Jump to content



Photo
* * * * * 12 votes

Dev thread: Road to DX9


  • Please log in to reply
2087 replies to this topic

#861 The Loafer

The Loafer

    Pinball Wizard

  • VIP
  • 3,471 posts
  • Location:Embrun, Ontario, Canada

  • Flag: ---------

  • Favorite Pinball: Superman, Firepower & Tron



Posted 28 February 2014 - 03:55 PM

Windows vol control

#862 bodydump

bodydump

    down and out

  • VIP
  • 578 posts
  • Location:California

  • Flag: United States of America

  • Favorite Pinball: High Speed, Creature

Posted 28 February 2014 - 04:25 PM

Version 6 fixed the ramp lip problem in World Cup Soccer but there are still some other issues.  In the screen shots below the edges of the ramp sticker "pass" are much worse looking in dx9 than in 9.20. This object is made from a solid ramp and is 1 unit above the alpha ramp (actual ramp).  The second thing is the sign above the ramp entrance.  The edges of it also look much worse in dx9 than in 9.20.  It is also a solid ramp.  Both ramps have images that are png's with a transparent background using 128,128,128 as the transparency color (true alpha).

Win7, I7, GTX660 3gb

 

DX9

dx9.jpg

 

9.20

920.jpg


Edited by bodydump, 28 February 2014 - 04:48 PM.


#863 Swisslizard

Swisslizard

    DOF inventor & coder

  • VIP
  • 152 posts

  • Flag: Switzerland

  • Favorite Pinball: The Machine

Posted 28 February 2014 - 09:32 PM

Hi

 

The hint by DJRobX about the right link for the DirectX installation did fix my problem about the missing DirectX dll. Just to make sure everybody knows, here is the link again:

http://www.microsoft...ails.aspx?id=35

 

Now, test6 is working fine on my cab. Tried AFM, Getaway; Monster Bash, Medevial Madness. They are all working fine here. I get about 850FPS for Getaway. Ball movement on all tables was absolutely smooth. Phantastic!

DOF is also working nicely (That was to be expected).

 

Regards

 

Tom


Programming is a race between software engineers striving to build  idiot-proof programs, and the universe trying to produce bigger idiots. So far, the universe is winning.


#864 Argo

Argo

    Enthusiast

  • Members
  • PipPipPip
  • 95 posts

  • Flag: ---------

  • Favorite Pinball: Jurassic Park

Posted 28 February 2014 - 10:05 PM

With test6 the AA from nvidia control panel looks even better!



#865 mukuste

mukuste

    Pinball Fan

  • VP Dev Team
  • PipPipPipPip
  • 854 posts

  • Flag: ---------

  • Favorite Pinball: Centaur

Posted 28 February 2014 - 10:22 PM

Version 6 fixed the ramp lip problem in World Cup Soccer but there are still some other issues.  In the screen shots below the edges of the ramp sticker "pass" are much worse looking in dx9 than in 9.20. This object is made from a solid ramp and is 1 unit above the alpha ramp (actual ramp).  The second thing is the sign above the ramp entrance.  The edges of it also look much worse in dx9 than in 9.20.  It is also a solid ramp.  Both ramps have images that are png's with a transparent background using 128,128,128 as the transparency color (true alpha).

 

 

Thanks for the report. I'm pretty sure that's a decal? Decals can't have antialiasing since they are rendered statically (anything that moved under the antialiased pixels wouldn't look right). So there are two possible options: 1) make the borders of the texture sharp by having alpha jump from 0 to 255 instead of the gradual slope (this will make the borders look a bit more pixelated) 2) convert the decal into an alpha ramp (there might be ordering issues until we have the manual depth bias option on alpha ramps).



#866 LoadedWeapon

LoadedWeapon

    The Night Owl..

