1031 replies to this topic

[DJRobX]

 

I can confirm if your on win7, and using Frezzy's 1.6.1 -1.7.0 DLL, all SAM tables will need at91jit value set to "0" 

I can also add doing so I have seen no performance issues...

 

 

You will have problems with LE SAM games (Star Trek LE, Walking Dead LE, Mustang, ACDC Premium, Metallica Premium) crashing (lights dying, slow DMD frames).   The IRQ timings are not tight enough in the non-JIT core (and making them tighter would make it considerably slower). 

atj91jit=0 is a time bomb in these games.
 

Edited by DJRobX, 01 July 2018 - 12:06 PM.

[Carny_Priest]

Yep, can confirm dead lights and slow DMD with TWD LE running with JIT disabled and freezy 1.6. I have not tested with freezy 1.7 but it does not sound like anything was attempted to address this. Too bad. Native VPM rendering does not cut it for SAM system. 

[toxie]

As said, please bug mjr about this. I failed to fix this and i do not see how i could in the near future. If enough ppl ask mjr (kindly!) he might invest some time again maybe??

[DJRobX]

As said, please bug mjr about this. I failed to fix this and i do not see how i could in the near future. If enough ppl ask mjr (kindly!) he might invest some time again maybe??

 

I've spent countless hours trying to reproduce the problem myself with no luck either.   I have two win7 machines and set up a Win7 VM ... none of them reproduce the crash.  

Yep, can confirm dead lights and slow DMD with TWD LE running with JIT disabled and freezy 1.6. I have not tested with freezy 1.7 but it does not sound like anything was attempted to address this. Too bad. Native VPM rendering does not cut it for SAM system. 

Upgrade to win10.   The problem will go away.    The issue is squarely a VPM one, but the additional "load" makes it trigger more readily.   I'd be happy to be wrong, but I don't think a fix is coming any time soon. 

Edited by DJRobX, 02 July 2018 - 06:02 AM.

[toxie]

I have the problem on Win10, too, from time to time (also with the latest update).. It was even more common/reproducible on an older version of Win10 for me..

 

So unfortunately Win10 is no silver bullet..  :/

[Carny_Priest]

Agreed. I'm running Win 10.

[Thalamus]

Hmm. Maybe that explains why I'm NOT seeing problems here. I trust Rob when he says that at91jit should be enabled for the games he mentions but, except those I turn it off without seeing any problems, like Hauntfreaks mentions. It looks to me that as long as I refrain from starting tables from double clicking their name, crashes doesn't happen. To me it sounds like this varies between hardware ?! I haven't played enough since 1.7 of freezy to conclude on anything. Still haven't seen any dmd delay or odd behavior. I know you guys have squashed a few bugs and I've found some myself - eg. loading pbx with animated gifs. I recommend reboot if you're on Win7 and the machine has been in sleep/hibernation state.

[DJRobX]

I have the problem on Win10, too, from time to time (also with the latest update).. It was even more common/reproducible on an older version of Win10 for me..

 

So unfortunately Win10 is no silver bullet..  :/

 

I had been running the original build of win10 on my cab until only recently.  I've never seen this crash on that system.   I really wish the bug happened to me more often, because then I'd be able to actually fix it.  

Note that re-reading Thalamus' post that it can sometimes crash on double-click starting the table *without* at91jit set ... that would make the bug something completely different.  As Toxie mentioned we'd want to get a crash txt of that, as we might be confusing the two completely different issues.

Edited by DJRobX, 02 July 2018 - 03:47 PM.

[Carny_Priest]

After Independence Day holiday I'll enable JIT and try to generate a crash file using my current set up if someone doesn't beat me to it. No crash with JIT disabled, just the dead lights and slow DMD. Sometimes. Sometimes TWD LE plays fine. I don't have any pattern I can point to for when and under what circumstances the slow down happens. 

[DJRobX]

The crash.txt/dumps of the JIT crash aren't very helpful because they occur in dynamically generated code.   Thalamus mentioned he sometimes gets crashes with at91jit=0.    In that case the crash is something completely different and we need a dump for that.

 

The DMD frame slowdowns when at91jit=0 is an understood issue - poor IRQ timing that is not easily solved without seriously killing the performance of the interpreted C core.   It is the whole reason I put all the effort into making the JIT core work for SAM.  

Edited by DJRobX, 02 July 2018 - 06:09 PM.

[bord]

Can we get a depth bias option added to the walls option menu? When using walls for plastics w/ transparency some elements aren't showing through the transparent part of the plastic. A depth bias adjustment should fix that.

[fuzzel]

rev3465 is up:

 

- fix obscure bug where starting the blank/built-in tables with the options pane closed causes the ball to levitate off the table

- Allow ESC to resume table if pause dialog is up (currently does nothing).
- update hankin.vbs so the "5" key,etc add coins on tables, too
- fix rendering of bumper base/skirt if ring is disabled
- built-in ball control: Enable in debug window (or via "ball control always on" in the global options)
   - left mouse click directs the ball towards the mouse cursor, left flipper key releases control
   - can also double click to make the active ball jump to the cursor
   - can hold left mouse button down to guide the ball
   - if "throw" and "ball control" are both checked, one can use the throwing action on the ball control instead of making a new ball

Edited by fuzzel, 03 July 2018 - 11:47 AM.

[hauntfreaks]

Real Ball control, now that's a super sweet feature we have needed from day one!!... awesome
and now all the people that wanted a pause button can shut up!!...

great thanks

[kiwi]

Of course there are no news if there are no answers, but I still try to ask the same if there are news about the bugs reported in the bug tracker?

 

Thanks

Edited by kiwi, 04 July 2018 - 02:34 PM.

[toxie]

No.. :/

 

But i looked just over them again quickly, which one(s) would you need 'desperately' to be looked into?

Cause i also track the other ones reported in here and always kinda like 'prioritize' based on how exotic an issue is in practice.

[kiwi]

No.. :/

 

But i looked just over them again quickly, which one(s) would you need 'desperately' to be looked into?

Cause i also track the other ones reported in here and always kinda like 'prioritize' based on how exotic an issue is in practice.

At the bumper issue..

[DJRobX]

Something up with the rubber changes seems to have affected FishTales.   Rubbers27, 28 and 29 and rendering weird on the preview, and it's pushing the POV way out of whack.   

https://imgur.com/a/C3VbThQ

[DJRobX]

Found it ... there was a block of code removed with a comment that rubbers are always a loop.   This table has some that aren't (you can make a non-looped rubber by deleting all but 2 control points).   I committed the fix. 

[toxie]

Okay, thanks, will double check if it behaves as expected for the other cases later-on..

[toxie]

I had to refix it, as it broke my previous fix, but now all cases should work..