Alexander has not published any posts yet.

We did some execution time measurements and had extremely high times in some nested calls compared between two machines. The calls were between some very big routines. The most extreme has its object code splitted to 46 ^rOBJ subnodes and has a total size of >1MB.

You realize the difference only in very long loops (as in the sample above) and deep nested calls to big routines (BTW: classes are compiled to routines too).

The measurements were made on a system with high load so the routine buffer was quickly invalidated by other processes.

I hope this helps.

One more thought:
Depending on the call structure of your methos/classmethods it's necessary to increase routine buffers too.

In our case 128 MB was too less. We configure to 256 MB on all production systems and it runs flawless since.

Alexander has no followers yet.
Alexander has not followed anybody yet.
Global Masters badges:
Alexander has no Global Masters badges yet.