[9fans] Attempt at making 3D graphics library

[9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

Hello. I realized that that last /usr/glenda/bin/rc/pull focused on a  
rewrite to the system libraries. I decided to start writing a library  
to convert 3D coordinates to 2D coordinates based on the data at  
http://www.shade.ca/getarticle.asp?i=11. You can find it at /n/ 
sources/contrib/pietro/3d.bundle. It contains: 3d.h (the header),  
3d.c (the library itself), and 3dtest.c (a test program). The test  
program doesn't seem to work right. It looks like the 2D coordinates  
are way off the charts - the x coordinates are between 500 and 600  
for xyz=(5,5,5)! I wonder... This makes the line a stub in the center  
of a 1024x768 screen. Plus, the threadmain crashes on threadexitsall.  
What's going on?

Re: [9fans] Attempt at making 3D graphics library

[Paul Lalonde]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

It's hard to tell what might be wrong with your transformation if you  
don't also show the transformation matrix you are using, including  
the camera parameters.  What you probably want is for your transform  
to go from some world space coordinates (your 5,5,5) to screen  
space.  It's also not clear if you are expecting device coordinates  
(0..1023, 0..767) or normalized coordinates (0..1, 0..1).   Your  
resulting coordinates aren't too far off for some camera vaguely  
pointing at the origin and a device coordinate transformation matrix.

Paul


On 23-Nov-07, at 7:43 PM, Pietro Gagliardi wrote:

> Hello. I realized that that last /usr/glenda/bin/rc/pull focused on  
> a rewrite to the system libraries. I decided to start writing a  
> library to convert 3D coordinates to 2D coordinates based on the  
> data at http://www.shade.ca/getarticle.asp?i=11. You can find it  
> at /n/sources/contrib/pietro/3d.bundle. It contains: 3d.h (the  
> header), 3d.c (the library itself), and 3dtest.c (a test program).  
> The test program doesn't seem to work right. It looks like the 2D  
> coordinates are way off the charts - the x coordinates are between  
> 500 and 600 for xyz=(5,5,5)! I wonder... This makes the line a stub  
> in the center of a 1024x768 screen. Plus, the threadmain crashes on  
> threadexitsall. What's going on?
>

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFHR8tTpJeHo/Fbu1wRAsrrAKC6Bmdz31+VYVSEGWRiGRkD133Y7QCgmCzS
56gVAizyJ4ndkHWUIixuJ1k=
=PMob
-----END PGP SIGNATURE-----

Re: [9fans] Attempt at making 3D graphics library

[anyrhine]

> I decided to start writing a library  
> to convert 3D coordinates to 2D coordinates based on the data at  
> http://www.shade.ca/getarticle.asp?i=11. You can find it at /n/ 
> sources/contrib/pietro/3d.bundle. It contains: 3d.h (the header),  
> 3d.c (the library itself), and 3dtest.c (a test program).

did't check the bugs nor the article, so i hope i'm not entirely
off here. there is already libgeometry, described in matrix(2),
will do the job for you. it does what 3d engines do, except
clipping and rendering.

Re: [9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

On Nov 24, 2007, at 1:57 AM, Paul Lalonde wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> It's hard to tell what might be wrong with your transformation if  
> you don't also show the transformation matrix you are using,  
> including the camera parameters.  What you probably want is for  
> your transform to go from some world space coordinates (your 5,5,5)  
> to screen space.  It's also not clear if you are expecting device  
> coordinates (0..1023, 0..767) or normalized coordinates (0..1,  
> 0..1).   Your resulting coordinates aren't too far off for some  
> camera vaguely pointing at the origin and a device coordinate  
> transformation matrix.
>
> Paul
>
>

I was using no transformation matrix. I'm also thinking I'm getting  
the screen width and height wrong. Either way, there are formulae in  
the middle of that article, reproduced here:

.EQ
x sub { 2d } -> x sub { 3d } * { N over z } + { { screenwidth } over 2 }
.EN
.EQ
y sub { 2d } -> y sub { 3d } * { N over z } + { { screenwidth } over 2 }
.EN

It also says at the top that z=N and because equivalence is  
bidirectional that N=z. I do expect device coordinates.

