Photos Blog Random Things Links Guest Book Resume : #Lecture

D3D8 - Intro to Vertex and Index Buffers

Session Start: Wed Jan 08 00:44:06 2003
[00:44] DCC Chat connection established
[00:44] <Error404> Sorry, I was reading logs.
[00:44] <RedBeard> gimme a min
[00:45] <Error404> ok.
[00:45] <RedBeard> i recommend you go through the code and find every single Direct3D function call, and look it up in the d3d docs to figure out what all the parameters are for
[00:46] <Error404> Ok.
[00:47] <Error404> msdn or just the ones that came with d3d8?
[00:47] <Error404> I have the ones open that came with the SDK.
[00:47] <Error404> They are usually open whenever i attempt to code dx
[00:47] <RedBeard> the d3d docs that come with the sdk
[00:48] <Error404> completely start over my d3d class?
[00:48] <RedBeard> not yet
[00:48] <RedBeard> lemme modify your initd3d function
[00:48] <RedBeard> uh, initvb
[00:49] <Error404> Ok.
[00:50] <RedBeard> bool cDXanadu::InitVB()
[00:50] <RedBeard> {
[00:50] <RedBeard> StartLog("cDXanadu::InitVB");loc;
[00:50] <RedBeard> if(FAILED( pd3dDevice->CreateVertexBuffer(50*3*sizeof(CUSTOMVERTEX), 0,
[00:50] <RedBeard> &vb)))
[00:50] <RedBeard> return false;
[00:50] <RedBeard> CUSTOMVERTEX * pVertices;
[00:50] <RedBeard> if(FAILED( vb->Lock(0, 0,(BYTE**)&pVertices,D3DLOCK_DISCARD)))
[00:50] <RedBeard> return false;
[00:50] <RedBeard> DWORD i;
[00:50] <RedBeard> FLOAT theta, theta2;
[00:50] <RedBeard> for(i=0;i<50;i+=2)
[00:50] <RedBeard> {
[00:50] <RedBeard> theta = (2*D3DX_PI*i)/(50-1);
[00:50] <RedBeard> theta2= (2*D3DX_PI*(i+1))/(50-1);
[00:50] <RedBeard> pVertices[3*i+0].position = D3DXVECTOR3 (sinf(theta), -1.0f,cosf(theta));
[00:50] <RedBeard> pVertices[3*i+1].position = D3DXVECTOR3 (sinf(theta), 1.0f,cosf(theta));
[00:50] <RedBeard> pVertices[3*i+2].position = D3DXVECTOR3 (sinf(theta2),-1.0f,cosf(theta2));
[00:50] <RedBeard> pVertices[3*i+3].position = D3DXVECTOR3 (sinf(theta), 1.0f,cosf(theta));
[00:50] <RedBeard> pVertices[3*i+4].position = D3DXVECTOR3 (sinf(theta2), 1.0f,cosf(theta2));
[00:50] <RedBeard> pVertices[3*i+5].position = D3DXVECTOR3 (sinf(theta2),-1.0f,cosf(theta2));
[00:50] <RedBeard> }
[00:50] <RedBeard> vb->Unlock();
[00:50] <RedBeard> return true;
[00:50] <RedBeard> }
[00:51] <RedBeard> no idea if that will work properly
[00:52] <RedBeard> i can't compile the app here, dont have certain headers
[00:52] <RedBeard> then in DoFrame, use this line in place of what's there
[00:52] <RedBeard> pd3dDevice->DrawPrimitive(D3DPT_TRIANGLELIST,0,2*50-2);
[00:54] <Error404> Err
[00:54] <Error404> Well.. it animated
[00:54] <Error404> but the cylinder is weird
[00:54] <Error404> and faces are everywhere
[00:54] <RedBeard> hrmm
[00:55] <Error404> Do I have to know the equation to design a cynlinder in a for loop to be a good dx programmer?
[00:55] <Error404> cylinder.
[00:55] <RedBeard> heh not really
[00:55] <RedBeard> but you need to know how the geometry is arranged
[00:56] <RedBeard> so that it makes a surface
[00:56] <RedBeard> i'd like to get you started on an index buffer method, but it would be a bit complicated to add that on-the-fly
[00:56] <Error404> I can rewrite the entire direct x class quick like
[00:56] <Error404> I've rewritten it over and over for the past week now.
[00:57] <RedBeard> heh hold on a sec
[00:57] <RedBeard> give it an IDirect3DIndexBuffer8 * ib
[00:57] <Error404> The current or new class?
[00:58] <RedBeard> whichever you're working with
[00:58] <RedBeard> mebbe start over with a sample app that works
[00:58] <RedBeard> and modify that
[00:58] <RedBeard> or sample code
[00:58] <Error404> Ok let me go back a few tutorials.
[00:58] <RedBeard> wait, lemme see if i can compile this stuff
[00:59] <Error404> I went back to the first file set I sent you.
[00:59] <Error404> It was working, it just draws a triangle with diff vertex shades.
[01:00] <RedBeard> mm ok
[01:00] <RedBeard> send me those again?
[01:00] <RedBeard> and i'll try to help you expand it
[01:01] <Error404> Ok.
[01:01] <Error404> If its possible can we open a DCC with the other computer, it would allow me to chat and code at the same time.
[01:01] <Error404> Actually nevermind.
[01:02] <Error404> This works.
[01:02] <Error404> Helps for copynpaste

-- switch from one DCC to another --

[01:02] <Error405> thanks
[01:03] <Error405> In the clean up I can just do if(!vb) right?
[01:03] <Error405> Since NULL is 0.
[01:04] <RedBeard> yes
[01:04] <RedBeard> it's technically correct, although some people like you to explicitly compare against NULL
[01:04] <RedBeard> i say don't worry about that unless it's not clear that you're using a pointer
[01:05] <Error405> Yeah. that makes since.
[01:05] <Error405> sense
[01:05] <RedBeard> using NULL makes it more readable when using pointers
[01:05] <Error405> So people aren't guessing what i'm doing. No 'clever code'.
[01:06] <RedBeard> right
[01:06] <RedBeard> don't try to outwit the compiler, it knows what it's doing
[01:06] <RedBeard> oh and another thing, don't EVER use the incrementer operator inside other code
[01:06] <RedBeard> it causes much pain when used as such
[01:06] <Error405> I try to force good coding practice on myself so I wont write bad code.
[01:06] <Error405> inside other code?
[01:06] <RedBeard> array[i++] = 12;
[01:07] <Error405> Oh.
[01:07] <RedBeard> instead, do: array[i] = 12; i++;
[01:07] <Error405> ah i see.
[01:07] <RedBeard> it's clearer and less bug-prone
[01:09] <RedBeard> alright, so what are ya gonna do?
[01:09] <Error405> Well if you think it'd be easier to restart, I can do that and remake the class.
[01:09] <Error405> Or I can use the class I gave you, which is really simple I think.
[01:10] <RedBeard> nah it's not worth tearing it down
[01:11] <RedBeard> IDirect3DIndexBuffer8 *ib; <== add that
[01:11] <Error405> got that
[01:11] <RedBeard> now you gotta create and initialize the vb to be used as being indexed
[01:11] <RedBeard> with a single triangle, that's no different
[01:11] <RedBeard> so now, create and init the index buffer
[01:12] <RedBeard> CreateIndexBuffer(...)
[01:12] <Error405> alright.
[01:12] <Error405> ok.. umm.
[01:12] <RedBeard> 3 indices, D3DUSAGE_WRITEONLY
[01:12] <Error405> Do I change the &vb to ib>?
[01:13] <RedBeard> you need a vertex buffer AND an index buffer, so you're making both still, right?
[01:13] <Error405> oh i do?
[01:13] <RedBeard> so in calls dealing with the index buffer, use "ib"
[01:13] <Error405> ah ok
[01:13] <RedBeard> yes, the index buffer just stores integer (short, 16-bit) indices
[01:14] <RedBeard> and the vertices are still stored in the vertex buffer
[01:14] <RedBeard> D3DFMT_INDEX16
[01:14] <RedBeard> default pool
[01:15] <Error405> ok.. ah please slow.. lol.. For the createindexbuffer do I keep it the same as it was just change the second variable to D3DUSAGE_WRITEONLY correct.. Ok I got that.
[01:15] <RedBeard> look it up in teh docs to see what im talking about with these values
[01:16] <Error405> yeah that is where i found out where to put writeonly
[01:16] <Error405> lol
[01:16] <Error405> Ok and I'm keeping the current vertice setup the same.
[01:16] <Error405> with the triangle.
[01:16] <Error405> So I still keep the 3*sizeof(customvertex)
[01:17] <RedBeard> right
[01:17] <Error405> Ok, now I leave the rest alone and set up the indexbuffer?
[01:17] <RedBeard> put in another thing there, make a new member variable num_tris
[01:17] <Error405> so all of this is happening in InitVB?
[01:17] <RedBeard> and make it num_tris*3*sizeof(customvertex)
[01:17] <RedBeard> yes
[01:18] <RedBeard> oh, and for the size of the index buffer...
[01:18] <Error405> just leave num_tris 1 for now?
[01:18] <RedBeard> sizeof(short)*num_tris
[01:18] <RedBeard> yes
[01:18] <RedBeard> actually, make another member var, num_verts
[01:19] <Error405> 3?
[01:19] <RedBeard> and the index buffer will deal with num_tris, and the vertex buffer will deal with num_verts
[01:19] <RedBeard> yes
[01:19] <RedBeard> remember, member var, not local var
[01:19] <Error405> Oh member vart
[01:19] <Error405> var
[01:19] <Error405> lemme move them
[01:20] <Error405> ok for the vertex buffer it isn't num_tris*3*sizeof.. it's num_verts
[01:20] <RedBeard> yes
[01:20] <RedBeard> and take out the 3
[01:20] <Error405> Ok when I call CreateIndexBuffer I should do it before or after the Lock?
[01:21] <RedBeard> the index and vertex buffers are completely independent
[01:21] <RedBeard> you have to create it before you lock it, but you can create and lock one, then create and lock the other
[01:21] <RedBeard> doesnt matter
[01:21] <Error405> alright
[01:22] <Error405> ok index buffer is..err.. damn... num_tris*3*sizeof(customvertex)?
[01:22] <Error405> oh you rtold me that
[01:22] <Error405> sorry
[01:23] <RedBeard> sizeof(short)*num_tris
[01:23] <Error405> oo short
[01:23] <Error405> .. why a short?
[01:23] <RedBeard> because you're using 16-bit indices
[01:23] <RedBeard> they're smaller, and you're using less than 65,536 vertices
[01:24] <RedBeard> also, your hardware probably doesn' support 32-bit indices
[01:24] <Error405> ok i understand that, but hmm. It's 1 triangle. I don't need 3 anywhere?
[01:24] <RedBeard> my gf2 doesnt
[01:24] <RedBeard> oh wait yeah my bad
[01:24] <RedBeard> sizeof(short)*num_tris*3
[01:24] <Error405> ah ok
[01:24] <RedBeard> :P
[01:25] <RedBeard> so, that's created. now to lock it and fill values in
[01:25] <Error405> ok the index buffer's usage, it isn't writeonly right?
[01:25] <RedBeard> yes, it is
[01:25] <RedBeard> you don't need to read values back out of it
[01:25] <Error405> oh.. ok
[01:25] <RedBeard> if you ever think you'll need to re-read the values in a vertex or index buffer, or texture, or anythng, just keep a copy of it around in your code
[01:25] <RedBeard> don't read stuff back through the API
[01:26] <Error405> it's slow right?
[01:26] <RedBeard> very
[01:27] <RedBeard> we can take this back to #silicon3d
[01:27] <RedBeard> since i'm not pasting stuff anymore
[01:27] <Error405> ah ok

