Print

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

[Schmaven]

I'm going to be upgrading my PC soon. How is the new version's multithreading/multiple CPU core use looking?

In the past I've prioritized processor speed over core count because of Dwarf Fortress, and I'd like to know whether I should continue to do so, lol

I was under the impression that it would involve a significant amount of work for Dwarf Fortress to be able to take advantage of multithreading/multiple CPU cores, like big wait kind of work.  CPU cache size might be relevant to some extent, given that it is what feeds data to the CPU to process, but I'm not sure of a way to check for bottlenecks there if they even exist.  RAM speed may also be a factor in fort FPS.  Large RAM volumes on the cheap typically involve sacrifices on speed.  Lower CAS latency is generally better, all other things being equal, but there are several variables there that all come into play. 

[PatrikLundell]

There is no reason to assume any multi threading efforts are being made in the near term: there is more than enough work on the plate for it to all be done before the release, so there is little room for new experiments.

Making DF multi threaded would be a monumental task that may not even provide that much benefit, given that the main bottleneck seems to be memory bandwidth rather than raw CPU speed.

[☼Obsidian Short Sword☼]

The main bottleneck seems to be memory bandwidth rather than raw CPU speed.

PatrikLundell is right, The wiki recommends fast ram and a large cache.

Reducing memory usage/size would probably be a better optimization, Since more data could fit in the L1 cache(Which is very fast).

[A_Curious_Cat]

The main bottleneck seems to be memory bandwidth rather than raw CPU speed.

PatrikLundell is right, The wiki recommends fast ram and a large cache.

Reducing memory usage/size would probably be a better optimization, Since more data could fit in the L1 cache(Which is very fast).

Correct me if I’m wrong, but doesn’t each core in a multi-core processor have it’s own L1 cache?  Currently (with DF being single-threaded), the game only uses the L1 cache on a single core.  Would it be possible to split the map into multiple l x m x n 3d segments and have each segment simulated by a separate core, with an additional core handling communications and synchronization between segments (and, of course, an additional core handling graphics)?

[☼Obsidian Short Sword☼]

PatrikLundell is right, The wiki recommends fast ram and a large cache.

Reducing memory usage/size would probably be a better optimization, Since more data could fit in the L1 cache(Which is very fast).

Correct me if I’m wrong, but doesn’t each core in a multi-core processor have it’s own L1 cache?  Currently (with DF being single-threaded), the game only uses the L1 cache on a single core.  Would it be possible to split the map into multiple l x m x n 3d segments and have each segment simulated by a separate core, with an additional core handling communications and synchronization between segments (and, of course, an additional core handling graphics)?

Sounds like a good system to me, But the overhead of communication might be slower than using the L2/L3 cache that is shared between cores,  I think benchmarking will be best. But this is just my opinion, I'm not a hardware guy.

On the topic of multithreading take a look at this(dwarffortress.exe):



I am using dfhack though, So it might be different.

[PatrikLundell]

Multi threading requires:
- Rewriting of the code to support it, which is a lot of work (or, more probably, an enormous amount of work).
- Code supporting multi threading requires data synchronization to ensure the data used is the current data, isn't in the middle of being changed (lots of "interesting" things can happen when half of a structure is new and the other half is old), and apart from the ample opportunity for hard to reproduce bugs, it also adds overhead that cuts into the savings (and can cost more than the time saved).

C(++) is also unsuitable for multi threading, although it's used for a lot of things it's not suited for because a lot of people's only tool is a hammer, so every problem is a nail.

[Fikilili]

Thanks for the replies Tarn! Here is a new batch of question for the next FotF!
Originally I was gonna ask about boats because you mentionned O'Brian's sailing books, but hey, too early to talk about that.

So given that we'll get the Myth and Magic update after the Steam Release (?), I wanted to ask a few questions regarding that.
Real-Life mythologies, although sometimes chaotic and having varying degrees of consistency depending on the period, have some form of inner-logic. There is a reason for this or that god to be there, or this or that agent of the gods to exist. DF's current handling of the gods is quite ubiquitous at times. Do you plan on actually retooling or entirely reworking the way Myths work so to generate some that are more accurate to the way real mythologies are formed/written?

