SpaceHack - Chronicles of the Final Push

[dislekcia]

This is the game that QCF is putting out for Dream Build Play this year... Been wanting to start a thread for it for a while, but today something rather funny happened:

Aeq started messing with adding bloom on the edges of our objects of interest while he's waiting for me to finish some framework stuff so he can populate the bosses with nasty fire patterns.

So he suddenly bursts out laughing and shouts "Hah! I've made the game 8-bit!"

[Chippit]

Unintended effects ftw. It's the only reason the ice in UQ2 looks the way it does.

[Azimuth]

That looks frickin' sweet.

[Fengol]

Frickin awesome Aeq! The 8bit legend will never die while we have dev's like you around :D

[Thaumaturge]

Heheh, that's a pretty cool effect - will you be keeping it, or removing it?

[FuzzYspo0N]

easter egg ftw

[edg3]

Talk about awesome! :O

I havent quite worked out how to use shaders yet...

[DukeOFprunes]

Didn't Darwinia's pixel effect do something in this vein? Looks cool!

[dislekcia]

Originally Posted by DukeOFprunes

Didn't Darwinia's pixel effect do something in this vein? Looks cool!

Yeah, it did. We can pretty much duplicate Darwinia's look now, all we have to do is turn the blur on the bloom effect down and change the texture filtering and it'll look all pixellated :)

But I prefer having glowing edges on important things, it makes the game a lot more alive.

[Gazza_N]

Well, with a second pass (and speed allowing) you could have pixellated graphics with extra bloom... >.>

[dislekcia]

But smooth particles are sooo neat :(

Just read the thread about the D3 health drops vs potions idea... I've been doing health drops since the first prototypes of this game over a year ago. It feels better, mainly because movement is the main player action in the game, so movement over something that's going to heal you provides players with very attractive 10 second goals.

[dislekcia]

Some more screenshots, I'm stressing out the game and profiling it to find slowdowns at the moment. Which means I've switched over to a windows build and I can take screenshots for a bit :)

You can see the random ship generation and a test particle trail which we totally stole off Geometry Wars 2. Aeq's working on new modules and I have to make sure we don't get slowdowns with 200+ enemies anymore...

[Coolhand]

Did you get nProf running yet D?

[FuzzYspo0N]

This is looking awesome Dis and Aeq. I love this style of game art. makes me wanna play rRootage again... The bloom looks more win!

[Cyberninja]

The game looks awesome D. :)

[dislekcia]

Originally Posted by Coolhand

Did you get nProf running yet D?

Yeah I did, thanks to your detailed instructions! (I felt like such a tool ;)) Next thing I knew it was 3am...

Originally Posted by FuzzYspo0N

This is looking awesome Dis and Aeq. I love this style of game art. makes me wanna play rRootage again... The bloom looks more win!

rRootage was a definite inspiration for how the enemies are generated. Kenta Cho is a hero of mine :) So if we can produce something that reminds people of his work, I'm going to be really happy.

[dINGLE]

looking good ...

after you have entered the comp, will it be publicly released?

[Coolhand]

Originally Posted by dislekcia

Yeah I did, thanks to your detailed instructions! (I felt like such a tool ;)) Next thing I knew it was 3am....

lol :)

I told you 3 months ago that it was not only simple, but ALSO super cool ;) Drop me an email if there are an XNA-side functions that are killing you and I'll let you know what I may or may not have to sort them our on our side (ie. the craziness that is 'IsSphereInFrustum()').

Originally Posted by dINGLE

after you have entered the comp, will it be publicly released?

I'm guessing if all goes well it will... for a cool 800 (give or take) MS Points ;)

[dislekcia]

Originally Posted by Coolhand

lol :) I told you 3 months ago that it was not only simple, but ALSO super cool ;) Drop me an email if there are an XNA-side functions that are killing you and I'll let you know what I may or may not have to sort them our on our side (ie. the craziness that is 'IsSphereInFrustum()').

Right now our main slowdown is a really stupid (N!) collision hack I did on enemies and good old Effect.Begin() :(

I'm probably going to have to dump meshes that share the same effect in buckets instead of brute-forcing it. Although on the 360 it doesn't seem to give a shit about rendering at all... My slowdown on that side appears to be from churning through long lists. I think.

I really wish I had something like nProf for 360. The remote performance monitor is kinda useful, but pales in comparison to the awesome that nProf is capable of diagnostically...

Originally Posted by Coolhand

I'm guessing if all goes well it will... for a cool 800 (give or take) MS Points ;)

Ding! Got it in one. Although we might go for a lower price-point to give people something to snap up when they don't have a ton of points to spare. Not sure yet.

[Coolhand]

Originally Posted by dislekcia

good old Effect.Begin() :(

Good times...

I think that and parameter uploading are our biggest ones. There really is no way to zero it out, but there are ways (ie. buckets...) to reduce it. Too bad you can't see if it is actually a culprit on the 360 though...

I just put a new system in that for every mesh loaded, it checks the effect (combination of an EffectID which I store as a .FX parameter, and the TechniqueID for that EffectID), and for every unique combination (like only 10 or 20 or so in our stuff), I do a single Effect.Clone(), and keep a list of these cloned effect around, and give each mesh a reference to the one that matches it (if a mesh requests an existing "cloned' effect, I just up the reference count and return an existing one).

Then at render time, I do ye-old-bucket-sort for the mesh of each object in Mr. Frustum, with a simple Effect.Equals() between existing bucket entries and each mesh/effect combo. Once this is done, just go through the bucket, set up the effect for rendering, then for each mesh that uses that effect just upload the unique parameters manually that change from mesh to mesh (matrices, textures, etc.). Then wash-rince-repeat for each bucket entry.

The effect setting and parameter uploads still take up a large percentage of the time, but if I remember correctly, a sample scene for us on my laptop went from like 10 fps to 60fps or so. Not that is of any kind of empirical value though.

But I'm sure you know all that already... I'm just rambling :)


Careful though... as I learned with this, Effect.Clone() is uuuuuuuuuuuuuuuuuuuuber-slow on PC at least... like a level load that should take 10 seconds takes 40 for us if I just ALWAYS clone every effect loaded (was only test code, but still scary...).

Oh snap, also be careful of running in windowed mode. The EndRender() (or whatever it is called) takes up huge amounts of time when running in windowed mode, so make sure you ignore that function in nProf. Again, the difference on my laptop between running windowed / fullscreen is something close to 100% (ie. 25fps to close to 50fps in one case I remember...).