-- switch to #silicon3d --

[01:34] <Error405> aight
[01:34] <RedBeard> ok, so now you have the index buffer, lock it
[01:35] <RedBeard> and then treat the locked data as an array of unsigned shorts
[01:35] <Error405> i'm trying to figure out what to memcpy to it.. one moment.
[01:35] <RedBeard> uint16 might work
[01:35] <RedBeard> and fill in 1,2,3
[01:35] <RedBeard> no wait... 0,1,2
[01:35] <Error405> wait
[01:35] <Error405> i don't have to memcpy
[01:35] <Error405> ok
[01:36] <RedBeard> you can do the same thing with the vertex buffer if you want, casting the pointer to another CUSTOMVERTEX pointer
[01:36] <RedBeard> and treating it as an array
[01:36] <RedBeard> but the initializer thingy is easy to do
[01:36] <RedBeard> so keep it as-is
[01:38] <Error405> ok. sorry but I'm having trouble figuring out what to pass in for the size of the index buffer to lock
[01:38] <Error405> would it still be g_vertices?
[01:39] * Quits: Exellon (guest123@ (Today is a good day to chat.)
[01:40] <RedBeard> passing 0 locks the whole thing
[01:40] <Error405> so i do want to lock the entire thing
[01:41] <Error405> do I want to do the same with the vertexbuffer since it's already the correct size.
[01:41] <RedBeard> but you wanna lock the full size, so you can either use 0 or sizeof(short)*3*num_tris
[01:44] <Error405> ok so i'm completely guestimating this, but i'm hoping i'm right, do I memcpy(pVertices,g_Vertices,sizeof(g_Vertices))?
[01:44] <Error405> I have a feeling I need to do something else
[01:44] <Error405> something dealing with uint16 like you said
[01:45] <Error405> damnit I wish I knew exactly what the hell I was doing, I hate being clueless.
[01:45] <RedBeard> i recommend you use sizeof(CUSTOMVERTEX)*num_verts
[01:45] <Error405> do I still use g_Vertices?
[01:45] <RedBeard> that's for the vertex buffer
[01:46] <RedBeard> for the index buffer, makea uint16 pointer
[01:46] <Error405> ok...
[01:46] <Error405> unsigned short?
[01:46] <Error405> or uint16 (is confused, thought they were the same)
[01:47] <RedBeard> same thing
[01:48] <Error405> holy flippin god i'm confused now.
[01:48] <Error405> Ok i have this pointer to an ushort
[01:48] <RedBeard> unsigned short * pIndices;
[01:48] <Error405> yeah made that
[01:49] * Quits: masterg ( (night)
[01:50] <RedBeard> if(FAILED( ib->Lock(0, sizeof(short)*3*num_tris, (BYTE**)&pIndices,0)))
[01:50] <RedBeard>         return false;
[01:50] * Joins: Washu[AWAY] (
[01:50] <X2> [Washu[AWAY]] <nps> masterg: what are you doing this weekend? <A*sassin> HOT SEX ON HIS PARENTS' BED! <A*sassin> uh, i mean... nothing <Masterg> omg, shhh!
[01:50] * X2 sets mode: +v Washu[AWAY]
[01:50] <RedBeard> that's your lock call
[01:50] <Error405> yeah I figured that out
[01:50] <Error405> after much thought and hair pulling.
[01:50] <RedBeard> now you can go... pIndices[0] = 0
[01:50] <RedBeard> and same for 1 and 2
[01:50] <Error405> whoa
[01:51] <Error405> That's different.
[01:51] <RedBeard> well, i wouldnt be teaching you anything if there wasnt thought involved :P
[01:51] <Error404> pIndices[0] = 0;pIndices[1] = 1;pIndices[2] = 2;
[01:51] <RedBeard> yes, you can treat the memory as an array
[01:52] <Error404> ?
[01:52] <RedBeard> pretty neat huh
[01:52] <Error404> yes.
[01:52] <Error404> This entire time I've been trying to figure out what the hell i'm memcpy'in
[01:52] <RedBeard> that's essentially what a vertex and index buffer store
[01:52] <RedBeard> an array of vertices or indices
[01:52] <RedBeard> and if you treat it in the correct manner, you can use it as such
[01:53] <Error404> awesome... now is that the line after ->lock?
[01:53] <RedBeard> yeah
[01:53] <RedBeard> and then you can unlock
[01:53] <RedBeard> i think that's all for that function
[01:53] <Error404> yay!
[01:53] <Error404> Ok wait.
[01:54] <Error404> I need to make sure I didn't write anything that will format my computer (j)
[01:54] <Error404> if(FAILED( pd3dDevice->CreateIndexBuffer(sizeof(short)*num_tris*3,D3DUSAGE_WRITEONLY,D3DFMT_INDEX16, D3DPOOL_DEFAULT, &ib)))
[01:55] <Error404> is to create it right?
[01:55] <RedBeard> just a slight optimization thing...
[01:55] <Error404> hm?
[01:55] <RedBeard> im not sure if MSVC is smart enough to resolve the syntax tree when constants are on opposite sides of a variable
[01:55] <RedBeard> ie, sizeof(short)*num_tris*3
[01:55] <RedBeard> so put constants together
[01:55] <RedBeard> just as a coding style thin
[01:56] <RedBeard> sizeof(short)*3*num_tris
[01:56] <Error405> ah thank you...
[01:56] <RedBeard> that works out to 6*num_tris
[01:56] <RedBeard> which is 1 less multiply
[01:56] <Error405> so the compiler will do it beforehand since they are constants.
[01:56] <RedBeard> im not sure if msvc will optimize it properly the otherway
[01:56] <Error405> neat, didn't know that.
[01:57] <Error404> ok.. i'm renaming this function, InitVB seems a little wrong now
[01:57] <RedBeard> heh
[01:58] <RedBeard> InitGeometry perhaps
[01:58] <Error404> InitBuffers is what i got... but Geometry seems smarter
[01:58] <Error404> lol
[01:58] <RedBeard> buffers works too
[01:58] <RedBeard> not a big deal
[01:59] <Error405> The compiler cares!
[01:59] <Error405> hehe ah well ok what's next?
[01:59] <Error405> wait wait lemme review this function
[01:59] <RedBeard> next, you have to modify your drawing code to use the index buffer
[02:00] <RedBeard> make VB, make IB, lock VB, lock IB, fill in vertex data, fill in index data, unlock both
[02:00] <RedBeard> roughly in that order
[02:01] <Error405> is it ok if you have lock vb, unlock, lock ib, unlock?
[02:01] <Error405> with fills between?
[02:01] <RedBeard> yeah no problem
[02:01] <Error405> ah ok good
[02:02] <Error405> ok now change drawing function to use the ib lemme see if i can't figure this out
[02:02] <RedBeard> it involves SetIndices
[02:02] <RedBeard> just as a starter hint
[02:03] <Error405> err.. ok lemme find that
[02:04] <Error405> Ok the call would be pd3dDevice->SetIndices(ib,0) but it says something about making memory leaks. so would I set it to pd3dDevice->SetIndices(NULL,NULL) afterwards?
[02:05] <RedBeard> possibly
[02:05] <RedBeard> not really a problem
[02:05] <RedBeard> just do that on program shutdown
[02:05] <Error405> it's not?
[02:05] <Error405> oh.
[02:05] <Error405> ok
[02:05] <RedBeard> because if you keep using the same index buffer, or even different ones, there's no need to set it to NULL all the time
[02:06] <Error405> ok lets see. so i set the indices then set the streamsource?
[02:06] <RedBeard> yeah, don't have to do that in any particular order
[02:06] <RedBeard> just set both before calling DrawIndexedPrimitive
[02:07] <Error405> ah another hint
[02:07] <Error405> one moment
[02:07] <RedBeard> heh, i hope im helping without bludgeoning you with the stuff
