Jump to content

smiley

Community Members
  • Posts

    212
  • Joined

Everything posted by smiley

  1. Unbelievable he performed so well in those Marvel movies. It's no easy feat even for someone at peak physical fitness. Rest in Power.
  2. The problem is caused by highdpi screens and how recent Windows versions scale them. Since Windows internally tries to handle application compatibility for highdpi screens, it gives scaled resolutions to SDL2, which is actually severely low. This caused the game graphics to be absolutely @#$%, (for instance, the main menu buttons were stacked etc. for me on my screen). I noticed this when I switched from Linux to Windows. Interestingly there were no green glitches at the time. And then there was an update within the year 2020, (I can't be sure if it was Intel or Windows, since I did both at the same time), that resulted in this atrocity. I have reason to believe the actual glitches come from within Intel drivers because the game resolution is so much lower than the native one. I spent like two whole days trying to figure this out. I really wish I had seen this thread earlier, this information would have really helped me when I was randomly reading SDL2 docs and ancient SDL forum threads. There are 2 solutions, 1) Specify exclusive fullscreen (the oldschool kind literally nothing uses anymore because its drawbacks outweigh the benefits). 2) Specify awareness in the binaries which stops Windows from meddling with the resolution by scaling them or whatever, which means SDL2 can actually know the true resolution and draw the game using the true native resolution. #2 is what MS itself recommends. Unfortunately, both of these solutions require WFG devs and as an end user there really is not much you can do other than wait for A24 which hopefully will have this fix. Or apply this metaphorical english patch yourself I guess.
  3. Not really true. They detect software built without proper versioning, publisher info, or unsigned exes and the like. I mean Blender, GIMP, Firefox are all open source too. The anti virus wouldn't even know what you execute is open source let alone read the source to make sure it's safe. That's an acceptable model for an AntiVirus because you don't really need one if you at all know what you are doing. And, the built-in AV tools are more than enough unless you do something very stupid.
  4. Hypothetically, would this be a desirable feature? idtech engine have this. It's not exactly rocket science to make it. But most importantly, the cost benefit analysis doesn't seem worth it since you lose the JS sandbox.
  5. > What is already done? >> Would there be a way to improve pathfinding performance by using formations? Commands to move and pathfind across a map apply to formations, unit pathfinding would then only be relative to the formation. > And what does this have to do with terrain auras? Technically, nothing. Terrain auras will be rather to simple to implement. > Also can you describe the architectural problems, what they are, and how this relates to the issue of territory auras. Well, considering that it was because of performance problems that this feature you are asking for was removed in the first place, I will say that it is somewhat relevant. For the other question, it's 2020, and the game can't run smoothly on overkill hardware. But yes, aside from unit movement affecting terrain and the consequence upon pathfinding, there really is no relevancy in that post.
  6. Already done. Making the pathfinder work with non uniform grids is a solved problem and various games already have it. I recently read a paper from AAAI 2019 that discussed this very in-depth. Performance problems today are mainly due to some major architectural flaws. Having worked with commercial engines, this one is just not there yet.
  7. Actually, you can't really have unit speed affecting properties. A unit will choose to walk through 2 feets of mud a foot away from a literal road in the current pathfinding scheme. Jump point search is based on the assumption of uniform cost grids. This feature was present in the past before it was decided to nuke it in favour of better performance.
  8. GPL is not a free license. Unless 0AD is relicensed under a permissive license, there is no way to use Steamworks SDK. STK is apparently using Steamworks SDK in a separate closed source process and communicating with said process as a middle man. This opens up only the most basic of steam functionality. GPL is a rather viral license in that it tries to attach itself to things it should not be. "Keep in mind that according to the Steam Distribution Agreement you warrant and represent that you have all necessary rights to distribute the game via Steam. If your application contains third party open source code that is incompatible with the Steamworks SDK, then YOU MUST NOT DISTRIBUTE YOUR APPLICATION VIA STEAM." "But I saw a GPL-licensed application on Steam! This can happen if the author of the code that is GPL-licensed has given the permission to do so. The author can of course always (a) decide to grant Valve a different license than the author grants everyone else or (b) decide that what the Steamworks SDK does is just a communication with a service that does not invoke the copyleft requirement of the GPL." Also, consumer protection laws can come in. This means two things, Wildfire Games should be a registered company (Hopefully an LLC), and Wildfire Games can potentially be open to law suits. Seems like a no win situation for the benefit achieved by a donate button.
