![[se]](/mwf/v12/flags/se.png "Sweden")
Hi,
I've been using your fantastic tool Aquarium for quite a while and have a pretty large number of projects.
Unfortunately, I didn't read about master trees until recently and have problems with very slow processing, while
many big projects uses the same tree.
However, as I understand it it is possible to do a migration. But before I do that I would like to have some expert opinion that it works and that it doesn't lose my valuable analysis...
The basic idea is to make the current common tree a master tree and create an individual tree for each project.
For each project:
1. Define a new tree for the project (named same as project)
2. Set the current common tree as the new master tree for the project
3. Import from the new master tree to the new local tree.
Will it work?
Thanks in advance,
Dean
I've been using your fantastic tool Aquarium for quite a while and have a pretty large number of projects.
Unfortunately, I didn't read about master trees until recently and have problems with very slow processing, while
many big projects uses the same tree.
However, as I understand it it is possible to do a migration. But before I do that I would like to have some expert opinion that it works and that it doesn't lose my valuable analysis...
The basic idea is to make the current common tree a master tree and create an individual tree for each project.
For each project:
1. Define a new tree for the project (named same as project)
2. Set the current common tree as the new master tree for the project
3. Import from the new master tree to the new local tree.
Will it work?
Thanks in advance,
Dean
![[us]](/mwf/v12/flags/us.png "United States")
Yes. Be sure and make backups of the tree before doing anything.
OK, thanks it seems to work.
Great! 
Another related thing:
What size of projects are possible before things get sluggish? Sometimes you definitely would like to have 50 000 positions or more.
Does using multiple root nodes solve the performance issues is some way?
Or put more technically: Is the scaling linear, polynomial or exponential?
What size of projects are possible before things get sluggish? Sometimes you definitely would like to have 50 000 positions or more.
Does using multiple root nodes solve the performance issues is some way?
Or put more technically: Is the scaling linear, polynomial or exponential?
![[es]](/mwf/v12/flags/es.png "Spain")
You will find that master size is almost unlimited, since they are not intended to be minimaxed
![[gr]](/mwf/v12/flags/gr.png "Greece")
working with 100K+ positions is fine. The problem in such big trees is not minimaxing but task generation. Because it has to scan and filter such a huge amount of data.
When i reach around that point, i 'upload" to the master tree then continue by "importing" the branch i am working on.
When i reach around that point, i 'upload" to the master tree then continue by "importing" the branch i am working on.
As I have experienced it, it is the "filling number of" (sub-nodes) that is the bottleneck when you close/finish your analysis.
Currently I'm waiting for just that to finish for a project with about 60 000 positions.... it still takes a while despite the master tree. Actually, the number of positions is a bit hard to tell while transpositions seems to be counted many times in some circular way.
Currently I'm waiting for just that to finish for a project with about 60 000 positions.... it still takes a while despite the master tree. Actually, the number of positions is a bit hard to tell while transpositions seems to be counted many times in some circular way.
i NEVER use filling number. I dont care about accurate node counts. They take a lot of time and they do not give useful information .
I actually think position counts should be completely removed and have a button to calculate the counts on demand for a given position (which is super faster than calculating for every node in the subtree)
That choice alone will save ALOT of valuable computer-hours spend on stupid position counts, even when you are asleep or IDEA is left unattended.
I actually think position counts should be completely removed and have a button to calculate the counts on demand for a given position (which is super faster than calculating for every node in the subtree)
That choice alone will save ALOT of valuable computer-hours spend on stupid position counts, even when you are asleep or IDEA is left unattended.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill