P.O.P.S.- Demo 3.1 available!
Moderator: Moderators
-
Master-Athmos
- Posts: 916
- Joined: 27 Jun 2009, 01:32
Re: P.O.P.S.- Demo 3 available!
Burns just fine here...
- Attachments
-
- Fire.jpg
- (40.38 KiB) Downloaded 29 times
Re: P.O.P.S.- Demo 3 available!
That's strange, Forb. I'll compare the last working version with this one, see what's up. IIRC, it was some minor difference in the order of the OpenGL setup instructions.
Re: P.O.P.S.- Demo 3 available!
OK, here's version 2. Please test it again, Forb, it's mysterious that this isn't working, but this matches version 22 of the Bubbles, in terms of the order of instructions. Hopefully that will resolve the issues (other than whatever I need to do to improve CPU efficiency- it's quite a bit faster than CEG, but is still slower than I'd like).
- Attachments
-
- FIRE_POPS2.sdz
- (3.74 MiB) Downloaded 14 times
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: P.O.P.S.- Demo 3.1 available!
No go. Am I supposed to be doing something besides starting the game with ctrl+enter?
I don't hear any sound effects either, not sure if that makes a difference. It just strikes me as really really odd that, something would work on ati and not work on nvidia.
I don't hear any sound effects either, not sure if that makes a difference. It just strikes me as really really odd that, something would work on ati and not work on nvidia.
Re: P.O.P.S.- Demo 3.1 available!
That is bizarre. What should happen is that the game should start, it should move to the Commander, and you should see the demo flames.
Really puzzled by this. Do you get any errors in the infolog? I've run this on the 7800GS and on the 9800GT at my folks'- it's been fine both places. I take it that the bubbles failed, too? Are you using custom drivers? Do you have other graphics problems? Just trying to get a feel for this- I use stock nVidia drivers, nHancer on the 7800GS to tweak a few things for best performance, but basically stock settings. Really weird, that it's running on ATi and on two nVidia cards, but not your hardware.
Really puzzled by this. Do you get any errors in the infolog? I've run this on the 7800GS and on the 9800GT at my folks'- it's been fine both places. I take it that the bubbles failed, too? Are you using custom drivers? Do you have other graphics problems? Just trying to get a feel for this- I use stock nVidia drivers, nHancer on the 7800GS to tweak a few things for best performance, but basically stock settings. Really weird, that it's running on ATi and on two nVidia cards, but not your hardware.
Re: P.O.P.S.- Demo 3.1 available!
hmm argh I find your fire and smoke effects start off looking beuatiful, and you post shots all over the place, then you do something to them and release and they dont look anywhere near as nice.
Fore xample remember when you first put up shots of the missiles in P.U.R.E:
I thought that looked prettttty, then you went and changed it before release and ti wasnt anywhere near as good as it lost all its contrast and colour.
Fore xample remember when you first put up shots of the missiles in P.U.R.E:
I thought that looked prettttty, then you went and changed it before release and ti wasnt anywhere near as good as it lost all its contrast and colour.
Re: P.O.P.S.- Demo 3.1 available!
I had to change it because CEG's performance was so slow, AF. It was a literally a "hey that's a nice slide show" kinda thing, if there was much else going on, and that's unacceptable, even though I know my CPU's fairly slow nowadays. I've found, pretty consistently, that anything that brings the old AMD XP to a halt doesn't perform *that* much better on newer chipsets. There are big differences in raw GPU power, but the total horsepower per CPU... not really. Now, MT Spring, that's another animal entirely, but it's not stable yet, so it's an issue where I have to be conservative and deal with what is probably going to work. And, of course, I can't test it MT until I have a machine with multiple cores to test with on a regular basis.
The goal is to be able to do that sort of thing, but meet the performance requirements. That system was throwing out new particles in the same range this demo is.
Right now, with (IIRC) 24 projectiles (gotta count the other Commander, they're both doing that) launching 3 events per gameframe, so 72 per gameframe, or 2160 new particle events per game second. IOW, it is writing almost half of stock Spring's particle LIMIT, with a total LIMIT more than 3X higher.
Note that now that CEG obeys a hard-coded limit, it really hasn't helped a lot with total load from CEG for me, because (at least in the case of P.U.R.E.) so many CEG events have short lifetimes anyhow. That's part of the beauty of this thing, imo- lifetime doesn't matter. Depthtests are a much more serious performance hurdle, although one that's very hard to avoid without taking unreasonable liberties with beauty.
Anyhow, it's performance is reasonable, but I think I can get it higher, maybe by writing to a table on the sync side or by finding further optimizations to the write stages, or other stuff. IDK yet. The target's somewhere like 30-35 projectiles (iow, about what you'd see in a big battle with a lot of smoke-spewing stuff, bits of buildings flying off, etc., fires that are continuous loops, etc.)... I think that I may write a limiter, and if > X are active, then I just won't let them register new events, to save people's CPUs.
