Print

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

[JamesK89]

Why does it even matter for game speed? Obviously there's a problem in a way game interacts with open tiles.

Pathfinding.

Quite simply, I know people looked into understanding how DF manages Dwarf AI and pathfinding -which remains the single biggest FPS killer. The more dwarves, creatures, paths, and items for them to interact with = worse fps.

The problem is that DF probably tries to solve pathfinding for everything at once every frame.
I believe the solution would be to do what RTS games do and put everybody in a queue and solve pathfinding for one or two units every frame or two.

[Lordinquisitor]

Hmm, didn`t want to open a new thread for this..

In every dwarf fortress version (Excluding 31.11 which i haven`t played.) i could play with ~180 FPS at embark. But now, in 31.12 i only get 80-100 FPS upon embark. Obviously i`m not amused.

I can also only use the 2D setting without getting a crash at the launch of the game.

Seriously, what happened?

[devek]

Computing power hasn't been about processing speed for a long time guys.

You are going to make DF faster by changing the pathing algorithm, this isn't 1960 anymore. The act of collecting the information to path in the first place takes more time than calculating it. Newer processors are faster because they can deal with lots of information better, not because they do individual calculations faster. Additionally, how you go about deciding when to path and where to store the information it needs will impact speed more than how you do the actually path.

[Shrugging Khan]

The problem is that DF probably tries to solve pathfinding for everything at once every frame.
I believe the solution would be to do what RTS games do and put everybody in a queue and solve pathfinding for one or two units every frame or two.

My opinion also. Pathfinding doesn't have to be done instantaneously. If the critters in DF have slower reaction times but make the game playably fast in return, that's a trade-off I'm a-OK with.

[jaked122]

I'm still trying to find out how to change the linux kernel so that it will defer all power over to DF. therefore we can have a DF-OS and it will be wonderful and nobody will die in the process.

[Marconius]

It should be noted that current processor cores aren't really getting more advanced. It's very hard to make a core even faster, given the limitations of space and heating. What most processor companies do these days is increase the number of cores in processors (dual cores are common these days and even quad cores are spreading) and delegate certain functions to other sub-processors; for example, graphic functions are done by GPU and lately, physics processors done via a so-called PPU are spreading too.

The problem is, as long as a program (or process, if you want to be accurate) runs on a single thread, it can only run on a single core as well. That means that even though I have a quad-core processor, DF can only use 25% of its maximum capacity.

Yeah, there are probably a lot of other places where optimization could be made (can't be exact here, since I obviously don't have an information as to how DF is designed and implemented). But as much as its cited on these forums, as long as DF remains single-thread, it won't be able to take advantage of modern hardware.

[Eugenitor]

DF-OS [...] nobody will die in the process.

I'm pretty sure that if you made a DF-OS, magma would start pouring out of your computer.

[Twitch]

No, there is in fact a wall. A giant, unbreakable wall named Physics. Processors can only get so small. And then comes the power issues that lead to unwanted antennae. Not fun.

Actually, they just figured out how to make a silicon-based laser. Since light beams can't intersect and bounce off each other, once silicon lasers get cheaper that will open up all kinds of space when beams of light replace leads inside the physical processor itself. Lasers are also alot cooler than wires because it only generates heat at the endpoints.
I predict that cores in the future will be tight rings or spheres of components firing binary lasers to each other through a vacuum, and a design like this should be able to hit the tens of gHz easily.
In the far far future processors should be using quantum entanglement instead of wires or lasers, and instead of binary they will use base 6 (instead of 0 and 1, the six flavors of quarks.)
Remember not long ago when people said computers had hit the limit because you couldnt make vacuum tubes any smaller? Exactly.

[nophilip]

No, there is in fact a wall. A giant, unbreakable wall named Physics. Processors can only get so small. And then comes the power issues that lead to unwanted antennae. Not fun.

Actually, they just figured out how to make a silicon-based laser. Since light beams can't intersect and bounce off each other, once silicon lasers get cheaper that will open up all kinds of space when beams of light replace leads inside the physical processor itself. Lasers are also alot cooler than wires because it only generates heat at the endpoints.
I predict that cores in the future will be tight rings or spheres of components firing binary lasers to each other through a vacuum, and a design like this should be able to hit the tens of gHz easily.
In the far far future processors should be using quantum entanglement instead of wires or lasers, and instead of binary they will use base 6 (instead of 0 and 1, the six flavors of quarks.)
Remember not long ago when people said computers had hit the limit because you couldnt make vacuum tubes any smaller? Exactly.

You're absolutely right. The wall I was referring to is inevitable under the current system. Eventually a new system will replace the current one, and that one will have its own limit. The trick is finding the new system every 15-25 years or so.

[nbonaparte]

No, there is in fact a wall. A giant, unbreakable wall named Physics. Processors can only get so small. And then comes the power issues that lead to unwanted antennae. Not fun.

Actually, they just figured out how to make a silicon-based laser. Since light beams can't intersect and bounce off each other, once silicon lasers get cheaper that will open up all kinds of space when beams of light replace leads inside the physical processor itself. Lasers are also alot cooler than wires because it only generates heat at the endpoints.
I predict that cores in the future will be tight rings or spheres of components firing binary lasers to each other through a vacuum, and a design like this should be able to hit the tens of gHz easily.
In the far far future processors should be using quantum entanglement instead of wires or lasers, and instead of binary they will use base 6 (instead of 0 and 1, the six flavors of quarks.)
Remember not long ago when people said computers had hit the limit because you couldnt make vacuum tubes any smaller? Exactly.

You're absolutely right. The wall I was referring to is inevitable under the current system. Eventually a new system will replace the current one, and that one will have its own limit. The trick is finding the new system every 15-25 years or so.

Most likely, the next system will be graphene based. graphene is basically a single atom thick sheet of graphite; it's a hexagonal carbon lattice.  It's also an extremely good conductor. By bonding hydrogen atoms to specific carbon atoms in the lattice, the resistance can be changed. This will allow for 'transistors' composed of only a few atoms. that was only a basic explanation, but google can explain further.

[devek]

Well, you're making computers more efficient but there is still a wall that will be reached.

Once 100% of the power/energy that goes into a computer is used for computations, you can't go further

The good news is, that computer wouldn't require a heatsink or any cooling...

[nophilip]

Well, you're making computers more efficient but there is still a wall that will be reached.

Once 100% of the power/energy that goes into a computer is used for computations, you can't go further

The good news is, that computer wouldn't require a heatsink or any cooling...

Actually, you could go further. There's always MOAR POWER

[martys1103]

150 FPS is TOO SLOW TO PLAY?! TRY TO PLAY AT 25-30 FPS AT EMBARK! (Stop complaining.)

i play at 10-15 FPS at embark...

[Heavenfall]

I play at 0 FPS at embark, AND THEN IT GETS SLOWER!

[carebear]

As computers will raise in power, DF will raise in complexity.

This needs a name. Like Moore's law.