Rebuild speed / quick flash
Case number: | 845813-987558 |
Topic: | Game: Tools |
Opened by: | steveB |
Status: | Closed |
Type: | Bug |
Opened on: | Wednesday, April 7, 2010 - 08:26 |
Last modified: | Thursday, May 6, 2010 - 21:03 |
Hi Seth,
Rebuild Speed
The rebuild speed varies drastically when bands are/are not attached to the protein. To speed up rebuilds I simply attach a band between any two random points on the structure, and I see a doubling or more in the rebuild speed. Why does this happen ?
Quick Flash
When rebuilding it is noticable that when the rebuild algorithm detects a high scoring configuration that it moves on so quickly to the next rebuild configuration that the graphics do not have time to render on the screen. In other words, when rebuilding you only get a quick flash of a really good rebuild and not enough time to see it. Could you insert some idling code into the algorithm so that these good rebuilds don't wizz by so fast ?
SteveB
Rebuild Speed
Anything that can double the speed on rebuilds would help me for time.
Quick Flash
For some players it may be a "quick flash" for other its more like a slow parade.
People are playing folding on a large range of computer hardware with a big difference in speed of over 10x. A quick look at cpu benchmark results would show this.
Suggestion
Give players more control of the rebuild tools.
1) A choice of algorithm.
2) Speed controller.
Hi B_H,
The band is completely at random. Even on an unrelated part of the structure to the rebuild, the addition of a band causes vast rebuild speed increases.
Status: Open » Open |
Type: Suggestion » Bug |
ok wow... that is crazy!
We'll try to figure out what is going on with that!
thanks steveB
Status: Open » Closed |
When you add bands to a rebuild, it will try to get the protein as close to those bands as possible. In this case it checks moves early on to see if the bands get stretched out too much, and uses that as an "early reject" criteria before doing any more processing on that move. Without bands there is no such early rejection check so each move does more processing.
I have fixed it so rebuild will check that the bands are in the rebuilding region, and delay a bit between moving between configurations if it happens to quickly.
This will go out in a future release.
Status: Closed » Closed |
This has been posted in an update.
My suggestion is, when doing a rebuild, that those possible solutions (when the shape changes) should be stored in the undo bar as unique point. This would help finding "score peaks" and you don't need to have fast relexes.
Hey steveB,
just to clarify your first question: you are banding residues on the section you are trying to rebuild, right?
(or are you banding 2 random residues that are nowhere near the area being rebuilt?)
Thanks!