Game Development

12:33 AM

I'm off to bed, we'll fart again tomorrow

@Wardy Looks good, but yeah, you definitely need to work on the colors :P

im experimenting with it now ... im not far off

i'll post another comparison in a bit

I'll have to head to bed. Be sure to ping me when you do post it. ttyl

ok dude

later

AttackingHobo

2:17 AM

gfycat.com/EnlightenedRemoteHomalocephale

2:27 AM

@William'MindWorX'Mariager Well i'm still not happy with it but its a ton closer than it was

slap an atmosphere around it and its good enough to move in I spose :)

I really need to get back to the size problem next

this planet just aint big enough!

lol

@Ali.S @Pip Figured I would just ping you guys ... you seemed to like the last pics I posted these are only going to get better!

@Lasse these planets are part of my scaleable voxel architecture ... the end game being that you can land on them and then play minecraft if that's the sort of game you're building or not if you just want a subset of that.

A decent engine with hardcore threading would make this code really awesome!

right im off to bed ... slap together some LOD architecture tomorrow I spose ... its either octrees or something custom ... not decided yet.

5:15 AM

7

Aidan Mueller

5:25 AM

@Wardy How close is that moon thingy? It looks a bit close.

5:46 AM

Pretty sure its just a comparison of his planet vs earth to show scale

At least that's my guess.

7:23 AM

I ALWAYS had a sneaking suspicion of this

:O

random ladies be all like... oh come in for some "heals".. never a dude

I don't see how that would possibly mean anything else

haha

That's clearly a brothel

A healer would have the red + on it

7:28 AM

very true

toby

it's obviously a nurse

i should load this up... see if my rupee count diminishes upon receiving "help", lol

user92578

7:50 AM

@toby Yeah, I'm using C++ and SLD2

user92578

@AttackingHobo true, and I already made it a lot smaller

AttackingHobo

user92578

morning guys!

user92578

@AttackingHobo looks interesting

9:05 AM

@Wardy I have been also building a whole platform to remake minecraft mechanics and fix some of the issues it has when it comes to modding, but that project is on pause at the moment

but yeah, that looks cool :)

different planets were also part of my plan, seems like it really can be implemented

user92578

turns out sdl's collision detection method is slowing me down quite a lot

user92578

oh wait no it isnt

SDL has collision detection?

user92578

yeah

I thought it was only a graphics libraray

user92578

9:12 AM

it's a multimedia library

user92578

it has a lot of things

yeah right that was the word I was looking for

what's the api for collision detection?

user92578

	if (SDL_IntersectRect(boundingBox->GetRect(), GetRect(), m_NullRect) == SDL_TRUE)
	{
		return true;
	}
	return false;

user92578

m_NullRect is a rect where the intersection amount is placed

user92578

but I don't care about it

9:14 AM

okay

resharper (code quality plugin) would whine about that

return SDL_IntersectRect(boundingBox->GetRect(), GetRect(), m_NullRect) == SDL_TRUE;

that would do the same thing

user92578

thanks :D

user92578

that is some of the first code I wrote when I started working in C++

user92578

so a 9x8 grid of dynamic boxes slows the game down a bit

user92578

that aint good

user92578

maybe I need to place those in the quadtree too...

user92578

9:17 AM

it's just slow to update quadtree positions

user92578

as I need to first find the item, remove it, and the re-insert it if it's position has changed

user92578

I guess I should've seen this coming as I had so much fun adding features yesterday, this day has to be a performance improvement day

there are always those days

user92578

9:37 AM

Hmm, didn't see this one happening:

user92578

@Wardy That looks super sexy. :D

user92578

rounding seems to be quite slow

user92578

when done many times

That sounds unlikely.

user92578

9:39 AM

SDL_Rect* CBoundingBox::GetRect() const
{
	SDL_ModifyRect(m_Rect, roundf(X), roundf(Y), roundf(W), roundf(H));
	return m_Rect;
}

user92578

This is one of the slowdown places

I've rounded around ~400k times a second without a hitch. :P

user92578

The method itself uses 19.96% of total cpu

user92578

SDL_ModifyRect takes 3.04%

user92578

and "External Code" (probably rounding) takes 12.03%

9:41 AM

Move the rounding into separate lines, and see if that is true. :)

Also, is that 19.96% of a time slice? Cause that doesn't mean it's slow. It just means you spend more time doing that than everything else.

