Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Putnam]

#4 is the game's software rescale function. 4.3%, 7.440s

Does playing the game at 100% scale help a lot with lowering this number? Then only 8x12 interface/text tiles would still need scaling.

Yes, zoom out twice for 1:1 and it avoids this I think? Keep in mind that it is on the graphics thread and thus has little-to-no impact on FPS

[Untrustedlife]

Yes, but negligibly, the main issue here is activities (storytelling, poetry etc.) with dozens or even over a hundred units involved

Did you see what I asked?

[Fantastic Damage]

In my first Steam Edition Fort back at the beginning of December'22 several actions that would spook the game, from strafing the map too quickly to saving the ###ing game
would result in Core 1 of 16 to be at 100%+ Capacity whilst the other 15 were 5-0%. On a 12600KF with 16GB 4000mhz RAM. It Struck me as a little excessive.

[Maolagin]

Fascinating results, and I'm currently playing a fortress that I suspect is demonstrating this perfectly. 120 citizens, all caverns+magma revealed (but sealed off), 250 year history, yet still sitting at my 50 FPS frame cap. What's different here? It's a reclaimed worldgen fort.

If the FPS issue was pathfinding, those should be awful - they're expansive and convoluted. But instead I think I'm benefiting from that. Spreading all the workshops and meeting areas through that convoluted space means there's rarely more than a dozen people in one place, and they can't see into neighboring areas at all. I wonder if it further helps that most of the fort was dug out for me already. I get all that space to spread out in, and even a huge surface building to block out the great outdoors, but not thousands of boulders for the game to track.

[Warmist]

Yes, but negligibly, the main issue here is activities (storytelling, poetry etc.) with dozens or even over a hundred units involved

Did you see what I asked?

Yes. Putnam and other people have been talking about this for (approx) 5 years if not more.

Edit: to be more precise this usually was talked about in dfhack contexts (i.e. dfhack irc or discord)

[Putnam]

Yes, I've talked a few times about rewriting basically the whole system to be more structure-of-arrays instead of array-of-structures, with more caching of things in a lot of places, but that's an ordeal. Like, it would take longer to implement than multithreading, albeit easier (a lot more work but it's far less likely to break for mysterious reasons)

[ab9rf]

A dwarf will never meet their grandfather in your fort if their grandfather is not in the fort.

How else are they supposed to know if their grandfather is in the fort?

You can do O(1) existence lookups, like say a hash table of creatures in the fort

Sure, why don't you get right on that.

[Treah]

SDL2 update is mostly done, I didn't do terribly much in the way of optimization besides the inherent optimization you get from moving to the GPU; however, you can just turn off the rescale function, so at least that's going on.

There's already a texture cache for rescaled stuff, it's just invalidated rather often. A bit less often with SDL2, though.

Kinda wish there was a Linux version of premium to be fair.. Running it in proton is buggy, workable but buggy and very slow for some reason then on native windows.

[ayy1337]

You can do O(1) existence lookups, like say a hash table of creatures in the fort

Sure, why don't you get right on that.

Sure mate just let me fire up my copy of the source code and write it in :')

[Putnam]

SDL2 update is mostly done, I didn't do terribly much in the way of optimization besides the inherent optimization you get from moving to the GPU; however, you can just turn off the rescale function, so at least that's going on.

There's already a texture cache for rescaled stuff, it's just invalidated rather often. A bit less often with SDL2, though.

Kinda wish there was a Linux version of premium to be fair.. Running it in proton is buggy, workable but buggy and very slow for some reason then on native windows.

working on it, main issue right now is concurrency woes (the game does use multithreading, to an extent, and even the very minor extent it does use it is causing issues lol)

[tonnot98]

(How much space is a 1x1 micro-fort these days?)  Something like the Epyc 7373X, which has the highest *steady state* clock rate of the 768 MB L3 chips, would only set you back around $4k. 

Savegame memory is heavily dependent on world size and history length. The Irritan Anam succession game is a med world with 100 years of pregen history and 7 years of fortress play, and it's only 141MB. The Museum III generated 1000 years with a very large world, and by turn 78 (the one I have on hand) it has gone through a few extra centuries of multiple fortresses and adventurers. It's 766MB, just under the max load of that L3 chip.

[Kestrel]

SDL2 update is mostly done, I didn't do terribly much in the way of optimization besides the inherent optimization you get from moving to the GPU; however, you can just turn off the rescale function, so at least that's going on.

There's already a texture cache for rescaled stuff, it's just invalidated rather often. A bit less often with SDL2, though.

Kinda wish there was a Linux version of premium to be fair.. Running it in proton is buggy, workable but buggy and very slow for some reason then on native windows.

working on it, main issue right now is concurrency woes (the game does use multithreading, to an extent, and even the very minor extent it does use it is causing issues lol)

Glad to hear this is a Linux issue (and disheartened but good to hear it's on the radar to be addressed). I've had to abandon forts so early because of severe FPS drops with sub-100 populations even. My Windows version is running fine so far.

[Treah]

SDL2 update is mostly done, I didn't do terribly much in the way of optimization besides the inherent optimization you get from moving to the GPU; however, you can just turn off the rescale function, so at least that's going on.

There's already a texture cache for rescaled stuff, it's just invalidated rather often. A bit less often with SDL2, though.

Kinda wish there was a Linux version of premium to be fair.. Running it in proton is buggy, workable but buggy and very slow for some reason then on native windows.

working on it, main issue right now is concurrency woes (the game does use multithreading, to an extent, and even the very minor extent it does use it is causing issues lol)

Glad to hear this is a Linux issue (and disheartened but good to hear it's on the radar to be addressed). I've had to abandon forts so early because of severe FPS drops with sub-100 populations even. My Windows version is running fine so far.

I haven't tried running it with pure wine however so it may actually run a little better that way. I may try this when I get my gentoo install completed on my laptop.

[Bralbaard]

(How much space is a 1x1 micro-fort these days?)  Something like the Epyc 7373X, which has the highest *steady state* clock rate of the 768 MB L3 chips, would only set you back around $4k. 

Savegame memory is heavily dependent on world size and history length. The Irritan Anam succession game is a med world with 100 years of pregen history and 7 years of fortress play, and it's only 141MB. The Museum III generated 1000 years with a very large world, and by turn 78 (the one I have on hand) it has gone through a few extra centuries of multiple fortresses and adventurers. It's 766MB, just under the max load of that L3 chip.

The main reason the museum world save is large is the ridiculous amount of player fortresses (60+) and the fact that the whole world has been explored/visited with people leaving a mess everywhere that needs to be saved.  It's only a medium sized world if you account for the weird dimensions, worldgen was 700 years, but with very low populations because a zombie apocalypse happened.

[Putnam]

Native Linux version seems to be working now, but thorough testing is quite difficult, so look forward to that. Mac version will need a setup to compile on.