在 2023/9/25 22:21, Ferruh Yigit 写道:
On 7/11/2023 11:24 AM, Dongdong Liu wrote:
From: Huisong Li <lihuis...@huawei.com>
Currently, hns3 SVE Tx checks the valid bits of all descriptors
in a batch and then determines whether to release the corresponding
mbufs. Actually, once the valid bit of any descriptor in a batch
isn't cleared, driver does not need to scan the rest of descriptors.
If we optimize SVE codes algorithm about this function, the performance
of a single queue for 64B packet is improved by ~2% on txonly forwarding
mode. And if use C code to scan all descriptors, the performance is
improved by ~8%.
So this patch selects C code to optimize this code to improve the SVE
Tx performance.
Signed-off-by: Huisong Li <lihuis...@huawei.com>
Signed-off-by: Dongdong Liu <liudongdo...@huawei.com>
SVE Tx optimized by removing SVE implementation :)
Yes. I didn't remove it, TBH. but I have no other idea.😂
Do you have any insight why generic vector implementation is faster?
I guess the number of instructions for this SVE code block is greater
than C code with scanning one by one.
But If the burst size is bigger, such as 128 and more, the SVE
performance can be enhanced a little.
.