Which makes sense if you round 4 times for every entity for every frame.

user92578

now the precentages are smaller, modifyrect takes 1.62% and external code takes 6.7%

user92578

the whole method 11.81%

user92578

maybe it's something about copying values when they are passed as parameters?

@AidanMueller thats deliberate so i can test the matching I was doing (the bigger of the two is earth the smaller is my planet, the idea being that they are both to approximate scale)

@Lasse These planets only need a seed, they are noise based: i'm using a customisable noise module stack. One of the things I really need to get working is my noise toolkit in either unity or VS (its a node based editor)

user92578

It's always that one method that slows everything down

9:47 AM

@William'MindWorX'Mariager I know right :) It's still quite all there, and I'm in the process of looking space partitioning / LOD systems to handle the orbit to ground transitions

@Tyyppi_77 I'm still curious how exactly you're measuring.

user92578

I'm using the VS profiler

@Tyyppi_77 is this that circle thing ?

user92578

?

Hmmm maybe it was someone else

You working with pygame?

user92578

9:51 AM

no

ok my bad ... crossed wires

If you have a strongly typed language a "cheap round" might be to simply cast the value to an int

Alright, consider this. Your program spends 0.4µs on everything else and 0.1µs on GetRect(). Does this mean that GetRect() is slow?

user92578

okay 50% of the total CPU usage of my game is taken by the collision checking method

It comes down to 20% for GetRect().

its not quite the same but it would save time

user92578

9:52 AM

1 min ago, by William 'MindWorX' Mariager

Alright, consider this. Your program spends 0.4µs on everything else and 0.1µs on GetRect(). Does this mean that GetRect() is slow?

The answer is no. It's still only a total of 0.5µs. It just means you added some code that is run often.

user92578

well true

user92578

okay so how would I figure out what "actually" takes a lot of time

Well, if you're actually experiencing lagging, you could use this.

But it's running at 20% CPU the whole time, so I doubt it's lagging.

How many times you calling that check entity collision method per frame?

user92578

9:57 AM

the amount of dynamic entities times two (one for the x axis and one for y)

why ?

not all entities can move

user92578

dynamic

surely you could handle that as part of updating an entity?

user92578

I have separated static entities and dynamic entities

when an entity is updated check for collisions

user92578

9:58 AM

Yeah, that's what I'm doing

but only between that and nearby entities

user92578

that's the problem, as I don't store the dynamic entities in a quadtree where I could query them

you probably want to use an oclusion map then

so say you have a player entity and then your "world" you can draw a texture that shows where the player can move to, then its just a lookup on a set of pixels

saves you having to query all your entities

could be as simple as a big array of booleans or something

user92578

like currently I can get the static collision boxes from a quadtree

user92578

but the problem is when I have a lot of dynamic entities that I can't just query from a quadtree

user92578

10:04 AM

so currently I need to loop over 116x2 boxes per entity per frame

collision boxes yes ... the idea behind an occlusion map is that all the collision boxes can looked up in 1 hit by querying just the map

user92578

I could also try putting the dynamic entities into the quadtree

say you had a texture for your occlusion map ... all the places where you dude could move could be white and all the places he cant could be black, then you could simply ask "are any ofthe pixels im using to draw my dude black" if yes collision

simplifies the problem down to a single lookup on a map

116 queries just became 1

user92578

but then I'd have to update that texture

user92578

and figure out what is on that black pixel I couldn't move to

10:06 AM

depends how your world works

user92578

and pick the correct response based on that

surely you got some sort of tile info for your world ?

when you gen a new world tile add to the map

and when you remove a tile take it out the map

Just go with the quad tree :P

It's fast and efficient.

And just add rects on top of every tile. Your worlds aren't huge, so it shouldn't be a problem.

in fact i think some even use the occlusion map as their raw map data storing say an array of numbers where up to a certain value is pass able and beyond that isnt

You could even run some optimization if you really felt like it.