[voliol]

Thanks for the replies Tarn! Here is a new batch of question for the next FotF!
Originally I was gonna ask about boats because you mentionned O'Brian's sailing books, but hey, too early to talk about that.

So given that we'll get the Myth and Magic update after the Steam Release (?), I wanted to ask a few questions regarding that.
Real-Life mythologies, although sometimes chaotic and having varying degrees of consistency depending on the period, have some form of inner-logic. There is a reason for this or that god to be there, or this or that agent of the gods to exist. DF's current handling of the gods is quite ubiquitous at times. Do you plan on actually retooling or entirely reworking the way Myths work so to generate some that are more accurate to the way real mythologies are formed/written?

I expect the answer from Toady to be simply "yes", as it is indeed planned to be half of that update. There are some old demos of it (as in showcases, the myth-demo software isn't available to us), and he also talks a bunch about it in the recent DF Talks. #24 I think has the highest concentration.

[Putnam]

C(++) is also unsuitable for multi threading, although it's used for a lot of things it's not suited for because a lot of people's only tool is a hammer, so every problem is a nail.

Definitely can't agree here, I've used it for multithreading without much pain in the past, mostly with C++17's execution policies

[BlueManedHawk]

[deleted, though frankly i'm not entirely sure why]

[BlackAion]

Would it be possible in the Myth and Magic expansion to create sky worlds? Like worlds where the continents are floating and the oceans are an endless sea of clouds? Or worlds where you live on the back of a turtle or worlds where it is literally one big giant tree?

[PatrikLundell]

Would it be possible in the Myth and Magic expansion to create sky worlds? Like worlds where the continents are floating and the oceans are an endless sea of clouds? Or worlds where you live on the back of a turtle or worlds where it is literally one big giant tree?

In principle the answer is yes. The goal is to allow for these kind of things. However, the first arc certainly won't implement everything, and may only have small entries into these kind of things, and it may even wait until the moving equipment arc to actually be implemented to any extent. Floating islands and cloud worlds are definitely thinks that the map rewrite part is intended to enable, but providing a foundation is not the same as implementing the contents on top of it.

[Fikilili]

Thanks for the replies Tarn! Here is a new batch of question for the next FotF!
Originally I was gonna ask about boats because you mentionned O'Brian's sailing books, but hey, too early to talk about that.

So given that we'll get the Myth and Magic update after the Steam Release (?), I wanted to ask a few questions regarding that.
Real-Life mythologies, although sometimes chaotic and having varying degrees of consistency depending on the period, have some form of inner-logic. There is a reason for this or that god to be there, or this or that agent of the gods to exist. DF's current handling of the gods is quite ubiquitous at times. Do you plan on actually retooling or entirely reworking the way Myths work so to generate some that are more accurate to the way real mythologies are formed/written?

I expect the answer from Toady to be simply "yes", as it is indeed planned to be half of that update. There are some old demos of it (as in showcases, the myth-demo software isn't available to us), and he also talks a bunch about it in the recent DF Talks. #24 I think has the highest concentration.

Right. Might as well come up with another question then.

[Shepanator]

Are there plans to add more gameplay features for forts with 'Mountainhome' status, e.g. ability to direct the overall civilisation, declare wars, expand the civ, etc? Or is there any other 'end-game' type content you have been thinking about?
Thanks and much love

[☼Obsidian Short Sword☼]

Are there plans to add more gameplay features for forts with 'Mountainhome' status, e.g. ability to direct the overall civilisation, declare wars, expand the civ, etc? Or is there any other 'end-game' type content you have been thinking about?
Thanks and much love

Having an optional 4x strategy game would be cool, But this would probably tie in better with Adventure Mode IMO. Becoming king could allow you to manipulate things on a larger scale.