  • Members
  • PipPipPipPipPip
  • 2,572 posts
  • Location:South Carolina USA

  • Flag: United States of America

  • Favorite Pinball: Star Trek TNG



Posted 28 February 2014 - 10:31 PM

so the next test about ready? :)



#867 mukuste

mukuste

    Pinball Fan

  • VP Dev Team
  • PipPipPipPip
  • 854 posts

  • Flag: ---------

  • Favorite Pinball: Centaur

Posted 28 February 2014 - 10:52 PM

Uh, I don't know. Nothing much of interest has happened between Test6 and now, the only two real points being the fix to the depth buffer problem (which most people won't even notice) and the revert to DInput 7. I keep looking into individual table glitches that people report, but as the DX9 renderer gets closer and closer to being bug-free, most of those turn out to be problems which have to be fixed table-side. So I think a Test7 right now would be pretty underwhelming.

 

Basically at this point I'm waiting for 9.2.1 to release so that we can start polishing this new version up to a real VP release, with whatever missing GUI options and whatnot else still to be done.



#868 arngrim

arngrim

    DJ Force Feedback

  • VIP
  • 2,188 posts
  • Location:Charleroi, Belgium

  • Flag: Belgium

  • Favorite Pinball: Monster bash



Posted 28 February 2014 - 10:55 PM

Bram Stoker Dracula GI8

 

http://www.vpforums....s&showfile=6222

 

vp9 9.23

http://www.youtube.c...h?v=x1-5caIrYxE

 

vp9 dx9

http://www.youtube.c...eature=youtu.be

 

not perfect yet :)



#869 bent98

bent98

    Pinball Fan

  • Members
  • PipPipPipPip
  • 1,077 posts
  • Location:NY

  • Flag: United States of America

  • Favorite Pinball: Roadshow, Haunted House, Safe Cracker

Posted 28 February 2014 - 11:04 PM

A

Uh, I don't know. Nothing much of interest has happened between Test6 and now, the only two real points being the fix to the depth buffer problem (which most people won't even notice) and the revert to DInput 7. I keep looking into individual table glitches that people report, but as the DX9 renderer gets closer and closer to being bug-free, most of those turn out to be problems which have to be fixed table-side. So I think a Test7 right now would be pretty underwhelming.
 
Basically at this point I'm waiting for 9.2.1 to release so that we can start polishing this new version up to a real VP release, with whatever missing GUI options and whatnot else still to be done.


Are text boxes and display reels working?

#870 mukuste

mukuste

    Pinball Fan

  • VP Dev Team
  • PipPipPipPip
  • 854 posts

  • Flag: ---------

  • Favorite Pinball: Centaur

Posted 28 February 2014 - 11:05 PM

 

Help me out here... I can't spot the difference.



#871 LoadedWeapon

LoadedWeapon

    The Night Owl..

  • Members
  • PipPipPipPipPip
  • 2,572 posts
  • Location:South Carolina USA

  • Flag: United States of America

  • Favorite Pinball: Star Trek TNG



Posted 28 February 2014 - 11:05 PM

one question?

I understand the lighting problems on Rudy but why dont he move or even open his eyes? can i fix it in the editor?



#872 arngrim

arngrim

    DJ Force Feedback

  • VIP
  • 2,188 posts
  • Location:Charleroi, Belgium

  • Flag: Belgium

  • Favorite Pinball: Monster bash



Posted 28 February 2014 - 11:10 PM

you can see lights off on dx9 when on 9.2 they are lit but not blinking.



#873 mukuste

mukuste

    Pinball Fan

  • VP Dev Team
  • PipPipPipPip
  • 854 posts

  • Flag: ---------

  • Favorite Pinball: Centaur

Posted 28 February 2014 - 11:17 PM

A

Uh, I don't know. Nothing much of interest has happened between Test6 and now, the only two real points being the fix to the depth buffer problem (which most people won't even notice) and the revert to DInput 7. I keep looking into individual table glitches that people report, but as the DX9 renderer gets closer and closer to being bug-free, most of those turn out to be problems which have to be fixed table-side. So I think a Test7 right now would be pretty underwhelming.
 