[02:08] <RedBeard> just it's a lot easier when you know which functions you need, it's much less tedious
[02:08] <Error405> i would have never figured it out
[02:08] <Error405> I don't know the difference lol
[02:09] <Error405> ok.. i see how this works but I'm a little confused with one thing.
[02:09] <Error405> I understand that i'm using the IndexBuffer now, but how do the coordinate information get rendered if the ib was just fed 0,1,2?
[02:09] <RedBeard> what are you giving to SetStreamSource?
[02:10] <Error404> d3dDevice->DrawIndexedPrimitive(D3DPT_TRIANGLELIST,0,num_verts,0,num_tris);
[02:10] <Error404> pd3dDevice->SetStreamSource(0,ib,sizeof(CUSTOMVERTEX));
[02:10] <Error404> should i keep it vb>
[02:10] <Error404> i thikn i should eh?
[02:10] <RedBeard> yes
[02:10] <RedBeard> streams are used for vertex data
[02:10] <RedBeard> indices get their own function
[02:11] <Error404> ah.
[02:11] <RedBeard> no need to set the same data more than once
[02:11] <Error404> ok so what is thedifference?
[02:11] <RedBeard> make sense?
[02:11] <Error404> I see how it works, somewhat. But I don't see the reasoning behind it.
[02:11] <RedBeard> so now you're giving the card vertices through SetStreamSource
[02:11] <RedBeard> and you're giving it a set of indices through SetIndices
[02:12] <RedBeard> now the card goes through the index array, and collects them in sets of 3 (triangle list)
[02:12] <RedBeard> and for each 3, it goes and grabs the vertices at the index indicated
[02:12] <Error404> So in this case I could make lots of triangles?
[02:12] <RedBeard> so in this case, it grabs vertex 0, vertex 1, vertex 2
[02:12] <RedBeard> and makes a triangle from them
[02:12] <RedBeard> then it renders that triangle
[02:13] <RedBeard> you can make 2 triangles by adding 1 vertex and 3 indices
[02:13] <Error404> Just increase num_tris and then transform it then I'd have lots of triangles using only one vertex buffer with 3 coords?
[02:13] <Error404> transform each on that is..
[02:13] <Error404> If that works.
[02:13] <Error404> (is hoping thats a reason)
[02:13] <RedBeard> well, if you only have 3 coords then you can only make 1 triangle
[02:14] <Error404> o
[02:14] <Error404> Ok. so it just seperates the data flow?
[02:14] <Error404> Oh wait I think I get it.
[02:14] <RedBeard> and you can't set the transformation matrices more than once per DP call (DrawPrimitve)
[02:15] <RedBeard> dont worry about transforms just yet though
[02:15] <Error404> Ok.. so what that one DrawIndexedPrimitive the right call?
[02:15] <Error404> d3dDevice->DrawIndexedPrimitive(D3DPT_TRIANGLELIST,0,num_verts,0,num_tris);
[02:16] <Error404> i'm drawing a triangle list, with 0 min indeces (i don't know why), with 3(num_verts), 0 as the starting, with 1(num_tris)?
[02:17] <RedBeard> ok
[02:17] <Error404> I used 0 in SetIndices, so I would use 0 in DIP as well?
[02:17] <RedBeard> and you're initializing num_tris to 3 and num_verts to 3*num_tris, right?
[02:17] <RedBeard> ah, those are a bit different
[02:17] <RedBeard> just use 0 for now, i can explain what they do if you want
[02:17] <Error404> Err I thought num_tris was 1 and num_verts was 3?
[02:18] <RedBeard> uhm yes, num_tris to 1, num_verts to 3*num_tris
[02:18] * RedBeard is a bit tired
[02:19] <Error404> is num_verts just 3 or 3*num_tris
[02:19] <RedBeard> 3*num_tris
[02:19] <Error404> Cause I know there are 3 vertices in a triangle..
[02:19] <Error404> Ah.. hrm ok.
[02:19] * Joins: Stinger (tcoleman@
[02:19] * X2 sets mode: +v Stinger
[02:19] <Error404> hm.. ah hmm. oh I see why.
[02:19] <Error404> Ah!@
[02:19] <Error404> ok
[02:19] <RedBeard> hrmm wait
[02:20] <RedBeard> no no, im talking rot
[02:20] <RedBeard> just make it 3
[02:20] <Error404> So if I increase the number of triangles then it'd increase it.
[02:20] <Error404> It makes sense
[02:20] <RedBeard> number of verts and number of tri's aren't really related
[02:20] <Error404> It does, really.
[02:20] <Error404> oh
[02:20] <Error404> damn
[02:20] * Error404 hits head
[02:20] <RedBeard> you can have 4 verts and 2 tris
[02:20] <RedBeard> using 6 indices
[02:20] <Error404> o
[02:20] <RedBeard> number of indices = 3*num_tris
[02:21] <Error404> ..
[02:21] <Error404> one moment, computing
[02:21] <RedBeard> but since that is always true, you can get away with only storing num_tris
[02:21] <RedBeard> but num_verts is stored separately
[02:21] <RedBeard> because it's an independent value
[02:21] <Error404> num_tris = 1;
[02:21] <Error404>     num_verts = 3;
[02:21] <RedBeard> yeah
[02:21] <Error404> ok.
[02:21] <RedBeard> see if it compiles
[02:21] <Error404> alrighty
[02:22] <Error404> err some errors, lemme debug
[02:23] <Error404> compiled.
[02:23] <Error404> .. i guess i run it eh?
[02:23] <RedBeard> yeah give it a go
[02:24] <Error404> I HAVE A FRICKEN TRIANGLE!!
[02:24] <Error404> MWHAHA
[02:24] <RedBeard> a word of caution when using index buffering:
[02:24] <Error404> hm?
[02:24] <RedBeard> it is entirely possible to fill in a bad index value
[02:24] <RedBeard> like 5637
[02:24] <RedBeard> or whatever
[02:24] <RedBeard> and it is likely that your computer will bluescreen
[02:25] <RedBeard> so take the utmost care when building index buffers
[02:25] <Error404> fill in bad index value? like putting in a wrong number in the array?
[02:25] <RedBeard> yes
[02:25] <Error404> ah, sorta if you for-loop it and it overwrites?
[02:25] <RedBeard> make sure you dont put in values that will try to reference non-existent vertices
[02:25] <Error404> Ok so lemme try to get this
[02:26] <Error404> The vertices i set up with err.
[02:26] <Error404> g_Vertices corisponds to the indeces
[02:26] <RedBeard> uh kinda yeah
[02:26] <Error404> So however many vertices i place in g_vertices there will be that many indeces?
[02:26] <RedBeard> the indices index into the vertex buffer
[02:27] <RedBeard> not quite
[02:27] <Error404> please explain?
[02:27] <RedBeard> you just want to make sure that you don't put values into your index buffer that are greater than the number of vertices you're using
[02:27] <RedBeard> ie, don't go pIndices[whatever] = 3452; if you only have 10 vertices
[02:27] * Error404 feels like making a class with error checking
[02:28] <RedBeard> yeah i'd recommend having some kind of error check in debug mode if you're wary of such things
[02:28] <RedBeard> but if you're just hardcoding things for now, it's not a big problem
[02:28] <RedBeard> just keep it in mind - if your program causes a bluescreen, it's likely an indexing problem
[02:28] <Error404> Ok can you tell me what the 0 does in SetIndices and DrawIndexedPrimitive?
[02:29] <RedBeard> i'll explain the DIP one first
[02:29] <Stinger> uh sdk?
[02:29] <Error404> alright.
[02:29] <RedBeard> MinIndex
[02:29] <RedBeard> and NumIndices
