How does MMF use video memory?

I believe in optimizing... BUT!.

what no one is getting is that these cards are supposed to have millions of samples from thousands and sometimes millions of surfaces in a fraction of a second; one surface with only a few skins loaded at a time is making a problem out of nothing.

...and despite that I'm somehow magically too close to the limitation of what my graphics card can handle. Everyone doesn't have the latest hardware, and I want my games to be optimized to run on as old PC's as possible.

I'm saying this again; I want my games to be optimized as much as possible in every aspect a game can be optimized in. I'm not asking for your opinion about whenever optimizations is necessary or not with today's GPU's. I've heard your opinion, I get what you're saying, and I disagree. I, just like DevonV, is interested in knowing how MMF2 uses the video memory, because knowledge about your dev tool is always more than useful.

very well, then the only other thing i might point out is use TGA for large images(the GPU handles them quicker)

Also the power of two thing is really just for skins if you can do it without visual loss then by all means...

on whim; ' I'm somehow magically too close to the limitation of what my graphics card can handle.'
how do you know?

Quote:

---

how do you know?

---

Slowdowns as a direct cause by the amount of graphical objects.

http://www.clickteam.com/epicenter/ubbthreads.php?ubb=showflat&Board=64&Number=71601

Quote:

---

Originally Posted by Yves

MMF2 doesn't do anything specific to optimize the textures, I've been trying to do something automatic but it doesn't work very well so it's not in this version. Not sure if I'll have time to do it before the release version. So you should try to use images with a size close to a power of 2 if you want to optimize the memory used by the application.

---

I hope that answers your first question sufficiently.

It does :) I hope we'll later learn the rest too!

wait... was'nt DevonV asking?

Quote:

---

Originally Posted by Joshtek

I hope that answers your first question sufficiently.

---

Yes, that's quite clear, thank you. :)

Quote:

---

Originally Posted by SEELE

wait... was'nt DevonV asking?

---

Yes, but the answer is useful for anyone interested in the technical aspects of the engine.

Quote:

---

Originally Posted by SEELE

what no one is getting is that these cards are supposed to have millions of samples from thousands and sometimes millions of surfaces in a fraction of a second; one surface with only a few skins loaded at a time is making a problem out of nothing.

---

The point I was actually trying to make wasn't about the card's ability to render sprites. Any 3D card should be blazing with sprites. It's that the cards are designed to process large amounts of graphic data very quickly, and they do so because they expect the data to be in a known, standardized form. An average MMF game designed without consideration of this fact will, on average, give them something other than expected, standard data, and therefore the results will be less predictable. Some cards might auto-convert without a hitch, some might even automatically optimize the VRAM usage, but others might have cheap components or weak drivers and have to spend more time converting the data into the format it expects in order to run optimally.

I do feel that heavy optimization can be important, because when you are making a 2D game today, most people are going to view it as a casual game by default, and they will expect casual system requirements. Trying to sell a 2D game that requires this year's GeForce model is going to confuse people. The people looking for 2D games don't necessarily have a 3D card beyond whatever integrated chip is on their motherboard, and most of the people who own major 3D cards probably want a 3D game that's going to (visually) push their hardware.



One more question if any of the Devs want to get into it:
How are the animation frames of a active object stored in VRAM? Do they get arranged into a square/near square sprite sheet-like single texture, or are they stored as individual textures? (Or a hybrid, like the former, but split into no larger than 1024x1024 textures to avoid issues with old cards?)

Again, this is better answered when the engine is closer to official release, but I'm putting it out there.

Quote:

---

Originally Posted by DevonV

One more question if any of the Devs want to get into it:
How are the animation frames of a active object stored in VRAM? Do they get arranged into a square/near square sprite sheet-like single texture, or are they stored as individual textures? (Or a hybrid, like the former, but split into no larger than 1024x1024 textures to avoid issues with old cards?)

---

I don't know about HWA, but I believe in regular MMF2 the image bank is separate to individual objects but referred to by them, meaning if you use the same animation frame anywhere it'll only store it once.

Quote:

---

Originally Posted by Joshtek

I don't know about HWA, but I believe in regular MMF2 the image bank is separate to individual objects but referred to by them, meaning if you use the same animation frame anywhere it'll only store it once.

---

I suspected it worked that way. As I recall HWA will crash if an object's size is greater than 1024 pixels (or at least it will if the card cannot render larger textures). I should just animate a huge object and see how it reacts.

Thanks.