I've just uploaded it to the V3 demos section. It's really cool. :D

...

More to come - sprite support for the particles. It'll be spiffy!

Spiffy. Triggered the snow a few times and had quite the blizzard going, good job on letting it stack.

El

Quote:Originally posted by Buckermann

Nice.
For your information:
at about 300 particles my lowly Athlon 1500/ Win98SE computer showed the first, serious signs of performance degration.

Mine was aching at 150 particles... but then again I run a lot of background crap.

sweetness and light:)

on my barton 3000+, I was able to get well over 500 particles before I started noticing managable lagging. Frames started to skip(very slightly) at around 250.

Third Release

Much changed. Vector calculations are actually correct. Streaks are in. Performance improved.

Pardon? My brain does flip-flops parsing that sentence. Do you mean that I should treat down as up?

Great.
Runs really better than the first release (sorry, didn't had the time to check release two).

But I have a few questions.
Is Max Particles dependent from the computer speed? Mine is 269.

After maybe 5 minutes on the first screen (Wheather-Rain-Streak) Max Particles turned to a large negative number, around -210000, and no particles were displayed. Again, after a while, Max Particles turned back to 269 and the rain was displayed again. As far as I can tell this behavior also happens with the other particle setups. Intentional?

Wheather-Rain-Streak had around FPS ~30 with ~60-70 particles.
Mystic-Vortex-Old had around FPS ~35 with ~260-270 particles.
A rather large difference. I'm curious, may I ask why?

For motion blur:
If your particles are moving at a constant speed, you could calculate the particle coordinates for a few ticks in the past, and put some (grayed out?) graphics on those coordinates too. But for particles moving with changing speeds, storing the last few values is probably quicker and easier.
I don't remember the Golden Sun effects all that well, so I don't know if this is what you're looking for.

I'm not sure if you've thought of this, but...

If you render all particles to a layer...

Then quit recalculating and reblitting each particle.

Each time you draw to the layer, save a copy. Then draw those saved layers (perhaps the last five or six) at a set translucency on top of each other. THAT'S good motion blur. That whole "re-render everything" isn't going to cut it.

Save the last five layers. At each new draw, set destination to the draw layer, then draw the previous five buffers at translucency, then the new batch of particles.

For additional blur, keep track of up to ten past buffers, maybe? It will still start slowing down after awhile, but blitting an entire screen in one pic is much faster than reblitting each individual item on that screen.

Remember, there's a difference between Blur and Shadow. A shadow effect requires that copies of the images be blitted (like Alucard's trail in Symphony of the Night), and that's what you're doing right now. Blur is where the eye can't immediately update what it's seeing, and so you see the current picture along with a blurred previous image (like motion blur in Majora's Mask).

Shadow == trails of elements
Blur == previous data + new data of all elements

The difference is pretty obvious (Alucard's trails vs. motion blur). I'm not quite sure if my method is a flawless way of doing it, but hey, it won't hurt to give it a try.

Well, the speeds of the particles are pretty much independent of the speed of the computer. A particle that moves 10 pixels on a faster machine is still going to move 10 pixels on a slower machine, the only difference will be the time it takes. They will both move the same amount. If I save a copy of the last 4 frames (standard for motion blur, I did some research a few days ago), then the amount of blur will be the same on all computers, just the amount of time it will take to show it will be different. Which brings me to my point: the slow performance of translucency.

I don't know why it's so slow in verge, but I'm guessing it's because it's being done in software. If I were to save a copy of the last 4 frames, place them one on top of each other translucently, the speed hit would probably be unbearable on everything but the fastest machines. One of the first things I did to try to improve performance of the particles was to render all the particles to a transparent layer, and then blit the layer, ensuring that the particles would only be redrawn after they were recalulated, which is, by default, 100/30 FPS (or 3 times a second). That's pretty cheap, really, and could improve performance quite a bit. Unfortunately, something went horribly wrong, and I lost a lot of work.

If all you were going to use were pixel particles, you would actually lose speed doing it the layer way, because drawing the pixel particles every single frame is much faster than blitting a 640x480 (or what have you) bitmap every frame. Now, when doing circle and the other particles (which require significant math), it should in theory be faster to just blit the image. But it wasn't. I don't know what I did, maybe I did it horribly wrong, I'll try it again. In theory, if it worked, you'd be able to display about 3 times as many particles at once (but not really). And, I'd be able to do nifty effects on the particle layer as well. Perhaps motion blurred, shadowed, streaking particles (all at once) at an acceptable FPS may be yet a reality. Here's to hoping.

-- edit

The odd thing is, in Golden Sun, I can't really tell WHAT the particles are doing to achieve that glowy, blurry effect. Common sense tells me that, based on the GBA's speed, it can't really be doing anything that complex. But it isn't simply a shadowing effect, and I can't really tell if it's a blurring effect. More studying will be required. Yeesh, I love reverse engineering. Without access to the engine.