[02:29] <RedBeard> basically, that's just an optimization thing
[02:29] <Error404> NumVertices?
[02:30] <Error405> just making sure we're looking at the same function
[02:30] <RedBeard> if the indices that you're using are in the range from 10-20, and don't reference any vertices in the 0-9 range, then you can use MinIndex of 10
[02:30] <Error405> Oh!
[02:30] <Error405> sorta like ofsetting during a lock?
[02:30] <RedBeard> so the driver only has to load the vertices it will actually use
[02:30] <RedBeard> not quite, but similar
[02:31] <RedBeard> note that it's the minimum index VALUE, not the minimum index index
[02:31] <RedBeard> the starting index is different
[02:31] <Error405> ok what do you mean the value?
[02:31] <RedBeard> so if you're drawing a quad using index values (10,11,12),(12,11,13), then you can use MinIndex of 10
[02:31] <Error405> because..err.. wait
[02:32] <RedBeard> because you aren't using any vertices below that index value
[02:32] <Error404> so if I had many shapes within the vertex buffer I could choose which ones to draw?
[02:32] <Error404> pIndices[0] = 0;    pIndices[1] = 1;    pIndices[2] = 2;
[02:33] <Error404> are those the values that you're talking about?
[02:33] <RedBeard> yes
[02:33] <RedBeard> so in that case, your minimum index is 0
[02:33] <RedBeard> so use 0
[02:33] <Error404> And those corispond to the vertex buffer right?
[02:33] <Error404> { 150.0f,     50.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(255,255,255,255) },
[02:33] <Error404>         { 250.0f,    250.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(0,0,0,128) },
[02:33] <Error404>         { 50.0f,    250.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(255,0,0,64) },
[02:33] <RedBeard> right, they index into the vbuffer
[02:33] <Error404> g_Vertices[0]-2
[02:33] <Error404> Awesome
[02:34] <Error404> It makes sense now.
[02:34] <RedBeard> StartIndex
[02:34] <RedBeard> you can store lots of indices into an index buffer, and some might pertain to different objects
[02:34] <Error404> Is the actual index to start out?
[02:34] <Error404> err with
[02:35] <RedBeard> so you can use 12 indices for making a set of 4 triangles, and then another 12 for another set
[02:35] <RedBeard> but you want to change some state between the renders
[02:35] <RedBeard> say, change the texture
[02:35] * Quits: SGT-T ( (Read error to SGT-T[]: Connection reset by peer)
[02:35] <RedBeard> you can store all the indices into the same index buffer, and then give it an offset to start at
[02:35] * Joins: Kicknfast (
[02:35] * X2 sets mode: +v Kicknfast
[02:35] <RedBeard> so the first time you'd use 0, and the 2nd time you'd use 12
[02:35] <RedBeard> make sense?
[02:36] <Error404> So if i have a triangle in the vb in the index 2-4 then another in 50-52 i'd use StartIndex of 50 to make that triangle?
[02:36] <Error404> of course I could just set the value to 50-52 when initializing right?
[02:36] <RedBeard> uhm, i think so
[02:37] <RedBeard> NumVertices i think is used in a similar fashion to MinIndex
[02:37] <RedBeard> essentially, it indicates the MaxIndex
[02:37] <RedBeard> except it's relative to MinIndex
[02:37] <Error404> so if the IndexBuffer was technically initialized with for(int i,i<50,i++)pIndices[i]=i I would just keep MinIndex and StartIndex the same right?
[02:37] <RedBeard> so NumVertices = MaxIndex - MinIndex
[02:38] <RedBeard> no, MinIndex and StartIndex are different things entirely
[02:38] <Error404> oh
[02:38] <RedBeard> for the time being, you can just use 0 for MinIndex and num_verts for NumVertices, in all your calls
[02:38] <Error404> lemme reread what you said.
[02:39] <RedBeard> but StartIndex corresponds to which index in your index buffer you want to start with
[02:39] <RedBeard> so, using indices=(0,1,2, 3,4,5)
[02:39] <RedBeard> you can draw the triangle 3,4,5 by setting StartIndex=3
[02:40] <Error404> Oh.
[02:40] <RedBeard> that's a kinda bad example, coz the numbers match
[02:40] <Error404> So really if I had just had a huge friken list of coords in the vertex buffer.
[02:40] <Error404> I could join ones generally next to each other in value
[02:40] <RedBeard> so, using indices=(12,15,22, 22,6,80) you still use StartIndex=3 to draw the 2nd triangle (22,6,80)
[02:40] <Error404> even if they don't make a shape or whatnot, just for example.
[02:41] <RedBeard> yes, that's a part of state-change optimization
[02:41] <Error404> Yeah it's the index.
[02:41] <RedBeard> ok, so now... add another vertex
[02:41] <Error404> Ok.
[02:42] * Quits: Stinger (tcoleman@ (Ping timeout for Stinger[])
[02:42] <Error404> I'm trying to remember how the coords look lol
[02:42] <Error404> It's in LHS.. right?
[02:43] <RedBeard> { 150.0f,    450.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(0,255,0,64) },
[02:43] <RedBeard> and add 3 more indices
[02:43] <RedBeard> (2,1,3) in that order
[02:44] <Error404> huh.. oh
[02:44] <Error404> oh. OH
[02:44] * Error404 hits head
[02:44] <Error404> Now it makes sense
[02:44] <Error404> It's the order you're drawing the polygon
[02:44] * Error404 kicks self
[02:45] <RedBeard> note that i'm using a clockwise ordering, since that's what the original sample you sent me uses
[02:45] <Error404> Do i change num_tris and num_verts..
[02:46] <Error404> I put num_verts to 4..
[02:46] <RedBeard> you could turn on culling... D3DCULL_CCW
[02:46] <RedBeard> since you're using clockwise winding, you can cull counter-clockwise faces
[02:46] <RedBeard> dont worry about culling just yet though
[02:47] <Error404> pIndices[0] = 0;    pIndices[1] = 1;    pIndices[2] = 2;
[02:47] <Error404>     pIndices[3] = 2;    pIndices[4] = 1;    pIndices[5] = 3;
[02:47] <Error404> I've read the entire section of d3d in Jim Adam's book, i just couldn't implement any of the knowledge.
[02:47] <Error404> is that about right?
[02:47] <RedBeard> looks good
[02:47] <RedBeard> and remember to set num_tris to 2
[02:47] <Error404> heh got that.
[02:47] <Error404> should be a quad
[02:47] <RedBeard> kinda
[02:48] <Error404> Now i understand what you said up there.
[02:48] <Error404> I changed the coords
[02:48] <Error404> to make a square
[02:48] <Error404> hopefully
[02:48] <Error404> YAR!!
[02:48] <Error404> I gots square!
[02:48] <RedBeard> :)
[02:48] <RedBeard> ok, now try turning on culling and see what it looks like
[02:49] <Error404> oh it was on
[02:49] <RedBeard> ok, flip it to use D3DCULL_CW
[02:49] <Error404> Or at least I thikn it was
[02:49] <Error404> I gots no square!
[02:49] <Error404> haha
[02:49] <RedBeard> alright
[02:49] <RedBeard> now change the indices to be 1,2,3 instead of 2,1,3
[02:49] <RedBeard> and then try both culling modes
[02:50] <Error404> ha that's cool
[02:51] <RedBeard> that's backface culling
