[Master Fibbles]

Mr.Flibble, on Jul 9 2010, 06:41 PM, said:

Using the July 6 build I can't get Mapster to run. It loads a blank screen and when I try to load something in 2D mode, it crashes.

I tried a fresh "install" and it works. I'm not sure what is causing the crash though. Whatever it was, it had to do with my newboard.map. That map was just a single room with a spotlight in a corner.

[RPD Guy]

updates?
^^

[Night Wolf]

Did the spark texture recently get lighting hardcoded to it?
I never noticed before untill now ....
looks good.

[TerminX]

Yeah. It was probably a couple of months ago.

[Night Wolf]

TX, on Jul 16 2010, 06:28 AM, said:

Yeah. It was probably a couple of months ago.

Hmm strange I only just noticed.
......
Any chance of the "red siren lights" being hardcoded?
maybe a twirling light source if thats even possible...

[TerminX]

Maybe. I thought about it at one point and I think I decided against it because the light would need to be a spotlight (more resource intensive) and it would be horrible in usermaps where you have a bunch of those sprites going around a helipad or something like that.

[Danukem]

TX, on Jul 16 2010, 09:06 AM, said:

Maybe. I thought about it at one point and I think I decided against it because the light would need to be a spotlight (more resource intensive) and it would be horrible in usermaps where you have a bunch of those sprites going around a helipad or something like that.

I agree, and I think that hardcoded lights should be kept to a minimum. In addition to the performance issue, there's the fact that mappers didn't consider many sprites to be light sources. Take switches, for example. Sometimes switches that are supposed to be hidden give themselves away in Polymer because of the light they give off. And there is another issue. In a TC, there may be a completely different tile of that number which the light doesn't fit, and it just adds one more annoying hardcoded thing for the scripter to deal with.

[Night Wolf]

TX, on Jul 17 2010, 02:06 AM, said:

Maybe. I thought about it at one point and I think I decided against it because the light would need to be a spotlight (more resource intensive) and it would be horrible in usermaps where you have a bunch of those sprites going around a helipad or something like that.

Hmmm I can understand that... maybe when eduke is more stable.

[Hendricks266]

DeeperThought, on Jul 16 2010, 11:51 AM, said:

In a TC, there may be a completely different tile of that number which the light doesn't fit, and it just adds one more annoying hardcoded thing for the scripter to deal with.

I've had to deal with that in Caribbean. Fortunately TX did add ways to disable the lights from CON.
Now I have to check that I disabled SHOTSPARK1's lighting... phew.

[High Treason]

I've had an idea, how difficult would it be to add a feature to Mapster32 where point lights show a circle on the 2D View like ambient sounds? This would be really useful. Just an idea.

[Micky C]

I always thought it'd be nice if there was some way to see what colour polymer lights are in 2D mode, the only way to tell at the moment is to switch into 3D mode, which usually takes a few seconds, and it can be a bit time consuming if you have to make constant corrections to get the colour right. Although High Treason's idea would be more practical and easier to implement.

[RPD Guy]

How do I fix a light, or spotlight on some object?

[The Commander]

Dk2, on Jul 26 2010, 05:27 AM, said:

How do I fix a light, or spotlight on some object?

You are going to have to explain that in more detail.

[Sobek]

I'm guessing he wants to 'attach' a light to an object (such as a sprite or model) in a way that will make the light move with that object.

I think you'd just need a bit of CON code to make that object cast a polymer light, basically. I'm sure someone else can properly expand on the idea.

[TerminX]

The first synthesis build with improved Polymer performance went up today. It also fixes problems with broken doors and sector effects in Duke Caribbean and some other maps.

[SwissCm]

The framerate increase is quite significant. If any of you were struggling to run Polymer before it's worth trying out again.

[0815Jack]

SwissCm, on Jul 27 2010, 07:10 AM, said:

The framerate increase is quite significant. If any of you were struggling to run Polymer before it's worth trying out again.

well i cannot confirm this.....i know its difficult to compare, but thats how i compare 1674 & 1676

i loaded E1L1 with latest SVN polymer HRP @ 1680*1050 (AF = none) and waited till the space ship exploded and doesnt move on the roof

1674: 28fps
1676: 12fps

my system: C2D-8400 / 4 GB / HD4350 (Catalyst 10.4) on XP32

[Danukem]

0815Jack, on Jul 27 2010, 06:51 AM, said:

well i cannot confirm this.....i know its difficult to compare, but thats how i compare 1674 & 1676

i loaded E1L1 with latest SVN polymer HRP @ 1680*1050 (AF = none) and waited till the space ship exploded and doesnt move on the roof

1674: 28fps
1676: 12fps

my system: C2D-8400 / 4 GB / HD4350 (Catalyst 10.4) on XP32

That's a dramatic reduction in frame rate. I would give it another try or two comparing 1674 and 1676. Maybe your computer was doing something intensive in the background during the 1676 test?