IDK about all of that stuff yet, that's all considerably ahead of where things stand atm, but I am fairly certain that my current loop is fail and that I am wasting a lot of cycles somewhere, probably at the transfer stage, but there may be a bit more efficiency that I can wring out of the main loop, too- but most of the real inefficiency there is GPU-side atm, unless I really missed something obvious. I just don't know yet.
First though, I need to figure out what is wrong when it encounters Forb's hardware, because that's really weird that it's broken NOW, when it works here, at my folk's place, it works at home, and it works on Master-Athmos's ATi.
It's really weird, but I really must get it reasonably bug-free before I worry too much about fancy applications, otherwise, once again, I may see 20-50% of the userbase going, "wtf, I can't see the fancy explosions u promised".
One thought is that perhaps it's the texture? I'll compare with version 22, maybe there's something odd going on. There has to be some sort of pattern, though. If it's not the fixed-function setup, then it's gotta be something dumb like that.
The goal is to be able to do that sort of thing, but meet the performance requirements. That system was throwing out new particles in the same range this demo is.
Right now, with (IIRC) 24 projectiles (gotta count the other Commander, they're both doing that) launching 3 events per gameframe, so 72 per gameframe, or 2160 new particle events per game second. IOW, it is writing almost half of stock Spring's particle LIMIT, with a total LIMIT more than 3X higher.
Note that now that CEG obeys a hard-coded limit, it really hasn't helped a lot with total load from CEG for me, because (at least in the case of P.U.R.E.) so many CEG events have short lifetimes anyhow. That's part of the beauty of this thing, imo- lifetime doesn't matter. Depthtests are a much more serious performance hurdle, although one that's very hard to avoid without taking unreasonable liberties with beauty.
Anyhow, it's performance is reasonable, but I think I can get it higher, maybe by writing to a table on the sync side or by finding further optimizations to the write stages, or other stuff. IDK yet. The target's somewhere like 30-35 projectiles (iow, about what you'd see in a big battle with a lot of smoke-spewing stuff, bits of buildings flying off, etc., fires that are continuous loops, etc.)... I think that I may write a limiter, and if > X are active, then I just won't let them register new events, to save people's CPUs.
IDK about all of that stuff yet, that's all considerably ahead of where things stand atm, but I am fairly certain that my current loop is fail and that I am wasting a lot of cycles somewhere, probably at the transfer stage, but there may be a bit more efficiency that I can wring out of the main loop, too- but most of the real inefficiency there is GPU-side atm, unless I really missed something obvious. I just don't know yet.
First though, I need to figure out what is wrong when it encounters Forb's hardware, because that's really weird that it's broken NOW, when it works here, at my folk's place, it works at home, and it works on Master-Athmos's ATi.
It's really weird, but I really must get it reasonably bug-free before I worry too much about fancy applications, otherwise, once again, I may see 20-50% of the userbase going, "wtf, I can't see the fancy explosions u promised".
One thought is that perhaps it's the texture? I'll compare with version 22, maybe there's something odd going on. There has to be some sort of pattern, though. If it's not the fixed-function setup, then it's gotta be something dumb like that.
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: P.O.P.S.- Demo 3.1 available!
AF, that screen has thousands and thousands of particles in use... Huge OMGWTF?!?!!? nono.
@argh, infolog is normal. Starts and exits correctly without errors.
@argh, infolog is normal. Starts and exits correctly without errors.
Re: P.O.P.S.- Demo 3.1 available!
OK.
One big difference was found between this and Bubble no. 22, that I overlooked: one version used TGA, one used DDS. Maybe some cards / DeVIL / OS still aren't handling DDS correctly.
Sooo... here's a version with PNG, which has been reliable in most Widgets that do textures in Spring. I can try TGA too. Please let me know if it works.
One big difference was found between this and Bubble no. 22, that I overlooked: one version used TGA, one used DDS. Maybe some cards / DeVIL / OS still aren't handling DDS correctly.
Sooo... here's a version with PNG, which has been reliable in most Widgets that do textures in Spring. I can try TGA too. Please let me know if it works.
- Attachments
-
- FIRE_POPS3.sdz
- (3.81 MiB) Downloaded 16 times
-
Master-Athmos
- Posts: 916
- Joined: 27 Jun 2009, 01:32
Re: P.O.P.S.- Demo 3.1 available!
It's still burning here... 
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: P.O.P.S.- Demo 3.1 available!
Still nothing 
I doubt this will help, but it shows that it "Should" be working correctly.
I doubt this will help, but it shows that it "Should" be working correctly.
Re: P.O.P.S.- Demo 3.1 available!
Yes, that is distressing. Nothing happens at all, eh?