[02:51] <RedBeard> handled by the video card
[02:51] * Error404 pats his gf
[02:51] <Error404> this is awesome
[02:51] <Error404> i feel like coding again
[02:51] * RedBeard wishes there were someone around to teach him the things he wants to know
[02:52] <Error404> so with pd3dDevice->SetIndices(ib,0); i can sorta make a triangle out of that vb? right?
[02:52] <Error404> Dude if you need help with guitar I'd help anyday
[02:52] <RedBeard> heh i meant programming stuff
[02:52] <Error404> Oh yeah you haven't told me what that does
[02:52] <RedBeard> dont have me a gi-tar
[02:52] <RedBeard> heh
[02:52] <Error405> HRESULT DrawIndexedPrimitive(
[02:52] <Error405> D3DPRIMITIVETYPE Type,
[02:52] <Error405> UINT MinIndex,
[02:52] <Error405> UINT NumVertices,
[02:52] <RedBeard> ok, the 2nd param of SetIndices
[02:52] <Error405> oops
[02:52] <Error405> yeah
[02:53] <Error405> that one
[02:53] <Error405> i hate copying on one computer then pasting on the other
[02:53] <RedBeard> that's the "index offset"
[02:53] <Error405> damn clipboard should be shared damnit
[02:53] <RedBeard> BaseVertexIndex
[02:53] <RedBeard> and essentially, it increases the value of all your indices
[02:53] <RedBeard> take note that it is a 32-bit value
[02:53] <RedBeard> not 16-bit
[02:53] <Error405> hm?
[02:53] <Error405> yeah
[02:53] <Error405> unit
[02:54] <Error405> int
[02:54] <RedBeard> you can only draw 65,546 things at a time
[02:54] <RedBeard> 65536
[02:54] <Error404> really?
[02:54] <Error404> doesn't seem like that much =(
[02:54] <RedBeard> heh, it is
[02:54] <RedBeard> that's per DP call
[02:55] <RedBeard> but you can store more than that many things into a vertex buffer
[02:55] <RedBeard> so if you want to access the vertex at element 65,537 let's say
[02:55] <Error404> ohh
[02:55] <RedBeard> you can put an index of 65535 into your index buffer
[02:55] <RedBeard> and then use 2 in BaseVertexIndex
[02:55] <Error404> and then my other code will stil work
[02:55] <RedBeard> or vice versa
[02:55] <Error404> if they were all triangles for instance
[02:56] <RedBeard> yeah
[02:56] <Error404> ok, so lets say that i only want to draw three of these vertices here to make a triangle instead of a quad.
[02:56] <RedBeard> so you can use the same index buffer, and index into separate parts of the same vertex buffer using the same indices
[02:57] <RedBeard> so you have your quad-drawing index buffer (0,1,2, 2,1,3) right?
[02:57] <RedBeard> you can put 8 vertices into the vertex buffer, grouped in 4's
[02:57] <Error404> and an index buffer can only have 65546 indices?
[02:57] <RedBeard> 65536
[02:57] <Error404> yeah?
[02:57] <Error404> ah 36
[02:57] <RedBeard> on a gf2 thats the max you can draw
[02:58] <RedBeard> you can use StartIndex to go to a higher index in the buffer
[02:58] <RedBeard> so with your 8 vertices in the vbuffer, you have 2 quads
[02:58] <RedBeard> you can then go... SetIndices(ib,0); DIP(...); SetIndices(ib,4); DIP(...);
[02:58] <RedBeard> and the DIP calls will be the same
[02:59] <Error404> ah yeah and it would draw both
[02:59] <RedBeard> pretty cool huh
[02:59] <Error404> yeah that is cool.
[02:59] <RedBeard> so you don't have to rearrange your vertices or your indices, you just set 1 value in a function call
[02:59] <Error404> is there any way to choose which vertices you're drawing?
[03:00] <RedBeard> like i said before, d3d enforces a pretty optimized method of doing things
[03:00] <Error404> say I want to make one very large quad instead of two smaller ones?
[03:00] <RedBeard> one large quad?
[03:00] <Error404> Yeah assuming that those two quads are next to each other
[03:00] <Error404> and the same size.
[03:00] <Error404> I could get the far corners and then just make one extremely large quad instead?
[03:01] <RedBeard> the positions of the vertices are in the vertex buffer, if you adjust the vertex data then the polygons will reflect the change
[03:01] <RedBeard> yes
[03:01] <RedBeard> but you'd have to change either the index buffer or the vertex buffer
[03:01] <Error404> which is slow right?
[03:01] <RedBeard> can be
[03:01] <RedBeard> if you do it as a preprocessing step then it's no big deal
[03:01] <RedBeard> just try to steer clear of doing it every frame
[03:02] <Error404> reminds me of LOD
[03:02] <RedBeard> in the case of individual quads, you won't see a big hit
[03:02] <RedBeard> but if you're talking about dynamically building index buffers for LOD, that can get slow
[03:02] <Error404> So it's good to try to draw as many things with one DP call as you can right?
[03:02] <RedBeard> better to adjust the indices than the vertices though, since the indices are 2 bytes each and the vertices are like 32 bytes each
[03:02] <RedBeard> BIG difference
[03:03] <RedBeard> yes, put as much stuff into one DP as you can
[03:03] <Error404> wouldn't that mean it's all one big arse triangle list?
[03:03] <RedBeard> yeah, there's a limit to how much you can do in one call
[03:04] <Error404> Do they all have to be connected, or is that just with tristrips and stuff?
[03:04] <RedBeard> all the triangles drawn must be consecutive in the index list
[03:04] <Error404> Ohh
[03:04] <RedBeard> they can be totally separate
[03:04] <Error404> yeah that's what I was talking about
[03:04] <RedBeard> each triangle in a list is drawn independently
[03:04] <Error404> Like drawing from index 1, then index 4,
[03:04] <Error404> Now i see why you can't
[03:04] <Error404> It has to be consecutive
[03:04] <RedBeard> there's also a trick when using strips to do the same thing
[03:04] <Error404> hm?
[03:05] <RedBeard> "degenerate triangles" are triangles that are actually lines or points
[03:05] <RedBeard> they have no area, and thus can't be drawn
[03:05] <Error404> Ooo
[03:05] <Error404> so you draw a line to the other triangles
[03:05] <RedBeard> now, to make a degenerate tri, you just give it the same vertex more than once
[03:06] <Error404> that's a neat trick
[03:06] <Error404> it's fast right?
[03:06] <RedBeard> you have to do 2 degenerate tri's for each "jump" though
[03:06] <Error404> 2?
[03:06] <RedBeard> the video hardware is relatively quick to reject them
[03:07] <RedBeard> yes, 2
[03:07] <Error404> what's the reason for needing two?
[03:08] <Washu[AWAY]> two get it to have the proper order for rendering the next triangle in the sequence
[03:08] <RedBeard> when dealing with a tristrip, the "current triangle" is composed of the vertices at the current, current+1, and current+2 positions
[03:08] <Washu[AWAY]> shouldn't that be current, current-1 and current-2
[03:08] <RedBeard> and because of culling, it flip-flops back and forth between CW and CCW culling on each triangle
[03:09] <Error404> ah that's right
[03:09] <Error404> I remember reading that
