20,741
edits
Philosopher (talk | contribs) m (Cut some unneeded stuff (that's what history is for - lucky us), add headings) |
m (NOT using gitorious) |
||
| Line 1: | Line 1: | ||
== Race Condition ? == | |||
In its current form, the test.nas script will profile the entire session - when running it, I am getting this when exiting: | |||
<pre> | |||
Done processing stack | |||
Found more work | |||
Processing stack | |||
Done processing stack | |||
End thread 0 @ 0 | |||
Done | |||
_dump | |||
Warning: it looks like the thread didn't release the lock | |||
Warning: it looks like the thread didn't release the lock | |||
</pre> | |||
When just profiling the test function itself, there's no such issue (less data & less threads presumably?): | |||
<pre> | |||
End thread 0 @ 0 | |||
**** Nasal garbage collection statistics: objects: 39231, references: 90961 | |||
Loading local weather routines... | |||
**** Nasal garbage collection statistics: objects: 66574, references: 150574 | |||
**** Nasal garbage collection statistics: objects: 72047, references: 162214 | |||
**** Nasal garbage collection statistics: objects: 17427, references: 64621 | |||
Exit | |||
Waiting | |||
Done | |||
_dump | |||
</pre> | |||
frame latency and frame rate get back to normal fairly quickly now, when profiling the entire sim, I am getting 180ms here - so, we'll still need to get it down by a factor of ~5-10. | |||
The "call() with extras" doesn't seem to be affected when profiling the entire sim session, not sure how this works internally, recursion-wise and WRT file/data processing ? | |||
== August 14 == | == August 14 == | ||