Hrmm. I used PNG, which is pretty much compatible with everybody using DevIL, DDS, which is supposed to work... the shader compiled, no errors... so, back to basics- it's failing either on the shader end, or on the setup end. I'll go stare at #22 s'more, there has to be something else I've missed here.
Hrmm. I used PNG, which is pretty much compatible with everybody using DevIL, DDS, which is supposed to work... the shader compiled, no errors... so, back to basics- it's failing either on the shader end, or on the setup end. I'll go stare at #22 s'more, there has to be something else I've missed here.
Re: P.O.P.S.- Demo 3.1 available!
or you could try TGA
Re: P.O.P.S.- Demo 3.1 available!
I am going to, after I get done examining the other stuff.
I switched back and forth between TGA and DDS several times, and #22 was the only one Forb reported working, so IDK if that's a factor or not, but I kinda doubt it. Problem is, there's so little to go on.
I switched back and forth between TGA and DDS several times, and #22 was the only one Forb reported working, so IDK if that's a factor or not, but I kinda doubt it. Problem is, there's so little to go on.
Re: P.O.P.S.- Demo 3.1 available!
OK, here's version 4.
If this doesn't work for Forb, it's back to basics- the shader could be doing something wonky that doesn't result in an error otherwise. Whatever the issue is, it's something strange.
If this doesn't work for Forb, it's back to basics- the shader could be doing something wonky that doesn't result in an error otherwise. Whatever the issue is, it's something strange.
- Attachments
-
- FIRE_POPS4.sdz
- (3.88 MiB) Downloaded 23 times
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: P.O.P.S.- Demo 3.1 available!
-
Master-Athmos
- Posts: 916
- Joined: 27 Jun 2009, 01:32
Re: P.O.P.S.- Demo 3.1 available!
Still fine at this end...
@Forb:
Do you see the "balls" at all?
Do you see the "balls" at all?
Re: P.O.P.S.- Demo 3.1 available!
Well, IDK Forb. The action's all right around the Commander- that's where you want to be looking.
At the very least, you should see a noticeable framerate dip (if you aren't running vsynced) when viewing the Commander (it'll be smaller, on new hardware, but it's definitely going to be noticeable, this last version is using a lot of fillrate).
If you aren't seeing anything, but you have a framerate hit, then I at least know that it's trying to operate, and that the particles are "there", as opposed to off near the world origin, etc. I have accidentally made that occur before.
One of the things you should try is maybe include a little Gadget I wrote that hides the map (for other experiments that I haven't finished yet). A copy of it is included here. This might show the particles at the world origin (0,0,0 or the upper-left corner of the map).
But chances are, the problem's something more fundamental than that. It's very weird, that out of all of those versions, just one worked. I assume you see the normalmapped sphere, the light show, etc.?
For those of you who haven't downloaded it yet, or checked out Master-Athmos's shot, here's a screen of about what it looks like:
At the very least, you should see a noticeable framerate dip (if you aren't running vsynced) when viewing the Commander (it'll be smaller, on new hardware, but it's definitely going to be noticeable, this last version is using a lot of fillrate).
If you aren't seeing anything, but you have a framerate hit, then I at least know that it's trying to operate, and that the particles are "there", as opposed to off near the world origin, etc. I have accidentally made that occur before.
One of the things you should try is maybe include a little Gadget I wrote that hides the map (for other experiments that I haven't finished yet). A copy of it is included here. This might show the particles at the world origin (0,0,0 or the upper-left corner of the map).
But chances are, the problem's something more fundamental than that. It's very weird, that out of all of those versions, just one worked. I assume you see the normalmapped sphere, the light show, etc.?
For those of you who haven't downloaded it yet, or checked out Master-Athmos's shot, here's a screen of about what it looks like:
- Attachments
-
- flame_test.jpg
- (28.19 KiB) Downloaded 40 times
-
- HideMap.lua
- (569 Bytes) Downloaded 10 times
Last edited by Argh on 10 Nov 2009, 19:22, edited 1 time in total.
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: P.O.P.S.- Demo 3.1 available!
O_O Balls?Master-Athmos wrote:Still fine at this end... @Forb:
Do you see the "balls" at all?
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
Re: P.O.P.S.- Demo 3.1 available!
OH FOR FUCKS SAKE! :_(
I am sorry argh, fire 1.0 works fine. I didn't realize that I needed to start it from the lobby as I saw bubbles from just running spring.exe so I assumed that it would be the same way with fire... Fuck damnit dude, I'm so sorry! :_(
Suggestion; idiot proof instructions for retards like myself.
I am sorry argh, fire 1.0 works fine. I didn't realize that I needed to start it from the lobby as I saw bubbles from just running spring.exe so I assumed that it would be the same way with fire... Fuck damnit dude, I'm so sorry! :_(