134 lines
3.1 KiB
Plaintext
134 lines
3.1 KiB
Plaintext
GT.M Rocks!!!
|
|
gpl
|
|
|
|
I wrote a standalone test driver that will work on any GT.M system.
|
|
( you have to change the calls to $$NOW^XLFDT to another way to get a timestamp..
|
|
and then do the number of seconds computation differently...I don't know how)
|
|
|
|
I'm attaching the mumps routine C0XTEST.m
|
|
|
|
The first time I ran it I got significantly worse results. Then I realized that I had
|
|
generated a random Graph number for each node.... In our triples load all the
|
|
Graph numbers are the same... Here's what that first run looked like:
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111105.234543
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=10000
|
|
3111105.234543
|
|
INSERTION ENDS AT 3111106.000401
|
|
ELAPSED TIME: 1098 SECONDS
|
|
APPROXIMATELY 9 NODES PER SECOND
|
|
|
|
Second run on gpl (which is a VM) got worse, even though all the Graph pointer are the same.
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.0032
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=10000
|
|
3111106.0032
|
|
INSERTION ENDS AT 3111106.00532
|
|
ELAPSED TIME: 1280 SECONDS
|
|
APPROXIMATELY 7 NODES PER SECOND
|
|
|
|
I put C0XTEST on raven and I couldn't believe my eyes....
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.024924
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=10000
|
|
3111106.024924
|
|
INSERTION ENDS AT 3111106.024924
|
|
ELAPSED TIME: 0 SECONDS
|
|
GTM>W ^C0X(101,0)
|
|
C0X TRIPLE^172.101I^238634^238634
|
|
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.024954
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=10000
|
|
3111106.024954
|
|
INSERTION ENDS AT 3111106.024955
|
|
ELAPSED TIME: 1 SECONDS
|
|
APPROXIMATELY 10000 NODES PER SECOND
|
|
GTM>W ^C0X(101,0)
|
|
C0X TRIPLE^172.101I^248635^248635
|
|
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.025009
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=10000
|
|
3111106.025009
|
|
INSERTION ENDS AT 3111106.02501
|
|
ELAPSED TIME: 1 SECONDS
|
|
APPROXIMATELY 10000 NODES PER SECOND
|
|
GTM>W ^C0X(101,0)
|
|
C0X TRIPLE^172.101I^258636^258636
|
|
|
|
So I upped the anti to 100000 nodes
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.025823
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=100000
|
|
3111106.025823
|
|
INSERTION ENDS AT 3111106.02583
|
|
ELAPSED TIME: 7 SECONDS
|
|
APPROXIMATELY 14285 NODES PER SECOND
|
|
GTM>W ^C0X(101,0)
|
|
C0X TRIPLE^172.101I^658640^658640
|
|
GTM>W $O(^C0X(101,"AA"),-1)
|
|
658639
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.025932
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=100000
|
|
3111106.025932
|
|
INSERTION ENDS AT 3111106.025936
|
|
ELAPSED TIME: 4 SECONDS
|
|
APPROXIMATELY 25000 NODES PER SECOND
|
|
GTM>W ^C0X(101,0)
|
|
C0X TRIPLE^172.101I^758641^758641
|
|
GTM>W $O(^C0X(101,"AA"),-1)
|
|
758640
|
|
|
|
GTM>D EN^C0XTEST
|
|
|
|
INSERTION STARTS AT 3111106.030006
|
|
|
|
USING BULKLOAD
|
|
LOCKING ZERO NODE
|
|
ZERO NODE UNLOCKED, IENS RESERVED=100000
|
|
3111106.030006
|
|
INSERTION ENDS AT 3111106.030015
|
|
ELAPSED TIME: 9 SECONDS
|
|
APPROXIMATELY 11111 NODES PER SECOND
|
|
GTM>W ^C0X(101,0)
|
|
C0X TRIPLE^172.101I^858642^858642
|
|
GTM>W $O(^C0X(101,"AA"),-1)
|
|
858641
|