I remember a while back about a discussion on 3D graphics  
programming. Andrey said that OpenGL was ported a long time ago but  
was never tied to rio or 8 1/2. He also said that it would be doable  
but difficult via APE. Unfortunately, I don't know how to program  
OpenGL, and by the code I've seen to produce a single object in a  
display I'm afraid of approaching it. I was hoping my simple point  
conversion mechanism would light a spark. Guess I'll try again later :-)

Re: [9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

I got it all to work now using the 3DMOL code, converted from C++ to C.

mkdir 3d # download
cd 3d
rc /n/sources/contrib/pietro/3d.bundle
8c 3d.c # compile
8c 3dtest.c
8l -o 3dtest *.8
3dtest # run

You can see a cube in the center of the screen. Use q to quit (and  
crash, for some threadexitsall-related reason). This could be the  
first step to a new 3D engine in C for Plan 9 - or for any system.  
I'll give them my regards now.

Re: [9fans] Attempt at making 3D graphics library

[andrey mirtchovski]

> Use q to quit (and
> crash, for some threadexitsall-related reason).

that usually means you need a bigger stack for your threads.

Re: [9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

On Nov 24, 2007, at 11:00 AM, andrey mirtchovski wrote:

>> Use q to quit (and
>> crash, for some threadexitsall-related reason).
>
> that usually means you need a bigger stack for your threads.
Setting both threads to 4096 stacksize fixed it. I'll note that in my  
tutorial. I used one thread to interpret mouse events, since for some  
odd reason eresized() is sent through the mouse. :-(

Re: [9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

On Nov 24, 2007, at 2:11 AM, anyrhine@gmail.com wrote:

>> I decided to start writing a library
>> to convert 3D coordinates to 2D coordinates based on the data at
>> http://www.shade.ca/getarticle.asp?i=11. You can find it at /n/
>> sources/contrib/pietro/3d.bundle. It contains: 3d.h (the header),
>> 3d.c (the library itself), and 3dtest.c (a test program).
>
> did't check the bugs nor the article, so i hope i'm not entirely
> off here. there is already libgeometry, described in matrix(2),
> will do the job for you. it does what 3d engines do, except
> clipping and rendering.
This library has Point3, which is Point3D with an object w that is  
the denominator to (x,y,z) in the final point. I didn't need this, so  
I made it 1 in my local copy and changed all Point3Ds to Point3s. The  
3D library now uses libgeometry and will have a box drawing facility  
the next update.

Re: [9fans] Attempt at making 3D graphics library

[Federico G. Benavento]

[-- Attachment #1: Type: text/plain, Size: 183 bytes --]

You're not supposed to combine event(2) with thread(2).

See mouse(2).


Federico G. Benavento

---
/bin/fortune:
Steal money you're a thief: steal a country you're a king.

[-- Attachment #2: Type: message/rfc822, Size: 4030 bytes --]

From: Pietro Gagliardi <pietro10@mac.com>
To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
Subject: Re: [9fans] Attempt at making 3D graphics library
Date: Sat, 24 Nov 2007 11:21:57 -0500
Message-ID: <9BD59CB3-3F41-415B-AC30-39773DDF99C2@mac.com>


On Nov 24, 2007, at 11:00 AM, andrey mirtchovski wrote:

>> Use q to quit (and
>> crash, for some threadexitsall-related reason).
>
> that usually means you need a bigger stack for your threads.
Setting both threads to 4096 stacksize fixed it. I'll note that in my  
tutorial. I used one thread to interpret mouse events, since for some  
odd reason eresized() is sent through the mouse. :-(

Re: [9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

Is there any purpose in qball(2) being in the 3D engine instead of my  
hit-and-run rotation system which I am unsure of?

Re: [9fans] Attempt at making 3D graphics library

[Pietro Gagliardi]

Then how do you check for the window resized event? I heard from IRC  
#plan9 that resize events are sent through /dev/mouse, so the  
following code is the event(2) way of ensuring resize events get done:

	for (;;)
		if (ecanmouse())
			emouse();

On Nov 24, 2007, at 5:48 PM, Federico G. Benavento wrote:

> You're not supposed to combine event(2) with thread(2).
>
> See mouse(2).
>
>
> Federico G. Benavento
>
> ---
> /bin/fortune:
> Steal money you're a thief: steal a country you're a king.
>
> From: Pietro Gagliardi <pietro10@mac.com>
> Date: November 24, 2007 11:21:57 AM EST
> To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
> Subject: Re: [9fans] Attempt at making 3D graphics library
> Reply-To: Fans of the OS Plan 9 from Bell Labs <9fans@cse.psu.edu>
>
>
>
> On Nov 24, 2007, at 11:00 AM, andrey mirtchovski wrote:
>
>>> Use q to quit (and
>>> crash, for some threadexitsall-related reason).
>>
>> that usually means you need a bigger stack for your threads.
> Setting both threads to 4096 stacksize fixed it. I'll note that in  
> my tutorial. I used one thread to interpret mouse events, since for  
> some odd reason eresized() is sent through the mouse. :-(
>
>