[swift and shift]

icantread49 posted:

Whats a good webm hoster? puush appears down at the moment, and I'm not going to risk my Dropbox account with garry's auto-play bullshit

can't you just put it on nullular.com

Edited:

ooof page king, have some content: http://codepad.org/ytBv4UwR

Edited:

this one might also give y'all a thrill: http://codepad.org/gqoOMXZV

Edited:

my personal favourite: http://codepad.org/jgg9NrKu

[Richy19]

swift and shift posted:

ooof page king, have some content: http://codepad.org/ytBv4UwR

How... I dont even...

[Lexic]

Naelstrom posted:

I thought we've been over this.

Yes, we did go over it, and we established that anyone who included loud pointless irritating music in their webm videos was a douchebag.

[Ziks]

Richy19 posted:

How... I dont even...

http://codepad.org/lgfIfGOe

[thingshappen]

icantread49 posted:

It looks like most of these issues come from Marmalade failing to set the correct GLES context.



And then it goes:



Which also explains the debug screen showing while the actual game is all black (until it exits).

I gotta investigate these context issues. Marmalade 5.2.4 supposedly fixed them, but I guess not.

Edited:

Do you mind if I post your log on the Marmalade developer forums?

Don't mind at all.

[COBRAa]

Ziks posted:

http://codepad.org/lgfIfGOe

Just to clarify, I assume the *(pi+1) increases pi and points to b (isn't memory assigned backwards normally? that is from high address space to low) and then sets its value to (int)3.14 (floored to 3).

[swift and shift]

COBRAa posted:

(isn't memory assigned backwards normally? that is from high address space to low)

yeah the x86 stack grows downwards

[COBRAa]

What is the reason they made it descend?

[swift and shift]

COBRAa posted:

What is the reason they made it descend?

it's x86, nobody has a clue

[COBRAa]

Also, a doesn't need to be volatile.

[swift and shift]

snip

[NovembrDobby]

icantread49 posted:

I don't want to spam my Youtube account with what are essentially short GIFs for WAYWO

Unlisted?

[Funley]

I just fucking FIXED the damn depth testing issue by just moving one line 5 lines down. Just that after messing with matrices and vertices for many hours.

Now i can actually start working on a better map generating, point sprites, collision, lighting, player, GUI, menus and all kinds of cool stuff.

Edited:

I have now separated the controls from the camera to a player class, took a while to figure out how to send variables to each other.

[Mordi]

Funley posted:

I just fucking FIXED the damn depth testing issue by just moving one line 5 lines down. Just that after messing with matrices and vertices for many hours.

Now i can actually start working on a better map generating, point sprites, collision, lighting, player, GUI, menus and all kinds of cool stuff.

I've been working on getting fonts to render in an efficient way for days now. Hopefully I won't be sick of C++/OpenGL when I get it to work properly. Looking forward to creating a GUI of some sort afterwards.

[synthiackup]

bonus realtime version of that video i posted on the last page (warning: requires oracle® java™)
http://synthiac.net/stuff/tetra.win32.rar

[Speedfalcon]

Matt- posted:

Oops I'm late but:



Rating winners of the last thread.

So most of my posts were either wrong or not programming :v

[Chris220]

synthiackup posted:

bonus realtime version of that video i posted on the last page (warning: requires oracle® java™)
http://synthiac.net/stuff/tetra.win32.rar

Is stuff like this more often done with realtime beat detection, or are the timings hardcoded? I'd assume the latter because the demo only has to work well with that one song, and it cuts out the fact that beat detection is rarely perfect.

[DrLuke]

Computer Science at school:

5% programming, 95% this:

[ahdge]

Phone picture beats screenshot when you need to tell a story.

[Icedshot]

DrLuke posted:

snip

Computer science at my school:

[BlkDucky]

Icedshot posted:

Computer science at my school:

I was waiting for an image to load for a good few seconds there.

[Funley]

DrLuke posted:

Computer Science at school:

5% programming, 95% this:
[dicks]

Computer Science at school:

5% do something with Office, 95% play games on Facebook

[DrLuke]

ahdge posted:

Phone picture beats screenshot when you need to tell a story.

Also I was too lazy to take a screenshot and paste it into paint and then save it onto my usb-drive

[icantread49]

New version: http://filesmelt.com/dl/Phyzicle3.apk

It should now work for all you guys that reported a black screen/crash/nothing happening!

