You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
By adding #150, I came to the conclusion that memory_allocate seems to have some peculiarities going on, which sometimes degrades its performance by a factor of 500 to 1000 - i.e. the system goes from having reasonable real-time performance (making it possible to play mod files with #145) to being completely sluggish and barely usable.
This screenshot illustrates the problem ("Last memory allocation took" is the key):
I believe the only reasonable way to solve this is by fixing #4 using the approach suggested in #4 (comment) and I will see if it can be done with a reasonable effort.
The text was updated successfully, but these errors were encountered:
perlun
changed the title
[storm] Find out why memory allocations start taking much longer after a while
[storm] Find out why global memory allocations start taking much longer after a while
Jul 8, 2019
Note: the screenshot is inverted because of a bug on my behalf; it's the global memory allocations that are taking an enormous amount of time, not the other way around.
Regarding the timings, I'm not sure how accurate they are. (We need better timers I think, especially ones that work in virtualized scenarios also. Perhaps using HPET or something would make sense; HPET in particular is blacklisted on Linux for some Intel chips though, so it might not be a great fit after all. 😒)
By adding #150, I came to the conclusion that
memory_allocate
seems to have some peculiarities going on, which sometimes degrades its performance by a factor of 500 to 1000 - i.e. the system goes from having reasonable real-time performance (making it possible to play mod files with #145) to being completely sluggish and barely usable.This screenshot illustrates the problem ("Last memory allocation took" is the key):
I believe the only reasonable way to solve this is by fixing #4 using the approach suggested in #4 (comment) and I will see if it can be done with a reasonable effort.
The text was updated successfully, but these errors were encountered: