Re: [Algorithms] General purpose task parallel threading approach

[Jon W. <jw...@gm...>]

Nicholas "Indy" Ray wrote:
> In my experience the order of events within a single thread is set and
> doesn't change, even though depending on the user certain tasks won't
> need to be spawned, It's trivial to setup the order of tasks *within*
> a single frame to have a static order, whether that is optimal or not
> is another matter altogether.
>   

Only if you legislate that reactions to events can only happen on the 
next frame. Else, you may have an event of type A that triggers an event 
of type B, and another event of type B that triggers an event of type A, 
and you can't get a static order that solves both.
For example: grenades exploding may put things on fire, but things 
burning may also explode grenades. Only in a system where you tie-break 
by introducing latency would you be able to use a static ordering. 
Meanwhile, in a system where work items are actual individual items 
(rather than static "pulsed" state machines or whatever), either case 
will be resolved within the same frame.

Sincerely,

jw