| |
| |
| |
|
Page: 1 2
Comments:
<0> Heya.
<0> I'm trying to revive Animadead. :P I've been going back through my code and cleaning it up a bit.
<0> One thing I'm trying to clean up is my matrix math. I originally wrote it to support both row major and column major indexing.
<1> Nice Return0, how's the user base?
<0> hehe. well
<0> there's a handfull :P
<0> It got me a job though :)
<1> Good :)
<0> how have you been mal?
<1> Quite well, I'm surprisingly having much fun writing a... high-performance distributed database
<0> nce?
<0> nm
<1> Nce?
<0> line wrapped :P
<1> Mmhm :)
<0> neat
<0> how bad is the network overhead?
<1> Very low, the point is being able to get a direct pointer to locally cached memory in a handful of cycles
<1> processor cycles that is, nothing you can compare to pathetically slow SQL databases
<0> a handful of parallel cycles anyway ;)
<0> the future.
<0> I'd like to improve my matrix cl*** a bit. At the moment it has a static member that defines format the data (from the result of an operation) is packed in, row-major or col-major.
<1> Sounds rather inefficient to read this variable on each call
<0> exactly
<0> especially because someone wouldn't likely ever switch
<0> for a single app anyway
<0> So at first I thought maybe a preprocessor, but that's a bad idea.
<0> so then, maybe different cl***es. which makes sense.
<1> Have functions handle one specific order, yes
<0> but I'm still working out the details
<0> It would be more trivial if I was just using C-style functions
<1> Quite :)
<0> but I was hoping to make the interface nice and use the C++ operator overloading.
<1> I'm really not fond of C++'s OO, it's abused and inefficient
<0> yeah, I sort of agree.
<0> But I'm not quite to the point where I just want to use strictly functions
<1> Sure, use cl***es when it's appropriate ; otherwise, don't
<0> I think it's appropriate that's the problem :P
<0> for matrices.
<0> I'm thinking I can fix a lot by using two different cl***es for row major and column major
<0> but I dunno.
<1> Sounds reasonable
<0> One optimization I originally implemented was allowing multiplication of row major and column major matrices together, and just using special indexing to make it proper.
<0> Which would safe the overhead of converting a matrix.
<0> But then again, it's probably not necessary
<0> s/safe/save/
<2> why allow both matrix modes?
<2> just choose one :)
<0> well... because I want to make it easy to use in both directx and opengl
<0> for pushing a matrix on the stack
<0> And you can't just after the fact convert a matrix that's compiled from several matrix multiplications, from one format to the other.
<0> it reverses the order of the multiplications
<1> Hrm, no? It's just a different representation, the data is the same
<0> Which is sort of why it doesn't make much sense to support mixed mode.
<0> are you sure?
<0> hrmm
<0> maybe you are right
<2> (S*R*T)*v = v*(S*R*T)^transpose
<2> = v*(T*R*S)
<0> so I'm right, right?
<2> no.
<2> mal is.
<3> heh
<0> you just showed me how (TRS) = opposite format(SRT), did you not?
<2> I just showed that if you swap from vM (d3d) to Mv (gl) you just need to transpose M
<2> (or that M is transposed...)
<0> oh, I get it.
<2> by that operaion
<0> or do I
<2> write out a 2x2 matrix and a 2d vertex
<0> I already knew to go from one format to the other it's the transpose... but
<2> put the vertex on either side of the matrix
<2> do the mult.
<2> look at the transpose of the matrix, do it again.
<2> but?
<0> I guess I get it, so by transposing, and changing the indexing nullifies the transpose
<0> I didn't think of it that way
<0> I always just looked at it from one perspective
<0> So I should definitely pick one format, and then just allow the user to access it either way.
<2> Just "pick" the C way.
<0> in either format I mean
<2> its what people expect.
<2> (row-major)
<0> ok.
<0> got it ;)
<0> Thanks, I have a much clearer understanding now.
<2> :)
<3> I saw a math once... it skeered me
<1> I couldn't look up "skeer", is the orthography correct?
<3> 'scare'
<3> with hillbilly accent applied o.O
<1> Oh :)
<0> :)
<4> How do the glTexEnv functions work without multitexturing?
<4> Do glTexEnv decide how the texel is applied to the fragment, before the lighting equations are done?
<1> The lighting formula sets the primary color for the fragments, which is then used by the fragment pipeline ( glTexEnv )
<4> thanks
<4> I'm implementing emboss bumpmapping on the PSP. And I'm not used to just have one texture unit.
<2> I hate GL's multitexturing.
<4> Well, yes it can be confusing.
<2> mm, not confusing as much as poorly designed
<2> .. but I guess I have to blame that on the client/server crap built in
<4> And the combiners give me a headache.
<4> I don't see any obvious reason for the client/server model either.
<4> Afaik, only X has a working implementation.
<2> a software impl at that.
<2> But the ARB knows GL is out of date.
<2> Thats why GL3.0 has been proposed.
<0> Dr^Nick, I haven't heard anything yet about gl3.0, are they proposing some major changes or just shader 3.0 compliance?
<0> We're only using shader 3.0 where I work.
<2> They're rewriting all of GL
<2> essentially a new api.
<2> then they're going to layer the legacy crap on top of i
<2> t
<2> since it will have to go through an emulation layer anyway to run on next-gen hardware.
<0> I know if you transpose a rotation matrix, you get the inverse, but does this work for any transformation matrix? Can I just transpose the first 3x3 part, and negate the position part?
<0> I'm guessing it might not work for a projection matrix
<0> but for storing translation, rotation, and scale would it work?
<0> about negating the position part, I'm also guessing that's a wrong ***umption
<0> I figure I'll probably have to rotate the position values by the 3x3 matrix, and use the negation of that.
<0> I guess I could just make sample cases and check that I get the identity when I multiply a matrix by its inverse.
<4> You probably never will.
<4> Because of precision errors.
<4> Using an epsilon might work.
<5> where can i download opengl for c++ in linux(ubuntu distribution)?
<1> Download "opengl"? Install drivers from your vendor, if they don't come with Ubuntu
<5> i dont know how to do that
<5> call it "install" or "use" opengl, what actions should i take in order to be able to use opengl in a c++ program?
<1> Just download and install the appropriate drivers for your video card, does Ubuntu provide the opengl headers by default? ( /usr/include/GL )
<5> yes i found that folder
<1> I don't know much of Ubuntu packages, you can surely get the drivers there as well, which could be preferable
<5> does that mean it's already "installed"?
<1> Do you have gl.h and other files in there? You also need drivers if you want to get any hardware acceleration
<5> i got gl.h and a all those other files
<1> Run glxgears, wait 5 seconds and tell me what frame rate you get
<1> In case your drivers are already installed
<5> i cant find glxgears
<5> i find no binary files at all
<1> Not there, run glxgears from a xterm
<5> ah
<5> :)
<5> i get three wheels in Red green and blue
<1> Alternatively, run glxinfo and see what is the implementation you got
<5> 1.2 (1.5 Mesa 6.2.1)
<1> Okay, you don't have drivers installed and rely on software... which is slow
<1> Download and install drivers for your video card, which I hope for you is Nvidia, or ATI
<1> ( or find how to get that through the Ubuntu package system )
<5> its Nvidia
<1> Then you'll have no problem at all, good
<5> i got one on download now
Return to
#opengl
or
Go to some related
logs:
sata install XP nx6320
#nhl
#sendmail
#delphi
#politics
#politics
ichigo's zanpaktou
+Jira Relay access denied
#beginner
#python
|
|