[RC5] CSC Clients Picking Wrong Cores?

Claude.Saam at alcatel.fr Claude.Saam at alcatel.fr
Fri Nov 19 12:43:03 EST 1999



agreed on this - tested on my own ppro and got this :

[Nov 19 10:59:30 UTC] CSC: using core #0 (6 bit - inline).
[Nov 19 10:59:48 UTC] Benchmarking CSC ... 100.00% done
[Nov 19 10:59:48 UTC] Completed in 0.00:00:17.73 [118,253.50 keys/sec]
[Nov 19 10:59:48 UTC] CSC: using core #1 (6 bit - called).
[Nov 19 11:00:05 UTC] Benchmarking CSC ... 100.00% done
[Nov 19 11:00:05 UTC] Completed in 0.00:00:16.76 [187,629.62 keys/sec]
[Nov 19 11:00:05 UTC] CSC: using core #2 (1 key - inline).
[Nov 19 11:00:24 UTC] Benchmarking CSC ... 100.00% done
[Nov 19 11:00:24 UTC] Completed in 0.00:00:18.64 [112,504.38 keys/sec]
[Nov 19 11:00:24 UTC] CSC: using core #3 (1 key - called).
[Nov 19 11:00:41 UTC] Benchmarking CSC ... 100.00% done
[Nov 19 11:00:41 UTC] Completed in 0.00:00:16.51 [190,469.81 keys/sec]

benchmark show core #3 is the best one, but when client starts :

[Nov 19 11:01:57 UTC] Automatic processor detection found 1 processor.
[Nov 19 11:01:57 UTC] Automatic processor type detection found
                      an Intel Pentium Pro processor.
[Nov 19 11:01:57 UTC] CSC: using core #2 (1 key - inline).
[Nov 19 11:01:57 UTC] Loaded CSC 4*2^28 packet 000BE67C:40000000 (21.40% done)
[Nov 19 11:01:57 UTC] 28 CSC packets (106 work units) remain in buff-in.csc
[Nov 19 11:01:57 UTC] 0 CSC packets (0 work units) are in buff-out.csc
[Nov 19 11:01:57 UTC] 1 cruncher has been started

ARG ! Client coders, please do something :)

PS : the client works fine on "normal" pentium clients, and on a Sparc Solaris Unix box - only ppro seem to be bugged

workaround : start the client with the -c CSC 3 option, to force the 3rd core to be used


--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list