On 3/20/2019 4:53 AM, Pavan Nikhilesh Bhagavatula wrote: > On Tue, 2019-03-19 at 16:48 +0000, Ferruh Yigit wrote: >> On 3/1/2019 1:47 PM, Pavan Nikhilesh Bhagavatula wrote: >>> From: Pavan Nikhilesh <pbhagavat...@marvell.com> >>> >>> Use mempool bulk get ops to alloc burst of packets and process >>> them. >>> If bulk get fails fallback to rte_mbuf_raw_alloc. >> >> I assume the motivation is performance improvement, do you have the >> numbers >> before and after this patch? > > We are seeing approximately 500% improvement with the patch on > octeontx2.
%500 is interesting. btw, Intel also confirmed the improvement: Tested-by: Yingya Han <yingyax....@intel.com> > >> >> cc'ed maintainers. >> >>> Suggested-by: Andrew Rybchenko <arybche...@solarflare.com> >>> Signed-off-by: Pavan Nikhilesh <pbhagavat...@marvell.com> >> >> <...> >> > >
