Bill, Thanks, will look at it.
> 在 2015年8月26日，下午7:46，Bill Fischofer <bill.fischo...@linaro.org> 写道： > > HePeng, > > As Mike mentioned, the TM Implementation is currently being staged in its own > repository for ease of development. The patches for that implementation > should be appearing on the ODP mailing list shortly. Barry provided an > overview of this implementation, which includes a discussion of the internal > hashing functions, a few weeks ago during one of the ODP public calls. You > can see the slides for that overview here: > https://docs.google.com/presentation/d/1bJJwgYiDTvE-OdjGq2ceDJYUQVVMkNgKUtHrCVYBv_A/edit#slide=id.p4 > > <https://docs.google.com/presentation/d/1bJJwgYiDTvE-OdjGq2ceDJYUQVVMkNgKUtHrCVYBv_A/edit#slide=id.p4> > > On Wed, Aug 26, 2015 at 5:43 AM, Mike Holmes <mike.hol...@linaro.org > <mailto:mike.hol...@linaro.org>> wrote: > Just as a note that as part of the traffic manager API that is being merged > there is some AES hash functionality being added to the linux-generic > implementation if not the ODP-API > > https://github.com/mike-holmes-linaro/odp/blob/api-next-traffic-mngr/platform/linux-generic/odp_name_table.c > > <https://github.com/mike-holmes-linaro/odp/blob/api-next-traffic-mngr/platform/linux-generic/odp_name_table.c> > > Mike > > On 26 August 2015 at 05:17, Benoît Ganne <benoit.ga...@kalray.eu > <mailto:benoit.ga...@kalray.eu>> wrote: > Hi all, > > I also agree that implementing another algorithm lib may duplicate > efforts. So at the first step, I think maybe contain some > hash tables and LPM functions are just fine. All in all, my original > intention is just to ease the porting from DPDK to ODP. > > So my list begins as follows : > 1. famous hash functions, such as jhash, crc, etc. > 2. hash table implementations, such as cuckoo hash table, etc. > 3. LPM implementations, such as Tree Bitmap algorithms or 16-8-8 > algorithms. > > I those would be very useful. > Regarding CRC, shouldn't it be part of the crypto API or a similar API? At > least on our HW we have asynchronous accelerators for CRC that you can invoke > directly from the cores. They can be used synchronously obviously, bu using > an API similar to crypto (or crypto itself FWIW) would allow to do work in > parallel. > Regarding jhash etc. I don't think there is any HW acceleration on any > platform do it could be in odph. > Regarding LPM, I have a 16-8-8 trie ready that I planned to relase soon to > ODP as part of a l3fwd app example. Tree bitmaps or similar would be better > but LPM algorithms is a patent minefield (and AFAICT tree bitmaps are > patented). > > ben > _______________________________________________ > lng-odp mailing list > lng-odp@lists.linaro.org <mailto:lng-odp@lists.linaro.org> > https://lists.linaro.org/mailman/listinfo/lng-odp > <https://lists.linaro.org/mailman/listinfo/lng-odp> > > > > -- > Mike Holmes > Technical Manager - Linaro Networking Group > Linaro.org <http://www.linaro.org/> │ Open source software for ARM SoCs > > > _______________________________________________ > lng-odp mailing list > lng-odp@lists.linaro.org <mailto:lng-odp@lists.linaro.org> > https://lists.linaro.org/mailman/listinfo/lng-odp > <https://lists.linaro.org/mailman/listinfo/lng-odp> > > > _______________________________________________ > lng-odp mailing list > lng-odp@lists.linaro.org > https://lists.linaro.org/mailman/listinfo/lng-odp
_______________________________________________ lng-odp mailing list lng-odp@lists.linaro.org https://lists.linaro.org/mailman/listinfo/lng-odp
