[v3,10/18] net/ice: add checks for max SIMD bitwidth

Message ID 20200930130415.11211-11-ciara.power@intel.com (mailing list archive)
State Superseded, archived
Delegated to: David Marchand
Headers
Series add max SIMD bitwidth to EAL |

Checks

Context Check Description
ci/checkpatch success coding style OK

Commit Message

Power, Ciara Sept. 30, 2020, 1:04 p.m. UTC
  When choosing a vector path to take, an extra condition must be
satisfied to ensure the max SIMD bitwidth allows for the CPU enabled
path.

Cc: Qiming Yang <qiming.yang@intel.com>
Cc: Qi Zhang <qi.z.zhang@intel.com>

Signed-off-by: Ciara Power <ciara.power@intel.com>
---
 drivers/net/ice/ice_rxtx.c | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)
  

Comments

Qi Zhang Oct. 9, 2020, 12:04 a.m. UTC | #1
> -----Original Message-----
> From: Power, Ciara <ciara.power@intel.com>
> Sent: Wednesday, September 30, 2020 9:04 PM
> To: dev@dpdk.org
> Cc: Power, Ciara <ciara.power@intel.com>; Yang, Qiming
> <qiming.yang@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> Subject: [PATCH v3 10/18] net/ice: add checks for max SIMD bitwidth
> 
> When choosing a vector path to take, an extra condition must be satisfied to
> ensure the max SIMD bitwidth allows for the CPU enabled path.
> 
> Cc: Qiming Yang <qiming.yang@intel.com>
> Cc: Qi Zhang <qi.z.zhang@intel.com>
> 
> Signed-off-by: Ciara Power <ciara.power@intel.com>

Acked-by: Qi Zhang <qi.z.zhang@intel.com>

> ---
>  drivers/net/ice/ice_rxtx.c | 20 ++++++++++++++------
>  1 file changed, 14 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c index
> fef6ad4544..5a29af743c 100644
> --- a/drivers/net/ice/ice_rxtx.c
> +++ b/drivers/net/ice/ice_rxtx.c
> @@ -2936,7 +2936,9 @@ ice_set_rx_function(struct rte_eth_dev *dev)
>  	bool use_avx2 = false;
> 
>  	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
> -		if (!ice_rx_vec_dev_check(dev) && ad->rx_bulk_alloc_allowed) {
> +		if (!ice_rx_vec_dev_check(dev) && ad->rx_bulk_alloc_allowed &&
> +				rte_get_max_simd_bitwidth()
> +				>= RTE_MAX_128_SIMD) {
>  			ad->rx_vec_allowed = true;
>  			for (i = 0; i < dev->data->nb_rx_queues; i++) {
>  				rxq = dev->data->rx_queues[i];
> @@ -2946,8 +2948,10 @@ ice_set_rx_function(struct rte_eth_dev *dev)
>  				}
>  			}
> 
> -			if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> -			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> +			if ((rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> +			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1) &&
> +					rte_get_max_simd_bitwidth()
> +					>= RTE_MAX_256_SIMD)
>  				use_avx2 = true;
> 
>  		} else {
> @@ -3114,7 +3118,9 @@ ice_set_tx_function(struct rte_eth_dev *dev)
>  	bool use_avx2 = false;
> 
>  	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
> -		if (!ice_tx_vec_dev_check(dev)) {
> +		if (!ice_tx_vec_dev_check(dev) &&
> +				rte_get_max_simd_bitwidth()
> +				>= RTE_MAX_128_SIMD) {
>  			ad->tx_vec_allowed = true;
>  			for (i = 0; i < dev->data->nb_tx_queues; i++) {
>  				txq = dev->data->tx_queues[i];
> @@ -3124,8 +3130,10 @@ ice_set_tx_function(struct rte_eth_dev *dev)
>  				}
>  			}
> 
> -			if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> -			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> +			if ((rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> +			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1) &&
> +					rte_get_max_simd_bitwidth()
> +					>= RTE_MAX_256_SIMD)
>  				use_avx2 = true;
> 
>  		} else {
> --
> 2.17.1
  
Qi Zhang Oct. 9, 2020, 1:05 a.m. UTC | #2
> -----Original Message-----
> From: Zhang, Qi Z
> Sent: Friday, October 9, 2020 8:05 AM
> To: Power, Ciara <ciara.power@intel.com>; dev@dpdk.org
> Cc: Yang, Qiming <qiming.yang@intel.com>
> Subject: RE: [PATCH v3 10/18] net/ice: add checks for max SIMD bitwidth
> 
> 
> 
> > -----Original Message-----
> > From: Power, Ciara <ciara.power@intel.com>
> > Sent: Wednesday, September 30, 2020 9:04 PM
> > To: dev@dpdk.org
> > Cc: Power, Ciara <ciara.power@intel.com>; Yang, Qiming
> > <qiming.yang@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>
> > Subject: [PATCH v3 10/18] net/ice: add checks for max SIMD bitwidth
> >
> > When choosing a vector path to take, an extra condition must be
> > satisfied to ensure the max SIMD bitwidth allows for the CPU enabled path.
> >
> > Cc: Qiming Yang <qiming.yang@intel.com>
> > Cc: Qi Zhang <qi.z.zhang@intel.com>
> >
> > Signed-off-by: Ciara Power <ciara.power@intel.com>
> 
> Acked-by: Qi Zhang <qi.z.zhang@intel.com>

