Comments on: Microcell Distribution http://hazelmckendrick.com/journal/microcell-distribution Programming and Games Technology Portfolio Sun, 20 Mar 2011 18:08:09 +0000 hourly 1 http://wordpress.org/?v=3.1.2 By: Hazel http://hazelmckendrick.com/journal/microcell-distribution/comment-page-1#comment-1209 Hazel Mon, 10 May 2010 11:58:41 +0000 http://hazelmckendrick.com/?p=1946#comment-1209 Hi, That's right, I'm just madly writing up now. I still have a few results I'd like to verify, but it has been hard to find a time when I can access higher numbers of computers at the university, even at the weekend, so I probably won't get a chance. Higher perimeters for a node are generally a bad thing, since there are likely to be more transfers that way, however the pathfinding maps mean that not every edge is even - some might never have any transfers at all if entities can't walk there. Per cell, the work is pretty much just some characters using A* to find paths to other random point on the graph as you say. The inter-cell communication isn't particularly interesting I'm afraid. Message queues would make sense if I was passing much more than entities between cells, but as it stands there are only message queues per-node for now. I'm not unhappy with how the project turned out, but I do feel that the topic area was too broad and that I should have focussed down on one area much earlier. In the end I investigated a lot of aspects of the simulation, but none of them in quite the detail I would have like. Hi,
That’s right, I’m just madly writing up now. I still have a few results I’d like to verify, but it has been hard to find a time when I can access higher numbers of computers at the university, even at the weekend, so I probably won’t get a chance.

Higher perimeters for a node are generally a bad thing, since there are likely to be more transfers that way, however the pathfinding maps mean that not every edge is even – some might never have any transfers at all if entities can’t walk there.
Per cell, the work is pretty much just some characters using A* to find paths to other random point on the graph as you say.

The inter-cell communication isn’t particularly interesting I’m afraid. Message queues would make sense if I was passing much more than entities between cells, but as it stands there are only message queues per-node for now.

I’m not unhappy with how the project turned out, but I do feel that the topic area was too broad and that I should have focussed down on one area much earlier. In the end I investigated a lot of aspects of the simulation, but none of them in quite the detail I would have like.

]]> By: Weeble http://hazelmckendrick.com/journal/microcell-distribution/comment-page-1#comment-1208 Weeble Mon, 03 May 2010 23:27:55 +0000 http://hazelmckendrick.com/?p=1946#comment-1208 Is high "surface area" a bad property for cells allocated to a node, since it increases the amount of inter-node communication? What sort of work is getting done per-cell? You mentioned pathfinding previously. Are you keeping the per-cell work pretty simple to focus on the high-level assignment of work? How interesting is the inter-cell communication? Do cells have message queues or are they only for nodes? *Looks back at Gantt chart* Is it pretty much finished now? Just writing up left? Is high “surface area” a bad property for cells allocated to a node, since it increases the amount of inter-node communication? What sort of work is getting done per-cell? You mentioned pathfinding previously. Are you keeping the per-cell work pretty simple to focus on the high-level assignment of work? How interesting is the inter-cell communication? Do cells have message queues or are they only for nodes?

*Looks back at Gantt chart*

Is it pretty much finished now? Just writing up left?

]]>