Hi Roland, Perhaps. We have some ICRC specific APIs as well as plain Jane CRC calculation that includes a copy at the same time. If I knew who maintained crc32.c I would be happy to talk to them and see what they think. The reason I wrote this Is that crc32.c takes about 6 clocks per byte and this one takes < 2 clocks per byte. We couldn't get over 200-300 MB/sec With crc32.c and this algorithm was hitting 900MB/sec. Sandy Bridge has a generic CRC instruction that should be able to Reduce the time to nothing on top of a copy I have heard.
Bob -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Roland Dreier Sent: Friday, July 01, 2011 5:04 PM To: [email protected] Cc: [email protected] Subject: Re: [patch 37/44] rxe_sb8.c On Fri, Jul 1, 2011 at 6:18 AM, <[email protected]> wrote: > Slice by 8 implementation of CRC32. > The code code is similar to the kernel provided crc32 calculation > except runs about 3X faster which allows us to get to ~1GB/sec. Wouldn't the sane thing to do be to fix lib/crc32.c instead? - R. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
