Bay 12 Games Forum

So you're saying that the part where speed matters, manipulating the actual wireframes, is done in assembly? The parts where speed doesn't matter is in C++? Color me surprised. Why don't they write those parts in C++? That is because it would be slow as shit.

Just the implementation of the matrix multiply etc functions. Functions where using specific assembly instructions that aren't available as normal c++ operators (e.g. SSE) is beneficial. And even then I'm sure I saw compiler vector intrinsics instead of assembly being used for some of the platforms, I'll check on Monday.
And you're sadly mistaken if you think speed doesn't matter anywhere else...

And for the record, the matrix ops aren't used for "manipulating the wireframes". Transformation of mesh vertices is much better handled by the GPU.

Incidentally, if you're wondering whether it'd be *possible* to write DF in assembly (regardless of any performance advantages it may bring), note that RollerCoaster Tycoon (http://en.wikipedia.org/wiki/RollerCoaster_Tycoon) was written 99% in assembly, with some C bindings to let the game interface with Windows.

Given that it's a sandbox game that gives the player considerable freedom to build and experiment just like DF, that's... incredible.

Based on the transport tycoon engine, though heavily modified. Both games are a huge favourite of mine.
There's a reason the openttd guys converted transport tycoon to c/c++ though...

Oh dear that site has no caching at all. It's quite emphatic about it in the HTTP headers.

At least putting ' in the user/pass fields didn't make it explode.

[It's a volcano on top of a glacier, go for it.]

This has given me new perspective and respect for the country of Iceland!

Between their choice of embark (Iceland was settled by the beardy Vikings, right?) their unpronouncable names, and their tendency to stay inside and drink, I'm pretty sure they are Dwarves.

They also eat pretty much anything (purified shark! Puffin!) and have a spirit (as in drink) nicknamed "Black Death".

EDIT: They also use geothermal power for everything, including heating. That's like taming lava, right?

Before you look at this, allow me to emphasize that
- You can change the font.
- This is still glitchy. The real thing might look better in many ways.
- If you don't like it at all, you can turn it off and get the old behaviour.

That said, a sneak peek:

Spoiler (click to show/hide)

Wow

You may not be able to tell, but for once I'm not being sarcastic.

Quote from: NerfJihad on August 31, 2010, 06:01:37 am

Kudos to Toady on this project, it's an incredible testament of coding prowess to brute-force a core-i7 (8 cores @ 2.0ghz) machine with 8gb of ram into a slow, wheezing demise due to the number of things running simultaneously. Adventurer mode is the only way I've managed to overload my machine. Then again, I wanted to play with actual rivers that were wide and deep and awesome to behold, so I made 'large regions' with 100's of z-levels of depth, so I guess i get what I pay for.

There's your problem. That's barely 25% stronger than my 5-year old laptop. DF cares about sheer processor strength as opposed to number of cores.

Wrong. CPU speed stopped going up years ago, ~2.0-3.0GHz is normal. What's been going up is instruction processing efficiency, cache size, (integrated) memory controller speed and number of cores. The last doesn't affect DF, but the rest all do. His i7 will get much much much more performance in DF than your laptop.

Quote from: macdonellba on August 24, 2010, 07:35:51 pm

Quote from: zagibu on August 24, 2010, 04:00:48 pm

...for AMD announced an architecture capable of running a single thread on multiple cpu cores: http://gizmodo.com/5620423/amd-announces-8+core-bulldozer-cpu

The mapping is actually one software thread of execution to each one integer core (ie. no magical one thread to multiple cores mapping); the novel design feature is that several integer cores share what would normally be redundant circuitry to cut down on power usage, increase cache coherency in some situations (ex. a VM allocated two physical cores could be allocated two sharing an L1 cache), etc.

You are right, my statement was a bit misleading. Actually, the thread is still running on one core only, but this core can use all of the shared resources. So it's not really 6x the performance, more like 150% or something. Still a step in the right direction.

Nah, you actually get nothing like that. Basically, each module is part-way between a full dual-core and a hyperthreaded single-core, in that there are some shared cpu resources (hyperthreading style) and some duplicated resources (dual-core style). AMD thinks that this gets a good balance between die space usage and two-thread performance, allowing them to get nearly the performance of a true 8-core cpu in a smaller die space. The only thing that could even remotely improve performance of a single-threaded application is the dual-FPU, and it would have to be a really floating-point-heavy application for that to make a difference.

I'm not expecting DF to benefit from this CPU at all.

I caught 2 dwarves trapped in tiny nooks by sudden tree growth. Had to get miners in to dig them out because the entire map's worth of trees was designated for chopping, and I didn't want to undesignate it to get a rapid response woodcutting team in there.

Use a burrow. Define one around the guys and assign a woodcutter to it temporarily.