Basically at this point I'm waiting for 9.2.1 to release so that we can start polishing this new version up to a real VP release, with whatever missing GUI options and whatnot else still to be done.


Are text boxes and display reels working?

 

 

No, they are not... you got me there. I was keeping them for the "polishing up" phase after 9.2.1 since I didn't consider them very high priority, and they are kind of annoying to fix.


you can see lights off on dx9 when on 9.2 they are lit but not blinking.

 

Ah, sorry. I was looking at my own current build, and it doesn't happen here anymore. So I guess another thing that got fixed with the depth buffer fix.

 

That table also is another one where the GI textures should have their transparency set to black, so you can try that. Weirdly, it doesn't seem to be necessary in my build.



#874 jimmyfingers

jimmyfingers

    Pinball Fan

  • VIP
  • 832 posts

  • Flag: Canada

  • Favorite Pinball: Comet



Posted 28 February 2014 - 11:31 PM

 

 

Help me out here... I can't spot the difference.

 

Insert lights are getting overwritten by GI updates due to the lights on the .2 ramp object not being correctly drawn above the GI lights table level (i.e. the whole depth bias issue again). The lights on the table GI images (8 of them) were originally used with pure black which is why you can see them turn black despite the fact that the insert lights / objects are on pedestals, if you will, and should not be overwritten. This is not the same as the other black insert issue as in this table I'm not relying on the pure black for the technique of allowing the lights to update through the GI and instead upgraded to the better method of placing the lights above the PF on their own surface (in this case a table sized ramp made up of just the lights cutout - easier than a wall for each light). It's at .2 and failing because test6 is not working correctly with near table PF objects.

 

This will also be occurring on other tables but people may not be noticing this, however, table lights will get overwritten and looked potentially in an "off" state when they might actually be on when a GI event takes place and the light has not changed. In the video that angrim provided you can see blacking out of the insert lights as the GI images have black where the insert lights are but in other tables the inserts will likely still be drawn normally but the state of them in the GI image overwriting the lamp image will cause at least blinking that shouldn't be present and, as I say, in other cases the light will be redrawn and potentially / effectively turned off so you don't know what you have lit or have not.

 

People are not noticing some of these issues but they indeed exist and would likely be detected by more people and with more awareness as to how they can affect the tables / game play as more time goes by - one of the reasons I personally think that rushing a final release will lead to some uncaught / unresolved problems with certain aspects missed / over-looked along the way.

 

@ mukuste - despite an update / test7 only fixing the depth bias I think it would be nice to still have an update with this fixed as I personally can think of a few dozen tables that need testing after it's release as you say it's got the depth issue fixed but will likely need some testing on a vareity of tables to ensure that it's fixed both problems being witnessed (the old one and now the current one messing up lights / light states and the roll-over animations). The release might be underwhelming to some as you describe but I really think it would be great if you could provide an update so this one key issue around the depth bias could be tested and in the time while other things are being worked on I and others could do some significant / detailed testing. If it's not a lot of trouble, I put my vote in for a test7 if you can and would be happy to spend some time testing it this weekend (along with some promised testing of 921rev930).


That table also is another one where the GI textures should have their transparency set to black, so you can try that. Weirdly, it doesn't seem to be necessary in my build.

 

This is not the case and it does not need any changes on the transparency setting to pure black like some other tables do.  See my notes above as I describe why the images originally had the black cut-out but did not use it any more as the technique was "upgraded" to placing the lights on a surface above the PF GI.  However, it now provides a better visual as to the bug that was occurring due to the depth bias problems.


Edited by jimmyfingers, 28 February 2014 - 11:32 PM.