For the guys with the screen corruption, I'll look into it. I believe it might just be your old device, though

[ahdge]

DrLuke posted:

Also I was too lazy to take a screenshot and paste it into paint and then save it onto my usb-drive

Just want to point out that you copy and pasted a throbbing member 14 times.

[Funley]

Oh man, seeing this makes me happy



Upgrading the map generator also allowed to have a maximum map size of 342x341, that's one wider that the last one! Though depending on the amount of wall pieces generated, this much can crash because the vertice list getting too big.

Now just to figure out how to flip vertices 180 degrees around to make a roof.

[Darwin226]

ahdge posted:

Just want to point out that you copy and pasted a throbbing member 14 times.

Tiled background.

Edited:

Funley posted:

Oh man, seeing this makes me happy



Now just to figure out how to flip vertices 180 degrees around to make a roof.

I think you just need to set y/z (what ever is the up vector) to -y/-z

[ahdge]

Shit.

[Funley]

Darwin226 posted:

I think you just need to set y/z (what ever is the up vector) to -y/-z

-snip. fixed-

[Icedshot]

Got back to work on my 3d opencl renderer

Objects are now asynchronously added during the runtime of the program - memory is allocated for the size of all the triangles currently in any object being rendered to the scene, this is all copied onto the graphics card, and then the current engine memory pointer is updated (mutex lock to prevent awful things from happening) to the new memory. The old memory can then be freed asynchronously as well, after this

This essentially means that adding new objects to the scene won't cause any delay whatsoever to the running of the program (though may take several frames for the actual memory itself to be allocated, and the object to be drawn to the screen). Also, because the memory new memory has to be allocated while the old memory is still present, this means that you are wasting a load of memory while transferring the two. I can't think of a way around this though at the moment

Currently though, I am very lazily getting around to implement a system to pass textures to the program. Unfortunately opencl is not a fan of being passed arrays of Image2ds, though perhaps i can pass a 3d texture that is actually the 2d textures on different layers - each layer being an object ID (or a variation of that).

I'll get around to it at some point

[Naarkie]

Powered by GeeUI™

(Collaborative project with Ziks and Geel9)

[synthiackup]

Chris220 posted:

Is stuff like this more often done with realtime beat detection, or are the timings hardcoded? I'd assume the latter because the demo only has to work well with that one song, and it cuts out the fact that beat detection is rarely perfect.

yeah i'm not the hugest fan of beat detection/fft. i find manual timing provides more rewarding results, and it takes as much time as one would spend making analysis results usable.

[Darwin226]

Funley posted:

Well, i generate the vertices like this:

verticesList.Add(new VertexPositionNormalTexture(new Vector3(position.X, position.Y + 1, -position.Z),
                new Vector3(0, 1, 0),
                new Vector2((currentwall * 2) / imagesInTexture, 1)));
            verticesList.Add(new VertexPositionNormalTexture(new Vector3(position.X, position.Y + 1, -position.Z - 1),
                new Vector3(0, 1, 0),
                new Vector2((currentwall * 2) / imagesInTexture, 0)));
            verticesList.Add(new VertexPositionNormalTexture(new Vector3(position.X + 1, position.Y + 1, -position.Z),
                new Vector3(0, 1, 0),
                new Vector2((currentwall * 2 + 1) / imagesInTexture, 1)));

            verticesList.Add(new VertexPositionNormalTexture(new Vector3(position.X, position.Y + 1, -position.Z - 1),
                new Vector3(0, 1, 0),
                new Vector2((currentwall * 2) / imagesInTexture, 0)));
            verticesList.Add(new VertexPositionNormalTexture(new Vector3(position.X + 1, position.Y + 1, -position.Z - 1),
                new Vector3(0, 1, 0),
                new Vector2((currentwall * 2 + 1) / imagesInTexture, 0)));
            verticesList.Add(new VertexPositionNormalTexture(new Vector3(position.X + 1, position.Y + 1, -position.Z),
                new Vector3(0, 1, 0),
                new Vector2((currentwall * 2 + 1) / imagesInTexture, 1)));

That generates one floor piece. How exactly can i change how each vertice is oriented?

Instead of position.Y + 1, try -(position.Y + 1). If Z is up then do that to z.