[03:09] <RedBeard> so to jump and keep the right culling mode, you need to jump twice
[03:09] <RedBeard> not if you start at 0
[03:09] <RedBeard> ^ washu
[03:09] * Error404 nods
[03:09] <Washu[AWAY]> that's right...sorry, its late
[03:09] <Washu[AWAY]> why would you start at 0 though when you can start at 2
[03:10] <Error404> So hmm, if I wanted to make a cube i'd have 8 vertices??
[03:10] <Washu[AWAY]> alright, I'm out
[03:10] <Error404> I'd also have to do a matrix thingy right?
[03:10] <Washu[AWAY]> Latz red, Err, other people that I can't see
[03:10] * Quits: Washu[AWAY] ( (I think we need to learn to recieve the entire buffer before closing the socket...!)
[03:10] <RedBeard> yes 8 vertices
[03:10] <RedBeard> how many indices (using a list)
[03:11] <Error404> lemme think
[03:11] <RedBeard> btw, the benefit of using tristrips is largely done away with when using indexing
[03:11] <RedBeard> which is why im getting you to use indexing with lists - more flexible, and still fast
[03:11] <Error404> indices are just triangles right? because d3d can't handle primitive quads?
[03:11] <RedBeard> yes, for almost everything, a "primitive" is a triangle
[03:11] <Error404> ogl can take quads right?
[03:11] <RedBeard> it's a little different when using n-patches, but i havent used those
[03:12] <RedBeard> uh it can, but it breaks them into triangles internally
[03:12] <Error404> so one could easily design directx to do it as well.
[03:12] <Error404> soif there are 6 sides.
[03:12] <Error404> there are 2 tris per side
[03:12] <Error404> 12 indices?
[03:13] <RedBeard> 12 faces
[03:13] <RedBeard> 12 tri's
[03:13] <RedBeard> 36 indices
[03:13] * Error404 blikns
[03:13] <RedBeard> 6 sides, 12 tri's
[03:13] <Error404> so index != tri
[03:13] <RedBeard> 36 indices
[03:13] <RedBeard> in a list, tri = 3*index
[03:13] <Error404> Oh yeah that's right!
[03:14] <Error404> Becuase you draw seperate tri's for each
[03:14] <RedBeard> right
[03:14] <RedBeard> lemme quickly explain why indexing is so good
[03:14] <RedBeard> before i head to bed
[03:14] <Error404> ok
[03:14] <RedBeard> the video card has a "vertex cache" which is different from an L2 cache
[03:14] <RedBeard> it also haas an L2 kinda thing
[03:14] * Error404 reminds to look up L2
[03:15] <RedBeard> the difference is that L2 is just memory
[03:15] * Error404 writes that down
[03:15] <RedBeard> and vertex cache is post-transform vertices
[03:15] * Quits: kOOktrOOp ( (Read error to kOOktrOOp[]: Connection reset by peer)
[03:15] <RedBeard> so if you're applying a complex vertex shader, and you go back and use an index again in a 2nd triangle, it can grab the already-transformed vertex from vertex cache without running it through the shader again
[03:16] <Error404> oo
[03:16] <RedBeard> ie, (0,1,2, 2,1,3) uses 1 and 2 twice
[03:16] <RedBeard> so you only compute 4 vertex positions and use them multiple times
[03:16] <Error404> i'm assuming that is much faster then
[03:16] <Error404> becuase you aren't reading from the vb
[03:16] <RedBeard> it can be up to three times faster
[03:16] <Error404> yeah that is fast.
[03:17] <RedBeard> "did you just say 'three times faster'?"
[03:17] <RedBeard> Yes.
[03:17] <Error404> haha
[03:17] <Error404> that's awesome.
[03:17] <RedBeard> remember that the vertex cache is quite small though, around 10 vertices on a gf2
[03:17] <Error404> I wonder how many are on gf3
[03:17] <RedBeard> i think 14 or 18
[03:17] <Error404> why so small?
[03:18] <RedBeard> takes a lot of transistors, and it's integrated into the core of the chip
[03:18] <RedBeard> = expensive
[03:18] <Error404> oh that makes sense then
[03:18] <RedBeard> and there are things in d3dx to optimize a mesh for a certain chip, based on the vertex cache size, so that it uses a vertex as many times as possible before it drops out of the cache
[03:18] <Error404> so is it smart enough to do all the computations on the available vertices before shoving them away and getting other thins?
[03:18] <Error404> ah just what you said
[03:18] <Error404> lol
[03:19] <RedBeard> that kind of mesh optimization is a freakin complex process
[03:19] <RedBeard> because once it drops out of cache, you want to delay using that vertex again until you can get a batch of uses together again
[03:19] <RedBeard> while keeping track of all the other vertices at the same time
[03:20] <Error404> sounds like it would be impossible to use with tristrips
[03:20] <RedBeard> i havent seen a paper that discusses how to do that yet, i might look for one tomorrow
[03:20] <Error404> as you're always borrowing at least one vertex from somewhere else.
[03:20] <RedBeard> not really, but tristrips make it harder to jump around to use vertex cache well
[03:21] <Error404> yeah that makes sense.
[03:21] <RedBeard> you don't have to use the vertex that you used on the previous triangle
[03:21] <RedBeard> you just have to use it while it's still in cache
[03:21] <RedBeard> so you are free to jump around a little
[03:21] <Error404> ah.
[03:21] <RedBeard> tristrips dont let you jump around, they force you to keep the last 2 vertices
[03:22] <RedBeard> so that's another reason to use lists
[03:22] <Error404> there isn't much of a diff though right?
[03:22] <Error404> Other than the extra drawing and more storage space
[03:22] <RedBeard> not a huge difference, but you might be surprised how much performance is gained by avoiding cache misses
[03:23] <RedBeard> basically the only difference between indexed tri lists and strips is the amount of memory used
[03:23] <RedBeard> and when you have textures taking up 4MB of memory, and index buffers that take up 50K, that's nothing to worry about
[03:23] <Error404> heh yeah
[03:24] <Error404> and with more and more cards going over 100mb of VRAM and more people getting over 1gb of system ram it's going to be less of a prob
[03:24] <Error404> i guess
[03:24] <RedBeard> yep
[03:24] <Error404> Thank you man, this really helped me a lot.
[03:24] <RedBeard> glad to hear it
[03:25] <Error404> I've been trying to get someone to help me, but no one seemed to want to.
[03:25] <Error404> even the great mittens..
[03:25] <RedBeard> hope i've given enough info for you to start exploring some more stuff
[03:25] * Error404 kicks trent.
[03:25] <RedBeard> .k mittens[smeep]
[03:25] * mittens[smeep] was kicked by X2 (<RedBeard> You're outta-here, mittens[smeep]!)
[03:25] * Joins: mittens[smeep] (
[03:25] * X2 sets mode: +o mittens[smeep]
[03:25] <RedBeard> heheh
[03:25] <Error404> Yeah I'm going to try to make a cube now, and read up on the other functions i called in the class.
[03:25] <RedBeard> good
[03:25] <Error404> I know more about DIP and SetIndices than any other.
[03:26] <Error404> I like knowing all the possibilities.
[03:26] <RedBeard> yep
[03:26] <RedBeard> im gonna get some sleep