I'd like to withdraw my ack, due to more consideration as below

> 
> > ---
> >  drivers/net/ice/ice_rxtx.c | 20 ++++++++++++++------
> >  1 file changed, 14 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
> > index fef6ad4544..5a29af743c 100644
> > --- a/drivers/net/ice/ice_rxtx.c
> > +++ b/drivers/net/ice/ice_rxtx.c
> > @@ -2936,7 +2936,9 @@ ice_set_rx_function(struct rte_eth_dev *dev)
> >  	bool use_avx2 = false;
> >
> >  	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
> > -		if (!ice_rx_vec_dev_check(dev) && ad->rx_bulk_alloc_allowed) {
> > +		if (!ice_rx_vec_dev_check(dev) && ad->rx_bulk_alloc_allowed &&
> > +				rte_get_max_simd_bitwidth()
> > +				>= RTE_MAX_128_SIMD) {
> >  			ad->rx_vec_allowed = true;
> >  			for (i = 0; i < dev->data->nb_rx_queues; i++) {
> >  				rxq = dev->data->rx_queues[i];
> > @@ -2946,8 +2948,10 @@ ice_set_rx_function(struct rte_eth_dev *dev)
> >  				}
> >  			}
> >
> > -			if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> > -			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> > +			if ((rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> > +			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1) &&
> > +					rte_get_max_simd_bitwidth()
> > +					>= RTE_MAX_256_SIMD)

As we have this max SIMD hint, the avx512 flag check is not necessary, 
and I think for old platform which not support avx512, the default RTE_DEFAULT_SIMD_BITWIDTH should be configured to 128, so AVX2 will not be over used.

> >  				use_avx2 = true;
> >
> >  		} else {
> > @@ -3114,7 +3118,9 @@ ice_set_tx_function(struct rte_eth_dev *dev)
> >  	bool use_avx2 = false;
> >
> >  	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
> > -		if (!ice_tx_vec_dev_check(dev)) {
> > +		if (!ice_tx_vec_dev_check(dev) &&
> > +				rte_get_max_simd_bitwidth()
> > +				>= RTE_MAX_128_SIMD) {
> >  			ad->tx_vec_allowed = true;
> >  			for (i = 0; i < dev->data->nb_tx_queues; i++) {
> >  				txq = dev->data->tx_queues[i];
> > @@ -3124,8 +3130,10 @@ ice_set_tx_function(struct rte_eth_dev *dev)
> >  				}
> >  			}
> >
> > -			if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> > -			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> > +			if ((rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> > +			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1) &&
> > +					rte_get_max_simd_bitwidth()
> > +					>= RTE_MAX_256_SIMD)
> >  				use_avx2 = true;
> >
> >  		} else {
> > --
> > 2.17.1
  

Patch

diff --git a/drivers/net/ice/ice_rxtx.c b/drivers/net/ice/ice_rxtx.c
index fef6ad4544..5a29af743c 100644
--- a/drivers/net/ice/ice_rxtx.c
+++ b/drivers/net/ice/ice_rxtx.c
@@ -2936,7 +2936,9 @@  ice_set_rx_function(struct rte_eth_dev *dev)
 	bool use_avx2 = false;
 
 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
-		if (!ice_rx_vec_dev_check(dev) && ad->rx_bulk_alloc_allowed) {
+		if (!ice_rx_vec_dev_check(dev) && ad->rx_bulk_alloc_allowed &&
+				rte_get_max_simd_bitwidth()
+				>= RTE_MAX_128_SIMD) {
 			ad->rx_vec_allowed = true;
 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
 				rxq = dev->data->rx_queues[i];
@@ -2946,8 +2948,10 @@  ice_set_rx_function(struct rte_eth_dev *dev)
 				}
 			}
 
-			if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
-			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			if ((rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1) &&
+					rte_get_max_simd_bitwidth()
+					>= RTE_MAX_256_SIMD)
 				use_avx2 = true;
 
 		} else {
@@ -3114,7 +3118,9 @@  ice_set_tx_function(struct rte_eth_dev *dev)
 	bool use_avx2 = false;
 
 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
-		if (!ice_tx_vec_dev_check(dev)) {
+		if (!ice_tx_vec_dev_check(dev) &&
+				rte_get_max_simd_bitwidth()
+				>= RTE_MAX_128_SIMD) {
 			ad->tx_vec_allowed = true;
 			for (i = 0; i < dev->data->nb_tx_queues; i++) {
 				txq = dev->data->tx_queues[i];
@@ -3124,8 +3130,10 @@  ice_set_tx_function(struct rte_eth_dev *dev)
 				}
 			}
 
-			if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
-			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			if ((rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+			rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1) &&
+					rte_get_max_simd_bitwidth()
+					>= RTE_MAX_256_SIMD)
 				use_avx2 = true;
 
 		} else {