[Tyco RC]

Icedshot posted:

Got back to work on my 3d opencl renderer

Objects are now asynchronously added during the runtime of the program - memory is allocated for the size of all the triangles currently in any object being rendered to the scene, this is all copied onto the graphics card, and then the current engine memory pointer is updated (mutex lock to prevent awful things from happening) to the new memory. The old memory can then be freed asynchronously as well, after this

This essentially means that adding new objects to the scene won't cause any delay whatsoever to the running of the program (though may take several frames for the actual memory itself to be allocated, and the object to be drawn to the screen). Also, because the memory new memory has to be allocated while the old memory is still present, this means that you are wasting a load of memory while transferring the two. I can't think of a way around this though at the moment

Currently though, I am very lazily getting around to implement a system to pass textures to the program. Unfortunately opencl is not a fan of being passed arrays of Image2ds, though perhaps i can pass a 3d texture that is actually the 2d textures on different layers - each layer being an object ID (or a variation of that).

I'll get around to it at some point

I'm jealous, I've had to pretty much halt my OpenCL work in order to deal with the end-of-course ramp-up in my other classes / final projects / procrastination because I was working on my photon mapper so much this quarter.

Depending on how complex your scenes are (last time you showed it it looked like maybe a cube-room and a couple of spheres) you shouldn't have too much to worry about in scene allocation. If you aren't already just make sure you're storing the minimum amount of information necessary, like how a sphere only needs 4 floats to represent it completely.

My scenes are currently ~200 bytes including my material definitions and I have the same number of primitives as you do, it shouldn't be terribly expensive.

[acpm]

beat detection is worthless in the first place

there is no magical universal tuning for frequency averages that will detect beats on every single last song provided, so there's no point in it for using it for the reasons you'd expect (to do less work)

the problem with beat detection is that to use it for anything useful, you'd need a set of algorithms that could detect beats across multiple different genres and play styles

that's one problem, but you also have the general "how do you define what a beat is?" issue

in the end, beat detection is best used in controlled, tuned cases. but if you're going to do that, you might as well just manually sync whatever your work is to whatever music you're using, because you'll end up actually doing more work tuning your beat detection for more than one song

[DeadKiller987]

Funley posted:

Computer Science at school:

5% do something with Office, 95% play games on Facebook

Computer Science at school:

10% do something with Office 2003, 15% play mari0, 75% sit quietly while secretly planning how to place the cd in my backpack into the cd-tray and reboot using kon-boot with admin privileges so I can fix their antivirus software and set up user rights properly to prevent some asshole from removing my mari0 saves forcing me to beat it each Tuesday.

[TerabyteS_]

I ended up accidentally creating the most attractive project of the science fair had to take part in (the robot arm) and got myself a backache and a sore throat trying to explain what the thing did to kids and their parents. The arm wouldn't work 50% of the times, yet the kids were still incredibly amused when I switched to manual control and grabbed pieces of LEGO and let them go in their hands.
The guys showing their projects in the stands next to mine came to me complaining that I were getting all the attention and I replied that I never asked for it and would gladly switch places

The kinds fucking about with the arm were incredibly unnerving though.

[esalaka]

amcfaggot posted:

beat detection is worthless in the first place

there is no magical universal tuning for frequency averages that will detect beats on every single last song provided, so there's no point in it for using it for the reasons you'd expect (to do less work)

the problem with beat detection is that to use it for anything useful, you'd need a set of algorithms that could detect beats across multiple different genres and play styles

that's one problem, but you also have the general "how do you define what a beat is?" issue

in the end, beat detection is best used in controlled, tuned cases. but if you're going to do that, you might as well just manually sync whatever your work is to whatever music you're using, because you'll end up actually doing more work tuning your beat detection for more than one song

Iunno Audiosurf seems to do it decently for pretty much everything although some songs have been recognized as waaaaay too high-tempo.

[acpm]

esalaka posted:

Iunno Audiosurf seems to do it decently for pretty much everything although some songs have been recognized as waaaaay too high-tempo.

audiosurf is far from perceptually perfect by a long shot

[synthiackup]

esalaka posted:

Iunno Audiosurf seems to do it decently for pretty much everything

placing a block in a random lane whenever there is an impulse in the spectrum is hardly decent.