10:08 AM

then 1 set of data solves the collision and rendering problems

my idea works the same way @William'MindWorX'Mariager just flattens the problem down

no need for the tree basically

I don't think it would work as well in 3d

but quad trees in 2d are overkill imo

user92578

So my world in terms of collision is currently divided into two parts

they add complication where really you don't need it

user92578

world collision areas are drawn in tiled over the terrain tiles

user92578

those collision boxes are stored in a quadtree

user92578

then there are dynamic entities (that update every frame) that are currently just stored in a vector

10:11 AM

surely tiles should just be marked passable or not then their bounding boxes are the collision bounds ?

user92578

basicaly if my level has just a solid straight floor on the bottom, it is drawn in the level editor as just one collision box

right i got work to do ... looks like im implementing an octree today :(

user92578

My individual tiles don't have their own bounding boxes

having an occlusion map basically gives you that and combines them

so you get an image of where overall things can and can't be moved

user92578

but I pretty much have that, and I have no problem with the static level geometry

10:15 AM

so do you have dynamic entities that are impassible ?

is that the problem ?

user92578

yeah

hmmm ... might make sense to do like you said before then given where you are

throw your dynamic entities in to the quad tree

job done

user92578

the thing is that removing from an quadtree is slow

slow why ?

user92578

as I need to remove and re-insert the dynamic entities there every frame

user92578

10:16 AM

it's a big recursive method

why would you do that?

user92578

when the position changes, I need to find a new place for it in the quadtree

do they definitely move every frame ?

user92578

no, but very likely

user92578

enough to cause slowdowns if removed from the quadtree

10:17 AM

but if you know the node the entity was in then basically the new insert is just a parent level request ?

you shouldn't need to crawl the whole tree

user92578

that's actually a good point

then if that parent deems you outside its range it can pass up

so you only go up as far as you need to

rather than starting from the root and working down

that could be pricey

user92578

but entities don't know where they are in the tree

ok but the treenodes know what entities they have right ?

user92578

yup

10:25 AM

so maybe entities could raise an event to say "i moved"

then the node they are in could subscribe to that

it can then handle moving if need be between itself and another node

likely by doing like i said and passing the event up the tree

I have a very complex problem to solve that im making absolutely no progress on

i keep putting it off :(

I have to generate planet regions in chunks and pass to an octree

2d feels real good right now

user92578

:D

my process goes something like this ...

for(each region)
   generate new game object
   kick off threaded task to gen surface
   gen mesh for the region
   when task completes, texture mesh with results

evil math

user92578

Omg I'm lerping a lerp factor

lol - that's evil too

user92578

10:41 AM

"once you go lerp, you never go back"

user92578

10:55 AM

what's the "best" way to pass around std::vectors?

user92578

std::vector<CBoundingBox*> CLevel::GetCollisions(const CBoundingBox* target) const
{
	// Starts with all the dynamic bounding boxes whose locations are unknown.
	std::vector<CBoundingBox*> result = m_LevelObjectStorage->m_BoundingBoxes;

	std::vector<CBoundingBox*> quadTreeQuery = m_LevelObjectStorage->m_StaticBoundingBoxTree->GetInArea(target);
	// Adds the quadtree query results to the vector.
	result.insert(result.end(), quadTreeQuery.begin(), quadTreeQuery.end());

	return result;
}

user92578

This code is called a lot, and my profiler is showing that a lot of allocations are being made

user92578

how would I make that more effecient?

user92578

in terms of passing vectors around

either create the vector on the heap and return a smart pointer or create a vector in a broader scope and use references

11:00 AM

Or get a pointer to a vector in and let the calling function take care of the allocation

or pass a vector in a parameter

Reference parameter also works (pass by value doesn't I think)

user92578

aight I think I'll try returning a reference

Where will you be creating the vector then?

be caureful, you can't do this as is

11:01 AM

Cause a reference is just a pointer with fancy syntax

user92578

yeahyeah, I'll create a vector on the CLevel

Aight :)

good idea, in that case it may be simpler to just pass the vector in the function and let it fill it instead of returning a reference

since you already have the object that you get the reference to

actually you don't have to pass anything since the vector would be usable inside the function

11:25 AM

gah ... anyone know how to turn degrees in to x,y,z

don't seem to be the same as polar coordinates

degrees from each axis?