#875 gtxjoe

gtxjoe

    VPF Veteran

  • VIP
  • 5,152 posts

  • Flag: United States of America

  • Favorite Pinball: Medieval Madness, AbraCadabra



Contributor

Posted 28 February 2014 - 11:36 PM

Move the VP9 DX9 registry entries to a new key, so us beta testers don't need keep changing our VP settings ;) With DX9 I bump all the settings up. With VP9.2.1, they are moderate

#876 BananaBoat

BananaBoat

    Enthusiast

  • Members
  • PipPipPip
  • 228 posts

  • Flag: Australia

  • Favorite Pinball: Tron LE

Posted 28 February 2014 - 11:42 PM

Could someone paste a screen shot of their nvidia 3d settings and VP9 DX9 video settings they use?
Thanks

Sent from my HTC_PN071 using Tapatalk



#877 koadic

koadic

    Pinball Fan

  • VIP
  • 1,363 posts
  • Location:Omaha, NE, USA

  • Flag: United States of America

  • Favorite Pinball: Addams Family/Fish Tales/Medieval Madness



Contributor

Posted 28 February 2014 - 11:46 PM

one question?
I understand the lighting problems on Rudy but why dont he move or even open his eyes? can i fix it in the editor?


I might have to update that table and the Simpsons to use ramps instead of gates when the dx9 version is officially released because obviously the .visible flag isn't being handled correctly for whatever reason.

#878 mukuste

mukuste

    Pinball Fan

  • VP Dev Team
  • PipPipPipPip
  • 854 posts

  • Flag: ---------

  • Favorite Pinball: Centaur

Posted 01 March 2014 - 12:00 AM

@ mukuste - despite an update / test7 only fixing the depth bias I think it would be nice to still have an update with this fixed as I personally can think of a few dozen tables that need testing after it's release as you say it's got the depth issue fixed but will likely need some testing on a vareity of tables to ensure that it's fixed both problems being witnessed (the old one and now the current one messing up lights / light states and the roll-over animations). The release might be underwhelming to some as you describe but I really think it would be great if you could provide an update so this one key issue around the depth bias could be tested and in the time while other things are being worked on I and others could do some significant / detailed testing. If it's not a lot of trouble, I put my vote in for a test7 if you can and would be happy to spend some time testing it this weekend (along with some promised testing of 921rev930).
 

 

Alright, you convinced me. Anyway these things are pretty trivial for me to release, I just wanted to avoid having an inflation of versions which nobody bothered to test anymore. But so far participation has been very good, so it should be fine.

 

I think after testing this new version you'll agree with me that finding and correcting the root cause (the wrong calculation of the Z planes) was the better approach over just reverting to an "almost" working state. It did take quite a bit more investigation and some broken tables in the meanwhile, but the end result is much cleaner and free of kludgy workarounds since I now actually understand what's going on. And as a bonus we get a fixed camera fitting routine in VP10. So I think it was well worth the effort.

 

Let me just write up the release post and off we go.



#879 randr

randr

    I'm just a hardware guy so...

  • VIP
  • 2,650 posts
  • Location:Minnesota

  • Flag: United States of America

  • Favorite Pinball: Twilight Zone

Posted 01 March 2014 - 12:06 AM

Sweet


Sent from my iPhone using Tapatalk

randr___pinball.png                         


#880 mukuste

mukuste

    Pinball Fan

  • VP Dev Team
  • PipPipPipPip
  • 854 posts

  • Flag: ---------

  • Favorite Pinball: Centaur

Posted 01 March 2014 - 12:08 AM

 

one question?
I understand the lighting problems on Rudy but why dont he move or even open his eyes? can i fix it in the editor?


I might have to update that table and the Simpsons to use ramps instead of gates when the dx9 version is officially released because obviously the .visible flag isn't being handled correctly for whatever reason.

 

 

Sorry, no, that's actually a bona fide bug. But since you stated it so succinctly, it took me all of three minutes to find and fix. I'm happy because now I have a third worthwhile change for the new test version! :)