1353 replies to this topic

[fuzzel]

I can only second Toxie's opinions. VPX is ready for free camera movement the only reason not to do it is performance issues on older/mid range gfx cards and another aspect that I've repeated multiple times yet: tables had to be built in a different way with much more primitives instead of custom shaped walls that use texture atlases as much as possible to reduce draw calls.
My current WIP Goin Nuts table is built mainly with primitives and where possible three primitives are joined together(like the plastics e.g). The plastics use just one big 2048x2048 texture so it can be easily cached by the hardware. Tables that use dozen of separate primitives/ramps/walls like posts that are copied multiple times on a table would kill the performance because the hardware must render each of then again and again and not just with one render call.
So at the end the author should know something about how a gfx card works to get the most out of it and this knowledge isn't fully necessary at the moment because of the static render state at the beginning.

[dark]

2048x2048 for all your plastics is kinda low res by today's standards, the new standard would seem to roughly be 1024x1024 for EACH plastic as primitives on average depending on size of plastic of course.  Would it be any better to have all those plastics jammed into one massive image?...and or combine as many plastics into one large prim?

[fuzzel]

Yes combine as many primitives into one and attach one big texture to it. I won't go so far to create a giant texture that could cause problems on gfx cards with less memory or older cards that don't support textures bigger than 4096x4096.
The rule of thumb is to combine those primitives together that belong together like one plastic primitive for the upper half of the table and one for the lower half instead of 5-10 separate plastics. And combine textures into one or more bigger textures. Don't use 5-10 primitives that use 5-10 different textures. Especially plastics, posts and screws are a good example. It's much more effective to render them with one or two render calls instead of 5-10 or more.

[toxie]

I avoid always-on timers whenever I can, but still one of the most common complaints I get is stutter related to script timers. And it kills me, because people mess with the timer intervals and that wrecks the timing for everything.

 

Highest on my wishlist is code fading for Flasher objects. I know Lampstates have to be handled by a timer, I'm not under any delusions that JP's light routine is slower than uselights=1 or whatever, but the fact that the timer has to handle fading as well just seems outdated to me.

 

Also now that we're in the age of modulated solenoids, it'd be great if there was a light state that could fade both up and down dynamically based on an input value. I've made a script that does this, but the timer has to be really fast or a lot of detail gets lost.

 

Thats exactly where you can now use timers with the interval set to -1. These are triggered only exactly once before each frame is then rendered. And using the GameTime call you can measure how much time has passed since the last update, so you can still fade correctly.

[dark]

Yes combine as many primitives into one and attach one big texture to it. I won't go so far to create a giant texture that could cause problems on gfx cards with less memory or older cards that don't support textures bigger than 4096x4096.
The rule of thumb is to combine those primitives together that belong together like one plastic primitive for the upper half of the table and one for the lower half instead of 5-10 separate plastics. And combine textures into one or more bigger textures. Don't use 5-10 primitives that use 5-10 different textures. Especially plastics, posts and screws are a good example. It's much more effective to render them with one or two render calls instead of 5-10 or more.

Yeah that's a problem though because how are you supposed to fit 10x 1024x1024 images into a 4096x4096 image so it put us back where we started.  Posts and screws are easy because they look the same, but plastics are more dynamic, they all have unique artwork and can't use the same map unless it's a map with multiple maps piled into one.  I guess you could do groupings of say 3 plastics but then we end up with more large sized images, will this really amount to a significant increase in performance even with taking a hit from using larger image sizes? (Fitting as much into 4096x4096 as you can for example.)

[vampirolatino2]

I don't want to sound rude or lack of respect to those using old hardware. But there is an already good version, VP9, VPX (or any other iteration) should move/look forward. This is not 2000 anymore, is almost 17 years. Heck, 1 year in Computer technology is like 5 of old years. So ... for those using old/mid hardware.... I think is time to move on. Time to catch up to current computer hardware specs. If technology doesn't move forward mankind have not achieve how we have accomplish today.

 

I remember when I tried to keep using old hardware and trying to squeeze all I wanted for new games to look awesome. But at the end I had to move forward because I was missing A LOT. Thank God I did. I now want those missing years back!!!

 

edit: typo

Edited by vampirolatino2, 30 November 2016 - 01:13 AM.

[vampirolatino2]

For those developers and table builders my most special gratitude. Because after all this is a hobby and is your free time on the line. Wish I appreciate a lot!

Table builders will always try to squeeze or push the program (VPX) to their max. They want the most awesome table made, look how much from VPX1.0 to now they have accomplish! This is because all (including developers) are trying to get the most of it. But sadly (I think) the software have its limits, right? That's the only reason I think we need to move to dynamic lightning and 3d objects only. With a good and optimized code it should do the trick.

 

But I don't know anymore I haven't code in yeeeeears, my last time I coded something, it was a game when the Xbox 360 started their indi program. I'm and old folk who programmed/used in old days C++, perl, java, javascript, vbscript, html, python, ruby, oracle, access, sql, flash, directx 9, opengl, etc etc. Ended when C# was introduced and only touched a little. So, I'm out of the game.

