User 677b9c22ff
09-12-2006 23:57:45
Hi,
I was just testing the new version Instant-JChem version (1.02) and now
the calculation of properties is really going faster. The 250k test set took
now 12 minutes on a 8-core machine (running with JAVA server version) instead of more than one hour for logP with only one CPU.
Thats what I call nice and cheap-chip multihreading, just split lists and its fun.
This translates to 350 molecules/sec for logP which is pretty good compared to the normal 70~100 molecules/sec.
But its still far away from the 1900 molecules/second from the compiled KOWWIN. But the JChem version is much more flexible
to use and not to forget pka and logD calculations and complex terms you can built from it.
Kind regards
Tobias Kind
I was just testing the new version Instant-JChem version (1.02) and now
the calculation of properties is really going faster. The 250k test set took
now 12 minutes on a 8-core machine (running with JAVA server version) instead of more than one hour for logP with only one CPU.
Thats what I call nice and cheap-chip multihreading, just split lists and its fun.
This translates to 350 molecules/sec for logP which is pretty good compared to the normal 70~100 molecules/sec.
But its still far away from the 1900 molecules/second from the compiled KOWWIN. But the JChem version is much more flexible
to use and not to forget pka and logD calculations and complex terms you can built from it.
Kind regards
Tobias Kind