yeh ... im confusing myself

Polar coordinate system

In mathematics, the polar coordinate system is a two-dimensional coordinate system in which each point on a plane is determined by a distance from a fixed point and an angle from a fixed direction. The fixed point (analogous to the origin of a Cartesian system) is called the pole, and the ray from the pole in the fixed direction is the polar axis. The distance from the pole is called the radial coordinate or radius, and the angle is the angular coordinate, polar angle, or azimuth. == History == The concepts of angle and radius were already used by ancient peoples of the 1st millennium BC. The...

so I have north south east and west as angles in degrees

and I need to determine the bounds for which I am going to generate a mesh

Spherical coordinate system

this may be useful

In mathematics, a spherical coordinate system is a coordinate system for three-dimensional space where the position of a point is specified by three numbers: the radial distance of that point from a fixed origin, its polar angle measured from a fixed zenith direction, and the azimuth angle of its orthogonal projection on a reference plane that passes through the origin and is orthogonal to the zenith, measured from a fixed reference direction on that plane. The radial distance is also called the radius or radial coordinate. The polar angle may be called co-latitude, zenith angle, normal angle,...

11:47 AM

yeh i think that's where i got my existing implementation from

Don't think my implementation is correct though

        public static SphericalPosition FromCartesian(float x, float y, float z)
        {
            var r = (float)(Math.Sqrt(Math.Pow(x, 2) + Math.Pow(y, 2) + Math.Pow(z, 2)));

            return new SphericalPosition
            {
                Radius = r,
                Inclination = y > 0 ? (float)(Math.Acos(z / r)) : (float)-(Math.Acos(z / r)),
                Azimuth = x > 0 ? (float)(Math.Atan(y / x)) : (float)-(Math.Atan(y / x))
            };
        }

        public void ToCartesian(out float x, out float y, out float z){

I'm sure i missed something when testing this the numbers don't feel right

might put together some unit tests

isn't Math.Atan only for -pi/2 to pi/2 ?

ignore

I don't really understand any of those functions

lol I just know that from the details on wiki ect if I call them I get the number i need (in theory) for my calc

it also depends from which axis you start measuring the angles

is the FromCartesian function correct?

given that my point of origin (sphere center) is at 0,0,0 in cartesian space if I know an angle up and down (north south) and an angle left and right (west east) I should be able to determine a point in cartesian space

@Veritas thats the problem I don't know

user92578

can you do ToCartesian(FromCartesian());?

user92578

12:01 PM

and see if the result is the same as the input? or does it not work this way?

thats the idea i should be able to do something like ...

var pos = new Vector(x,y,z);
var spherical = SphercalPosition.FromCartesian(pos.x, pos.y, pos.z);
var newPos = spherical.ToCartesian();

I'm not convinced its all working right as I would expect.

Try Math.Atan2

user92578

@Wardy and newPos == pos, right?

it should

I'm just putting together a test app now

I am almost sure it's the Math.Atan but take my work with a grain of salt

check this part in wiki: The inverse tangent denoted in φ = arctan(y/x) must be suitably defined, taking into account the correct quadrant of (x,y)

basically if x < 0 you need to differentiate on y < 0 and y >= 0

you don't take into account the sign of y when calculating Azimuth so i think that's the bug

12:19 PM

ok something definitely aint right

rounding issues

hmm can you try larger x,y,z values?

ignore

in all cases (except 0,0,0 of course) surely the radius should come out 1

you forgot to divize z

with the sqrt of the squares

ignore

guys I'm off too tired

doh

:(

Hmmm actually i think this looks correct

if im at say 1,0,1 then im outside the sphere so the radius will be high

Q: C++ Very Sleepy strange results

12:37 PM

that's probably it

yes this makes sense

well it's not that it's outside the sphere

the radius is of the sphere's after all, the radius of the sphere on which (1,1,0) for example lies just doesn't have to be 1

the end results also look pretty good

user92578

I'm getting really weird profiler results

user92578

on the top of the list is a function that is supposed to log stuff but no output is seen

user92578

Basically I have the same problem as in this

user92578

-5

I'm currently programming an SDL Network Game, everything is perfect ... But my game only runs at 30fps, so I ran a Very Sleepy analysis. First strange effect, very sleepy increase the perfs of my Game to 60Fps ... ( Why ? ) And secondly, there is a lot of SDL_LogCritical calls ( you can get the...

