Folks, I am looking for a review for the following bugs/RFEs - 4781345 me_mutex lock in kcf_mech_entry_t can be broken up 6771819 Use of atomic increment in KCF causes scaling problems on multi-socket T2 systems 6705174 C_EncryptInit scaling issues on T2plus 6813873 assertion failed: (prov_desc)->pd_refcnt != 0, file: ../../common/crypto/core/kcf_prov_tabs.c, line
This work improves the MT performance on all the platforms for both software providers and hardware providers. For example, on an AMD64, throughput with 4 threads went up from 120 MB/sec to 656 MB/sec with a kcf MD5 test that used 64-byte input. I am attaching the performance numbers measured with these changes. The webrev is at http://cr.opensolaris.org/~krishna/kcf_perf_fixes I would appreciate a review by 4/24/09. Thanks, -Krishna -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: perf_numbers.txt URL: <http://mail.opensolaris.org/pipermail/crypto-discuss/attachments/20090415/ecaa7dd1/attachment.txt>
