User talk:Philosopher/Nasal introspection: Difference between revisions

User talk:Philosopher/Nasal introspection (view source)

261 bytes added , 16 August 2013

m

20,741

edits

@@ Line 85: / Line 85: @@
 ::: Yeah, completely agreed - ideally, we really WANT the full session to be profiled, but obviously there are still some issues, so we should slow down a second and fix the call() implementation first to be able to profile individual scripts. That would give us an idea about what's going on - profiling the whole session is still too massive timing/memory-wise, so that we need to use call() as a testbed to understand what's happening before we can look again at profiling the entire sim session, which I also consider a primary goal. Regarding listeners and timers, I think we could patch the FG side of things to "inherit" extras based on the initial call().--[[User:Hooray|Hooray]] ([[User talk:Hooray|talk]]) 21:29, 14 August 2013 (UTC)
-:: then just make it a variable and use a property switch/listener to make it runtime configurable?
+:::: It would be really interesting if we could call() the test.nas infrastructure to get ITSELF profiled when profiling the sim session, to identify which parts of test.nas are heavy (i.e. responsible for 200ms frame spacing), and why (GC overhead, callgraph, data structures). Need to do some more thinking to figure out if/how that could be made to work...
 <hr/>