[MusicallyInspired]

Just tried it. Hmm, a bit strange. I tried E1L1 with the Polymer HRP. Even though the framerate on the roof says it's 90-100 it's very jumpy and skippy. Once down in the street this strange phenomenon disappears, though, (along with the improved framerate) until you go back up. Performance seems to vary from area to area moreso than before. Some bigger areas are worse than they were before and some isolated areas are silky smooth when they weren't before. All in all it's not quite as playable as it was before as far as gameplay is concerned. It takes the game a bit of time to come to grips with new textures and I lose framerate altogether for a bit until it figures itself out and then it's partially playable.

Then the game crashed when it attempted to load the second level. Before the texture caching. Whereas before the game would just close (no error message) after the textures were cached on random levels.

Still there was definitely an improvement in certain areas. Optimization seems to be coming along nicely.

My specs:

WinXP SP3
P4 2.8 Ghz
3GB DDR RAM
ATI X1650 AGP

I tried posting my eduke32.log but it wouldn't show. There were a lot of texture "GL_FALSE!" errors, though.

EDIT: I just deleted my texture cache files and tried starting up the game again and it crashes on startup now. Can't even start the game.

EDIT 2: Ok, now it works again. Bizarre.

EDIT 3: Apparently some of my framerate issues are gone this time around. I forgot that I was downloading a pretty hefty file during my first run. Now it seems much more manageable. I still have quite a bit of framerate issues where more than a couple lights are involved. I also no longer get lighting from objects such as switches and atomic health.

Also, here's some screenshots of the bizarre shadow effect spotlights have for me:

http://brandonblume....es/duke0000.jpg
http://brandonblume....es/duke0001.jpg
http://brandonblume....es/duke0002.jpg

[Danukem]

MusicallyInspired, on Jul 27 2010, 09:21 AM, said:

EDIT 3: Apparently some of my framerate issues are gone this time around. I forgot that I was downloading a pretty hefty file during my first run.

No offense, but that undermines your whole report, and I was wondering the same thing about 0815Jack's report.

[MusicallyInspired]

Maybe, but I later confirmed everything about my report afterwards.

[Danukem]

MusicallyInspired, on Jul 27 2010, 11:46 AM, said:

Maybe, but I later confirmed everything about my report afterwards.

Well, nevermind then.

[Plagman]

MusicallyInspired,

is the shadow issue illustrated in your screenshots something recent, or has this always happened? I think your GPU falls into the "polygon offset" workaround where I had to disable a functionality that makes shadow look more correct because of a driver bug that caused a software fallback (causing huge performance hits and sprites showing above everything else, not sure if you remember). I think that's the cause of the shadow problems you're seeing, just wanted to confirm that it was always like that and wasn't anything new with the new version.

[MusicallyInspired]

Plagman, on Jul 27 2010, 02:18 PM, said:

MusicallyInspired,

is the shadow issue illustrated in your screenshots something recent, or has this always happened? I think your GPU falls into the "polygon offset" workaround where I had to disable a functionality that makes shadow look more correct because of a driver bug that caused a software fallback (causing huge performance hits and sprites showing above everything else, not sure if you remember). I think that's the cause of the shadow problems you're seeing, just wanted to confirm that it was always like that and wasn't anything new with the new version.

Yeah, it's been like that for quite a while. As far as I can remember anyway. And I'm the person who had that exact problem you're describing back then. So it's probably my card. Just making sure.

[Mark]

POST DELETED. I will fix my other problem before commenting on the newest version.

[Sobek]

I've been away for a while, but I just started working on some maps again and figured I'd test some particularly nasty sections of some maps (performance-wise) to see if the latest build fared any better on my crappy work PC (which I have since ditched for mapping).

There's definitely a marked improvement. It no longer randomly crashes when switching between 2d/3d with the Polymer renderer active (I normally turn it on and off as needed but it would be nice to not have to), and the overall framerate in some particularly bogged down areas has improved in the order of maybe 20 or 30%... One spot in particular was running at around 15fps with polymer active in Mapster, but now sits at a good 30fps or so the whole time with nowhere near as much random pausing or stuttering.

This is a pretty tame PC by today's standards though - it's just a single core A64 3200+ with an HD3870 and 1gb of ram.

[Mark]

I made the jump from version 1661 to 1676. I did a clean install.
There is about a 10 FPS increase in many parts of the regular duke levels in Eduke32 with this new version. And no other bugs found yet But no increase in the large outdoor areas of my W.I.P. map that could really use a boost

I am having problems with Mapster in this new version. In the Polymer mode sprites are showing up green all over the place.
Second problem is Mapster quits and returns back to my desktop every time I try to load any second map if polymer mode is enabled.



I'll post later with more details and my system info and log file in the other area of the forum for bugs and problems.

[TerminX]

Heh, I actually saw the green sprite thing at some point earlier but didn't get around to figuring out the cause yet. See if you can determine which build started the problem.

[Mark]

I just downloaded and ran the last 4 versions. 1674 and before, no green sprites. 1676, green sprites.

There was no 1675 listed to try out.

[Skulldog]

Had a glich with r1676. On e1l1 I jump off the building in the air duct. and the duct was sealed up at the end, I was
traped inside the air duct. I turn on god mode for jet pack, flew to the top and it was sealed closed too. Restarted
the level and then it was fine.