  9. White peacocks are almost always the result of leucism. And since it's a recessive gene, there is like no chance of encountering one of these in the wild. A random mutation will surely get killed, and even if not, will probably not pass on the gene. Leucism almost always occurs in captivity, so, i would assume recent. 1/100 is still too frequent. 0/100 is preferable in my point of view to be honest. (this is regarding that peacock discussion from last month btw)
  10. Maybe build artifacts? The SM binary distributed with 0AD is not that large as well. Somewhat matches Firefox sizes.
  11. That's actually a pretty good concept. Rather unique as well. Even worth exploring. Any concrete plans on this?
  12. Huh. I see. I assumed it was like the Intel integrated graphics.
  13. Integrated graphics will still be using the VRAM no? Aside from the onboarding buffer (which your dedicated card will also use)? I don't think the current ram bus allows to not tank GPU performance while using system ram. You should be fine as long as you are not buying an Intel and it can push 4ghz on a single core.
  14. Can you abstract it a bit and make it part of rmgen/math.js? And submit a patch of course. (code.wildfiregames.com is fine by me). It can be reviewed more easily and committed (at least to the one I commit to).
  15. Get used to it dude. You are gonna have to go through the same thing when you finish your thesis. Literally everyone in academia is skeptical of everything. Assume it's wrong until proven correct.
  16. General comment on the flood fill algorithm. When testing reachability, there is no reason to not stop once a connection has been established. Edit: disregard the comments below. You are already using a stack it seems. ~~Keeping that in mind, for most cases, using a stack (the data structure) might lead to a connection being found faster than if you were using a queue. If you visualise your algorithm, you would see the fill expanding like a diamond (actually a square oriented diagonally). If you were using a stack, the fill would be expanding line by line. Vertical lines going across the whole area. The probability of hitting the target is much higher. Especially if the target is an area rather than a point. Basically, last in first out instead of first in fast out. But implementing a stack is more complex than a queue.~~
  17. density_placer.js Submit it to upstream if you think that's worth having there. I am not sure I have much use for that now.
  18. You don't really need the pathfinder to do that. You just need to flood fill for non mountain tileclasses and see if your two points are connected. But you have to use the underlying storage of tileclasses for optimal performance. But this should not be that slow since the N in O(N^2) would remain in acceptable values. For nicer mountains, I suggest using fractal generation. More specifically, subdivision generation. The concept is rather simple. Split into half, add height, repeat. The only issue might be that there just aren't enough vertices in 0ad terrain. I might have some code to throw your way which I used to generate some screenshot worthy maps with regards to more natural looking forests.Will post if I managed to find it.
  19. What about PBR? I am not sure if it's really worth it. In RTS games, you can't really see the details. And more simpler code from certain points of view. But most importantly, if a gltf migration is planned, that's something to think about since it's kinda centered on PBR.
  20. > What is nowadays the possibility to have lightning / storms during a match? >> Not possible yet. @vladislavbelov worked on something but didn't release anything usable. I guess I should have quoted the relevant part. That was my bad. But yes, any plans in the graphics area of stuff going on? If I may suggest something, do what STK did and get the graphics engine of open source code and maintain it as a fork.
  21. With the hardware you want to support, the lack of deferred shading (or just forward+), the use of alpha preventing a reasonable deferred rendering system, and all around archaic rendering system compared to today I am really not sure if dynamic lighting "works". I guess completely ignoring shadows casted would make it simpler. Aproper implementation of dynamic lighting is pretty straightforward. But in this 20 year old system, it's just too much work. Without instancing, dynamic lights in a forward system will fry cards. So many fragments per pass. And it scales with number of lights too.
  22. Not honoring this request within 30 days is literally illegal per GDPR laws. A full name is personal data and the subject is within his right to erase it. The nickname is also identifiable information and he is within right to rectify it. "[information must be provided to the individual without delay, and at the latest, within one month of receipt of the request.]" Meaning that Wildfire Games is already in violation of EU GDPR since this request was filed early May. Honestly very disappointing that this was not taken care of.
  23. [Something about being mandated by law to release source and well wishes to anyone trying to merge stuff]
  24. Not necessarily. Since the kernel over commits memory, there are specific circumstances where it might not lag at all. OOM kill is the best case scenario. Without logs, cant really say I suppose.
  25. 0AD most likely starved system memory plus swap so the Kernel sent a Kill(9) signal to the process terminating it. And yes, the Kernel only does this in the most dire of cases. One of these three should help figure out why. The first one is the most likely. I am not sure to be honest. dmesg -T| grep -E -i -B100 'killed process' /var/log/kern.log /var/log/messages Good luck fixing that.
×
×
  • Create New...