushil Gupta <rushilg@google.com>
>> Sent: Friday, September 15, 2023 04:52
>> To: Zhang, Qi Z <qi.z.zhang@intel.com>; ferruh.yigit@amd.com
>> Cc: Guo, Junfeng <junfeng.guo@intel.com>; dev@dpdk.org; Rushil Gupta
>> <rushilg@google.com>; Joshua Washington <joshwash@google.com>
>> Subject: [PATCH] net/gve: Refill all descriptors that were consumed
>>
>> Refill all consumed descriptors instead of refilling a set amount of
>> descriptors everytime. Also, lower the rx-free-thresh to refill more
>> aggressively. This yields in fewer packet drops for DQO queue format.
>>
>> Signed-off-by: Rushil Gupta <rushilg@google.com>
>> Reviewed-by: Joshua Washington <joshwash@google.com>
>
> Acked-by: Junfeng Guo <junfeng.guo@intel.com>
>
Applied to dpdk-next-net/main, thanks.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2023-09-21 16:37 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-09-14 20:52 [PATCH] net/gve: Refill all descriptors that were consumed Rushil Gupta
2023-09-15 2:28 ` Guo, Junfeng
2023-09-21 16:37 ` Ferruh Yigit
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).