[dpdk-dev] [PATCH v3 04/18] net/i40e: add checks for max SIMD bitwidth

[Power, Ciara]

Hi Konstantin,

 
>-----Original Message-----
>From: Ananyev, Konstantin <konstantin.ananyev at intel.com>
>Sent: Thursday 8 October 2020 16:22
>To: Power, Ciara <ciara.power at intel.com>; dev at dpdk.org
>Cc: Power, Ciara <ciara.power at intel.com>; Xing, Beilei
><beilei.xing at intel.com>; Guo, Jia <jia.guo at intel.com>
>Subject: RE: [dpdk-dev] [PATCH v3 04/18] net/i40e: 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: Beilei Xing <beilei.xing at intel.com>
>> Cc: Jeff Guo <jia.guo at intel.com>
>>
>> Signed-off-by: Ciara Power <ciara.power at intel.com>
>> ---
>>  drivers/net/i40e/i40e_rxtx.c | 19 +++++++++++++------
>>  1 file changed, 13 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/net/i40e/i40e_rxtx.c
>> b/drivers/net/i40e/i40e_rxtx.c index 60b33d20a1..9b535b52fa 100644
>> --- a/drivers/net/i40e/i40e_rxtx.c
>> +++ b/drivers/net/i40e/i40e_rxtx.c
>> @@ -3098,7 +3098,8 @@ static eth_rx_burst_t
>> i40e_get_latest_rx_vec(bool scatter)  {  #if defined(RTE_ARCH_X86) &&
>> defined(CC_AVX2_SUPPORT)
>> -	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2))
>> +	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) &&
>> +			rte_get_max_simd_bitwidth() >=
>RTE_MAX_256_SIMD)
>>  		return scatter ? i40e_recv_scattered_pkts_vec_avx2 :
>>  				 i40e_recv_pkts_vec_avx2;
>
>Hmm, but that means - if user will set --simd-bitwidth=128 we'll select
>scalar function, right?
>Even though sse one is available.
>Is that what we really want in that case?
>

If the max SIMD is 128, the second return in this function is used, which I believe is SSE:

	return scatter ? i40e_recv_scattered_pkts_vec :
			 i40e_recv_pkts_vec;

And that function is only called if the max SIMD is >=128, scalar is used otherwise.

Am I missing something else here?

Thanks,
Ciara 

>>  #endif
>> @@ -3115,7 +3116,8 @@ i40e_get_recommend_rx_vec(bool scatter)
>>  	 * use of AVX2 version to later plaforms, not all those that could
>>  	 * theoretically run it.
>>  	 */
>> -	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F))
>> +	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) &&
>> +			rte_get_max_simd_bitwidth() >=
>RTE_MAX_256_SIMD)
>>  		return scatter ? i40e_recv_scattered_pkts_vec_avx2 :
>>  				 i40e_recv_pkts_vec_avx2;
>>  #endif
>> @@ -3154,7 +3156,8 @@ i40e_set_rx_function(struct rte_eth_dev *dev)
>>  		}
>>  	}
>>
>> -	if (ad->rx_vec_allowed) {
>> +	if (ad->rx_vec_allowed  && rte_get_max_simd_bitwidth()
>> +			>= RTE_MAX_128_SIMD) {
>>  		/* Vec Rx path */
>>  		PMD_INIT_LOG(DEBUG, "Vector Rx path will be used on
>port=%d.",
>>  				dev->data->port_id);
>> @@ -3268,7 +3271,8 @@ static eth_tx_burst_t
>>  i40e_get_latest_tx_vec(void)
>>  {
>>  #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
>> -	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2))
>> +	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) &&
>> +			rte_get_max_simd_bitwidth() >=
>RTE_MAX_256_SIMD)
>>  		return i40e_xmit_pkts_vec_avx2;
>>  #endif
>>  	return i40e_xmit_pkts_vec;
>> @@ -3283,7 +3287,8 @@ i40e_get_recommend_tx_vec(void)
>>  	 * use of AVX2 version to later plaforms, not all those that could
>>  	 * theoretically run it.
>>  	 */
>> -	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F))
>> +	if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) &&
>> +			rte_get_max_simd_bitwidth() >=
>RTE_MAX_256_SIMD)
>>  		return i40e_xmit_pkts_vec_avx2;
>>  #endif
>>  	return i40e_xmit_pkts_vec;
>> @@ -3311,7 +3316,9 @@ i40e_set_tx_function(struct rte_eth_dev *dev)
>>  	}
>>
>>  	if (ad->tx_simple_allowed) {
>> -		if (ad->tx_vec_allowed) {
>> +		if (ad->tx_vec_allowed &&
>> +				rte_get_max_simd_bitwidth()
>> +				>= RTE_MAX_128_SIMD) {
>>  			PMD_INIT_LOG(DEBUG, "Vector tx finally be used.");
>>  			if (ad->use_latest_vec)
>>  				dev->tx_pkt_burst =
>> --
>> 2.17.1