[dpdk-dev] [PATCH v3 18/19] net/hinic: optimize RX performance

[Wangxiaoyun (Cloud, Network Chip Application Development Dept)]

Hi Ferruh,
      Thanks for your comments. I think you're right, i will modify it with Patch V4. Also I change it with the same structure
for X86-64 and ARM platform with cache-aligned, and test the RX performance, all is OK.

Best Regards
Xiaoyun Wang

在 2019/9/30 23:10, Ferruh Yigit 写道:
> On 9/30/2019 3:00 PM, Xiaoyun wang wrote:
>> This patch optimizes receive packets performance
>> on arm platform.
>>
>> Signed-off-by: Xiaoyun wang <cloud.wangxiaoyun at huawei.com>
> <...>
>
>> @@ -28,6 +28,7 @@ struct hinic_rq_ctrl {
>>   	u32	ctrl_fmt;
>>   };
>>   
>> +#if defined(__X86_64_SSE__)
>>   struct hinic_rq_cqe {
>>   	u32 status;
>>   	u32 vlan_len;
>> @@ -36,6 +37,16 @@ struct hinic_rq_cqe {
>>   
>>   	u32 rsvd[4];
>>   };
>> +#elif defined(__ARM64_NEON__)
>> +struct hinic_rq_cqe {
>> +	u32 status;
>> +	u32 vlan_len;
>> +	u32 offload_type;
>> +	u32 rss_hash;
>> +
>> +	u32 rsvd[4];
>> +} __rte_cache_aligned;
>> +#endif
> This change makes "struct hinic_rq_cqe" only exist for 'x86_64' and 'ARM64', if
> the intention is to add '__rte_cache_aligned' for the neon, would following
> work, which will also work for all archs:
>
>
>   struct hinic_rq_cqe {
>   	u32 status;
>   	u32 vlan_len;
>   	u32 offload_type;
>   	u32 rss_hash;
>
>   	u32 rsvd[4];
>   #if defined(__ARM64_NEON__)
>   } __rte_cache_aligned;
>   #else
>   };
>   #endif
>
> If this works, can you please send a new version with this update?
>