<snip>
> Subject: RE: [RFC PATCH v1] net/i40e: put mempool cache out of API > > > From: Feifei Wang [mailto:feifei.wa...@arm.com] > > Sent: Monday, 13 June 2022 07.52 > > > > Refer to "i40e_tx_free_bufs_avx512", this patch puts mempool cache out > > of API to free buffers directly. There are two changes different with > > previous version: > > 1. change txep from "i40e_entry" to "i40e_vec_entry" > > 2. put cache out of "mempool_bulk" API to copy buffers into it > > directly > > > > Performance Test with l3fwd neon path: > > with this patch > > n1sdp: no perforamnce change > > amper-altra: +4.0% > > > > Suggested-by: Konstantin Ananyev <konstantin.v.anan...@yandex.ru> > > Suggested-by: Honnappa Nagarahalli <honnappa.nagaraha...@arm.com> > > Signed-off-by: Feifei Wang <feifei.wa...@arm.com> > > --- > > Once again bypassing the mempool API and copy-pasting internal code from > the mempool library to a PMD for performance optimization. > > Certainly not happy about it, but it's already done elsewhere, and thus we > should not deny it here... > > Acked-by: Morten Brørup <m...@smartsharesystems.com> This was done with the intention to understand the performance of the zero-copy mempool APIs that Konstantin suggested. Even though this patch does not have the zero-copy mempool APIs, it reflects the code/instructions of creating such as API. We are trying to show here that zero-copy mempool APIs will not provide equivalent performance of direct-rearm method.