-- the header file, DXanadu.h --

_***font face=Courier New***_
#include <d3d8.h>

#ifndef _DXANADU_
#define _DXANADU_

class cDXanadu

bool InitD3D(HWND hWnd);
bool InitFrame();

bool DoFrame();

void CleanUp();

float x,y,z,rhw;
DWORD color;


bool InitBuffers();
IDirect3D8 *pD3D;
IDirect3DDevice8 *pd3dDevice;
IDirect3DVertexBuffer8 *vb;
IDirect3DIndexBuffer8 *ib;
DWORD num_tris;
DWORD num_verts;



-- the source file, DXanadu.cpp --

_***font face=Courier New***_
#include "DXanadu.h"

pD3D = NULL;
pd3dDevice = NULL;


bool cDXanadu::InitD3D(HWND hWnd)
if(NULL == (pD3D = Direct3DCreate8 (D3D_SDK_VERSION)))
return false;

if(FAILED (pD3D->GetAdapterDisplayMode(D3DADAPTER_DEFAULT, &d3ddm)))
return false;

d3dpp.Windowed = true;
d3dpp.BackBufferFormat = d3ddm.Format;

return false;
pd3dDevice->SetRenderState( D3DRS_CULLMODE, D3DCULL_CCW );
return true;

bool cDXanadu::InitFrame()
return false;
return true;

//make VB, make IB, lock VB, lock IB, fill in vertex data, fill in index data, unlock both
bool cDXanadu::InitBuffers()
num_tris = 2;
num_verts = 4;
CUSTOMVERTEX g_Vertices[] =
{ 150.0f,     50.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(255,255,255,255) },
{ 250.0f,     50.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(0,0,0,128) },
{ 150.0f,    150.0f,    0.5f,    1.0f,    D3DCOLOR_RGBA(255,0,0,64) },
{ 250.0f,    150.0f, 0.5f,    1.0f,    D3DCOLOR_RGBA(0,255,0,255) }
if(FAILED( pd3dDevice->CreateVertexBuffer(num_verts*sizeof(CUSTOMVERTEX), D3DUSAGE_WRITEONLY, D3DFVF_CUSTOMVERTEX,D3DPOOL_DEFAULT, &vb)))
return false;
VOID * pVertices;
if(FAILED( vb->Lock(0, 0, (BYTE**)&pVertices, 0)))
return false;

// Creating the Index Buffer
if(FAILED( pd3dDevice->CreateIndexBuffer(sizeof(short)*3*num_tris, D3DUSAGE_WRITEONLY, D3DFMT_INDEX16, D3DPOOL_DEFAULT, &ib)))
return false;
unsigned short * pIndices;
if(FAILED( ib->Lock(0, 0, (BYTE**)&pIndices, 0)))
return false;
pIndices[0] = 0;    pIndices[1] = 1;    pIndices[2] = 2;
pIndices[3] = 2;    pIndices[4] = 1;    pIndices[5] = 3;

return true;

bool cDXanadu::DoFrame()
pd3dDevice->Clear(0, NULL, D3DCLEAR_TARGET, D3DCOLOR_XRGB(0,128,255), 1.0f, 0);
// Begin the scene

// Render Stuff
pd3dDevice->SetIndices(ib, 0);
pd3dDevice->SetStreamSource(0, vb, sizeof(CUSTOMVERTEX));

pd3dDevice->DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, num_verts, 0, num_tris);

// End the scene

// Show the backbuffer
return true;

void cDXanadu::CleanUp()
if(vb != NULL)
if(ib != NULL)
if(pd3dDevice != NULL)
if(pD3D != NULL)


Text file last modified on April 28, 2007, 10:28:21 PM, MST.

All content Copyright 2017 Andy Campbell.
Permission to reproduce content, in any format,
by explicit written permission of author only.
In just two days, tomorrow will be yesterday.