Poor Performance After Yesterday's Update

[SilentRuin]

And to clarify: I have complete sympathy for you being a small team. But that doesn't release you from taking responsibility from delivering what you promise to work instead of blaming users for "holding it wrong".

FGU performs calculations on the full image despite tokens only seeing up to the next wall (LoS line). FGU causes load so high that a rather expensive 2018 Macbook Pro cannot deal with it. FGU causes high GPU load on high-performance graphic-cards while displaying a blank white image. These are either bugs or consequences of bad design, which a small team has harder times tackling than a large one. They still need to be tackled, though, while asking users to be patient instead of blaming them.

Nor does it release you from not stating "I'm guessing here but" instead of acting like you know it for sure. IMHO. I'd not bother engaging in some sort of "I know better" - I don't disagree that there are other performance issues going on right now in TEST server - but really... Duplicating it and providing campaign data is how you show you know what your talking about. Not just stating what you think it is. As the saying goes "put up or..." well that sounds to harsh - but you get the drift.

[Weissrolf]

Performance issues have been replicated and SW can have all the campaign data they need. That's not the problem here. Also not much guessing going on here, unless explicitly stated.

And frankly, if image size was the main culprit then this would just point to design problems again. There is no need to calculate stuff that's not visible or meaningful to gameplay anyway. For example, Foundry (any many other games) only calculates and displays animated FX for areas currently visible by a token, CPU/GPU load decreases accordingly.

FGU is in fact very capable at displaying even enormous images, except for memory consumption being too high for the size of images being displayed. You only run into problems when LoS then calculates vision through two dozen closed walls from one end of the image to the other.

[SilentRuin]

Performance issues have been replicated and SW can have all the campaign data they need. That's not the problem here. Also not much guessing going on here, unless explicitly stated.

And frankly, if image size was the main culprit then this would just point to design problems again. There is no need to calculate stuff that's not visible or meaningful to gameplay anyway. For example, Foundry (any many other games) only calculates and displays animated FX for areas currently visible by a token, CPU/GPU load decreases accordingly.

FGU is in fact very capable at displaying even enormous images, except for memory consumption being too high for the size of images being displayed. You only run into problems when LoS then calculates vision through two dozen closed walls from one end of the image to the other.

This is about TEST in laboratory. You realize this correct? For example the 100% duplicatable test case I provided to SW has nothing to do with your theories here. Provide your 100% duplicatable test with a small campaign made specifically for this with a step by step procedure to show them and they can fix it. But really? Your theory here? Does not match what I see. For example...

https://www.fantasygrounds.com/forum...l=1#post592039

They fixed the blackout one already with this type of "here is what it does" and hopefully they will fix the lag case. But your "have all the campaign data they need" is ... I'm too polite to say. But build your campaign and step by step PROOF and stop just shouting about what you think the problem is.

[Weissrolf]

My main theory here is that repeated "your image is too large" is not the way to improve the software, even less so the Beta.

[SilentRuin]

My main theory here is that repeated "your image is too large" is not the way to improve the software, even less so the Beta.

Exactly. Theory. Prove it or just hope it gets magically fixed and stop theorizing.

[celestian]

My main theory here is that repeated "your image is too large" is not the way to improve the software, even less so the Beta.

I agree in this case. Even a modest map will be 15 meg and 8Kx8K. I really don't want to hack up maps into smaller pieces because the platform can't deal with it. I was hoping we'd get past that in FGU. I get they are concerned about users with older systems and bad connectivity but do not make that the baseline for the rest of us.

[Moon Wizard]

That particular point is a valid discussion point.

Here is the information I have on image size:

* According to Carl, the size of the graphic asset should not have an affect on display performance directly.

* However, Carl has also repeatedly stated that the following items do affect performance:
** Size of the image window view. (i.e. pixels needed to be drawn)
** Number of tokens
** Number and placement of LoS blocker points
** Number of lights (lesser)

* Therefore, the size of the image "does" impact the above points, because larger images typically have larger numbers of all the other pieces of information. Large image files usually equate to larger data overall.

* Also, users typically want to open those "larger" images in larger windows to "see more"; which impacts the image view size which does impact performance.

* On a similar note, I still personally look at Photoshop as an example of a program that does similar things with layers, and how it definitely is performance impacted by large images 10Kx10K or larger. So, I'm not completely convinced that asset size does not have an impact; since it does take more memory, more load time and more network transfer time.

Regards,
JPG

[Weissrolf]

Here is a comparison of a single large image, no walls, 1 token in both FGU and Foundry on the same hardware. FGU's performance is considerably worse to begin with and it surely won't get better once more assets are added to the map.

https://www.fantasygrounds.com/forum...l=1#post592285

[HywelPhillips]

I hope the FG team will take this in the spirit in which it is intended. I'm a big fan of FGU's capabilities, it's my platform of choice. It is a hugely difficult thing to rewrite all the code base into a new framework and maintain products which date back a decade or more. It's a huge achievement to do it with a small team, and I think we all salute you.

We're here because we want to help the platform deliver on its promise. But it does have significant performance limitations that need to be addressed. I run FGU on an iMac or iMacPro that are a couple of years old but among the fastest machines Apple has ever produced, with maxed out memory, SSD and GPUs. I can edit and colour grade 4K video professionally on these machines.

FGU isn't - or at least from my back-of-the-envelope calculation shouldn't - be doing as much GPU-intensive math as a 3D game running at full stretch. For all the lighting and rendering multiple points of view for all the tokens, it should be dealing with 2D spaces, not 3D spaces. As Weissrolf has reported there are reasons to point to algorithmic improvement to reduce the LOS workload rather than limiting the size of user assets. I don't want to have to abstain from using features like lighting because of performance issues on a top-spec machine.

There's also no doubt that the single-threaded nature of a lot of FGU's processing is a serious limitation. I have nine cores sitting idle, one fully loaded and a spinning beach ball every time I search assets for "pumpkin". That's not a bandwidth issue - that's an inefficient search algorithm which not only incurs that time penalty the first time it does a search, it incurs it all over again when you move to page two. And again for page three. There's no caching implemented there, for example, which would at least moderate the problem it if there is no more efficient search algorithm available. And really the rest of FGU shouldn't freeze while the asset window has to go away and do a big old disk search.

Performance can be a bugger to improve, but there are some serious bottlenecks right now which need to be addressed. On the plus side I don't see any fundamental reason why it cannot be improved, since other products have achieved it. Happy to help with the testing when I can, but others are more set up to do it more efficiently than I am, I think.

Cheers, Hywel

[SilentRuin]

Here is a comparison of a single large image, no walls, 1 token in both FGU and Foundry on the same hardware. FGU's performance is considerably worse to begin with and it surely won't get better once more assets are added to the map.

https://www.fantasygrounds.com/forum...l=1#post592285

I trust that is in TEST as your posting it here? Also - I also trust you mailed them the campaign with step by step instructions to duplicate it. Else... you know. They have to duplicate it with what you have.