 

I occasionally do some vbscript and/or powershell scripts for my work, but that's it. I'm not an active programmer anymore. I'm more in the hardware/maintenance part of the industry (IT Security Specialist / Service Engineer) working all day/night with hundreds of Servers...etc,etc, cant disclose too much. That's why I don't offer any help. I would be like an old car moving in the fast lane. Hehehe

 

So, after all this blah blah.... at the end I follow what you guys decide. I can only voice my opinion. Kudos to ALL of you! Keep up the good mood and work.

Edited by vampirolatino2, 30 November 2016 - 01:48 AM.

[DJRobX]

I don't want to sound rude or lack of respect to those using old hardware. But there is an already good version, VP9, VPX (or any other iteration) should move/look forward. This is not 2000 anymore, is almost 17 years. Heck, 1 year in Computer technology is like 5 of old years. So ... for those using old/mid hardware.... I think is time to move on. Time to catch up to current computer hardware specs. If technology doesn't move forward mankind have not achieve how we have accomplish today.

 

I tend to agree.  But on the flipside, I think it's worth looking into optimization opportunities.   If I look at some of the graphics in commercial 3D games, we seem to need an AWFUL lot of power to draw a little rolling ball around a static table.   I can play TPA for hours on my old iPad.  The graphics in TPA are quite good and it's not a big issue.    That's not to say anyone's doing a bad job, we've come such a long way in the last few years from the old DX7 VP, just that I think performance improvements are worth investigating.   

 

And while you might say it's not 2000 anymore, CPU power hasn't accelerated at the rates it used to.  Now people are interested in smaller, quieter machines.   They want to play VP on tablets.      The more we optimize the engine, the more developers can go crazy with special effects and other fun things to make it more realistic. 

[robertms]

Just a side note.. l love following this thread, so much insightful information discussed and shared daily by skilled and super smart people. Thank you for allowing all of us regardless of skill set to participate in the discussion and witness the birth of new ideas and functionality that we all so crave. Truly awe-inspiring stuff!

Having said that, we're nearing 1000 posts here but how's it possible this topic been rated only 4 times so far??? C'mon people, I know it's a silly thing but for crying out loud show the VP devs your appreciation. 5-stars from me!

[arngrim]

what is old hardware as you see?
the recommandation to use vpx is i think a gtx960 4gb or better isn't it?

Sent from tapatalk

[TerryRed]

I use an i5 2500K CPU and a GTX 960 2GB at 1920x1080.

 

All VPX graphics options are maxed out with 4x anti-aliasing. VPX runs silky smooth in exclusive fullscreen for me, with B2S, DOF, and all cabinet toys running.

 

As a reference, FP with BAM "New Renderer" runs almost maxed out (no ambient occlusion, kills fps with new renderer) with my setup on 95% on the tables, while running DOFLinx.

 

PFX2 runs maxed out no problem, while running DOFLinx.  

 

TPA, maxed out as well.

Edited by TerryRed, 30 November 2016 - 07:09 AM.

[fuzzel]

Yes combine as many primitives into one and attach one big texture to it. I won't go so far to create a giant texture that could cause problems on gfx cards with less memory or older cards that don't support textures bigger than 4096x4096.
The rule of thumb is to combine those primitives together that belong together like one plastic primitive for the upper half of the table and one for the lower half instead of 5-10 separate plastics. And combine textures into one or more bigger textures. Don't use 5-10 primitives that use 5-10 different textures. Especially plastics, posts and screws are a good example. It's much more effective to render them with one or two render calls instead of 5-10 or more.

Yeah that's a problem though because how are you supposed to fit 10x 1024x1024 images into a 4096x4096 image so it put us back where we started.  Posts and screws are easy because they look the same, but plastics are more dynamic, they all have unique artwork and can't use the same map unless it's a map with multiple maps piled into one.  I guess you could do groupings of say 3 plastics but then we end up with more large sized images, will this really amount to a significant increase in performance even with taking a hit from using larger image sizes? (Fitting as much into 4096x4096 as you can for example.)

Well it depends on how detailed a plastic really is. Not every texture needs to be 1024x1024 especially if the plastic primitive is not that big. And yes it makes a difference if the whole table is build in that way because the gfx card can cache the important textures and hasn't to switch them.

[toxie]

i just fixed the ESC bug, so everything should be back to normal with the next build (otherwise let me know please).

[jpsalas]

It also used to quit when I pressed the "D" key to open the debugger (1 out of 10)

[toxie]

i hope this should also be fine. if not, then please let me know.

[wrd1972]

Oh the humanity!

You mean I might have to upgrade my Core2-quad CPU one day?

[vampirolatino2]

Oh the humanity!

You mean I might have to upgrade my Core2-quad CPU one day?

 

  kill it! hahahaha

[freneticamnesic]

I thought I was going crazy with it constantly quitting (even had it quit when I pressed D as JP pointed out) - glad I wasn't alone! Thought some new "features" were added while I was MIA

[toxie]

maybe we could settle on some super-easy intermediate solution?

if a specific option is set, then the file open dialog is automatically opened on VP start and also acts like open and play?

and the default would be that this option is enabled??

 

opinions??

 

this is now implemented, e.g. in the next build.

[ICPjuggla]

has the next build been published yet with the esc bug fixed? its driving me crazy!