Timeline for Using an Entity System Architecture with Task Based Parallelism
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 27, 2013 at 1:44 | comment | added | Steven Lu | Another comment I wanted to make is it seems to me that the more parallelism you build into the way your components interact, the more difficult it may be to enforce determinism to achieve things such as achieving state sync with minimal network traffic (only sending controller input values). | |
| Sep 27, 2013 at 1:41 | comment | added | Steven Lu | @SeanMiddleditch Could you explain a little further the purpose of the Proxies? If they are only meant to be read from, then how is this any different from allowing all the non-authoritative threads to directly read (and only to be able to read) from a one and only copy of the entity? | |
| Sep 27, 2013 at 1:15 | comment | added | Sean Middleditch | I recommend whatever you find easiest to implement. Just remember that the point I'd to be able to query proxies without taking any locks, without using any atomics, and without deadlocks. | |
| Sep 27, 2013 at 0:30 | comment | added | Ross Hays | @SeanMiddleditch When I say attribute, I am essentially referring to components I think. The attributes are not just single values like floats and strings if that is what I made it sound like. Rather they are classes that contain specific information like a PositionAttribute. If component is the accepted name for that then perhaps I should change. But would you recommend proxying at the entity level rather than the component/attribute level? | |
| Sep 27, 2013 at 0:17 | vote | accept | Ross Hays | ||
| Sep 27, 2013 at 0:06 | comment | added | Sean Middleditch | @roflha: yeah you could put the proxies at the EntityAttribute level. Or make a separate entity with a second set of attributes. Or just drop the concept of attributes altogether and use a less granular component design. | |
| Sep 27, 2013 at 0:01 | comment | added | Patrick Hughes | There's a good chance that in trying to avoid waiting you spend so much time analyzing the dependency graph that you lose time overall. | |
| Sep 26, 2013 at 23:47 | comment | added | Ross Hays | I had read some things about the fork join you mentioned before and I was under the impression that while it does allow you to utilize some parallelism, there are situations in which some worker threads may be waiting on one group to finish. Ideally, I am trying to avoid that situation. The proxy idea is interesting and slightly resembles what I was working on. An Entity has EntityAttributes and those are wrappers for the actually values stored by the entity. So values could be read from them anytime but only set at certain times and could contain a proxy value in the attribute, correct? | |
| Sep 26, 2013 at 23:33 | history | answered | Sean Middleditch | CC BY-SA 3.0 |