user92578

but there's no real answer to his problem

12:50 PM

@Veritas its the conversion back I think that's wrong

eg: (1,-1, 0) ends up being (0.9999, -1, -6.181)

the spherical values look reasonable and within the sort of range I would expect (not sure if they are exactly correct though)

but from those it gets a way out cartesian

1:04 PM

it's E-8

so that's pretty close to 0

hmmm

so maybe i just need to apply some accuracy rounding

that's like 0.00000006

not really worth it imo but depends on the context

hmmm

Jon

Ali.S

1:53 PM

@Wardy it seems way too good to be real ...

what does ?

@Ali.S ?

Ali.S

your planet textures

textures are easy to make look nice ... just crank up the resolution

the hard part is tying that back to the data

2:13 PM

So @Wardy what was the problem?

I don't think there was one

I was reading it wrong

Didn't notice the -8E

I think i mostly use the values against my voxel array anyway so the float values get cast to ints

dammit i still don't have what i need though

user92578

2:31 PM

I'm kinda in a point where I don't know what I need to do next

same here

Jon

2:43 PM

store.steampowered.com/app/327890

user92578

I think I pretty much have gameplay wise everything programmed

user92578

except if I wish to add something like falling floor etc.

user92578

So I think I'll get to programming the main menu

user92578

or something

got the map editor finished?

user92578

2:48 PM

I don't have a map editor

how do you implement game content?

user92578

tiled

That diagram might solve my spreading.

Basically, I look for the situation on the left, and then mark the lower right particle that it should move right.

Next step, I move it right if it can.

I could use the alpha channel for "desired" movement.

Q: How to get a position from a radial distance and angles?

3:23 PM

0

I have a radius and the 2 angles for northSouth and eastWest rotations in degrees. I need a Spherical position or a cartesian position (I have code translate one to the other) Where a position is defined as one of the following ... class Vector { public float x; public float y; public f...

c# unity mathematics coordinates

Meh, I have no idea where to even start with this

4:12 PM

Better sand spread, but it seems to have a weird stable state that I can't figure out how is happening.

@William'MindWorX'Mariager "a wierd stable state"?

Yeah, those single particle lines are stable. They're just standing there, not spreading out.

I guess I'll have to try and run through the steps by hand and see where the problem is.

It's pretty simple, which is why it confuses me.

Basically, if there is air below, it moves down. If there's air on the left and dust below, it moves left. If there's air on the right and dust below it moves right.

is it because you have a condition where the grain of sand above is perfectly positioned above the one below so is considered to be "supported"

It does this in random order.

you probably want some sort of diagonal check or something

oh i c

hmmm

Pip

4:18 PM

@Wardy looks good

:D

they are awfully close together ... is there some sort of edge case / race condition occurring where the column between 2 stacks is considered populated

@Pip I know right :)

just trying to figure out how to get the points to apply my mesh generator

Actually.

@Pip gamedev.stackexchange.com/questions/95876/…

It's possible that it just moves right, and then immediately left again before the dust above falls.

can't get my head round that at all

@William'MindWorX'Mariager that sounds plausible

@William'MindWorX'Mariager maybe cellular automata could help here

you'd have to sequentially process though

4:25 PM

Yeah, that's not good for GPU :)

or do some sort of pattern based processing

That's what I did before :)

But that resulted in pyramid stacking :P

Not really good looking.

how does water simulation usually work on the gpu ?

don't they do it over many frames

I'm not sure.

I wonder ...

Could it be that it can't pick out the individual pixels properly?

Nah, then I shouldn't be able to render it properly either.

Hmh

hmmm i think (it's been a while so don't quote me on this) but from when I last looked the usual trick was to render a frame in the initial state then apply changes to half the data then on the following frame apply to the other half

that'sa really old school way of doing it though

has to be a smarter way

particle coupling perhaps ?

4:34 PM

Hm

Well, what I'm doing now should work. I just have to figure out why it does the weird stacking.

scirra.com/forum/…

is that what you're trying to achieve?