Jump to content
• Advertisement

# DX11 Getting around non-connected vertex gaps in hardware tessellation displacement mapping

This topic is 1264 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

## Recommended Posts

Sorry for the long title, couldn't figure out how to express it shorter without being overly ambigious as to what this post is about.

Anyway, I've been poking around with displacement maping using the hardware tessellation features of DX11 for getting some more vertices to actually displace the last few days, for no particular reason other than to try it out so I'm not really looking for other ways to solve some specific problem.

Displacing a sphere or some other surface with completely connected faces work out as intended but issues obviously occur where there are multiple vertices with the same position but different normals (these vertices then get displaced in different directions and thus become disconnected => gaps appear in the geometry). I tried to mock up some simple solution to this by finding out which vertices share positions in my meshes and then setting a flag for these to tell my domain shader to not displace those vertices at all; it wouldn't be overly pretty but at least the mesh should be gapless and it hopefully wouldn't be too noticeable I reasoned. Of course this didn't work out very well (the whole subdivision patches generated from such overlapping vertices had their displacement factors set to 0 creating quite obvious, large frames around right angles and such). What I'm wondering is basically if this is a reasonable approach to try to refine further or if there are other ways to go about it that may be better? The only article on the topic I've managed to find mostly went on about the exquisitness of Bezier curves but didn't really seem to come to any conclusions (although maybe those would've been obvious to anyone having the required math skills).

Thankful for any pointers on this, the more I try to force this, the more it feels like I'm probably missing something.

As for my implementation of the tessellation, I've mostly based it around what is described in chapter 18.7 and 18.8 of Introduction to 3D Game Programming With DirectX 11 (http://www.amazon.com/Introduction-3D-Game-Programming-DirectX/dp/1936420228).

#### Share this post

##### Share on other sites
Advertisement

Thanks MJP.

I don't suppose there's any video / audio recording of the presentation using those slides available somewhere, possibly for a nominal fee?

On a more off-topic note, I recognize that avatar of yours but have been unable to remember the name of the show (or was it possibly a book?) it featured in, care to enlighten me?

#### Share this post

##### Share on other sites

I think i have implemented index buffer for PN-AEN as described in that pdf document, but once i finished i saw that patch funcion HS_Constant is missing from shader that they have provided in appendix.

Copy-paste error.

D3D11 ERROR: ID3D11DeviceContext::DrawIndexed: Mismatched topology. Current Hull Shader expects input Control Point count of 3, but Input Assembler topology defines a patch list with 9 Control Points per patch. [ EXECUTION ERROR #2097222: DEVICE_DRAW_HULL_SHADER_INPUT_TOPOLOGY_MISMATCH]

Damn. Why did they provide code that does not work.

edit 6841:

I get shader to work this time but i am not sure if i have indices right as i see no displacement.

if i add displacement along normal:

... // domain shader
float3 n = mul(f3Normal, (float3x3)g_f4x4WorldView);
n = normalize(n);
f3EyePosition += n * 0.02f;


CRACKS!

void Test::calcPNAENIndices(const std::vector<USHORT>& ind, const std::vector<VERTEX>& verts, std::vector<USHORT>& out)
{
out.resize(ind.size() * 3);
for (std::size_t i = 0; i < ind.size(); i += 3)
{
out[3 * i + 0] = ind[i + 0];
out[3 * i + 1] = ind[i + 1];
out[3 * i + 2] = ind[i + 2];

out[3 * i + 3] = ind[i + 0];
out[3 * i + 4] = ind[i + 1];
out[3 * i + 5] = ind[i + 1];

out[3 * i + 6] = ind[i + 2];
out[3 * i + 7] = ind[i + 2];
out[3 * i + 8] = ind[i + 0];
}

struct Edge
{
float3 p[2];
USHORT inx[2];

bool operator == (const Edge& o) const
{
if (inx[0] == o.inx[0] && inx[1] == o.inx[1])
return true;

if (Equal(p[0].x, o.p[0].x) && Equal(p[0].y, o.p[0].y) && Equal(p[0].z, o.p[0].z))
{
if (Equal(p[1].x, o.p[1].x) && Equal(p[1].y, o.p[1].y) && Equal(p[1].z, o.p[1].z))
{
return true;
}
}

return false;
}
};

// reverse edges
std::vector<Edge> edges;
edges.resize(ind.size());
for (std::size_t i = 0; i < ind.size(); i += 3)
{
edges[i + 0].p[1]   = verts[ind[i + 0]].pos;
edges[i + 0].p[0]   = verts[ind[i + 1]].pos;
edges[i + 0].inx[1] = ind[i + 0];
edges[i + 0].inx[0] = ind[i + 1];

edges[i + 1].p[1]   = verts[ind[i + 1]].pos;
edges[i + 1].p[0]   = verts[ind[i + 2]].pos;
edges[i + 1].inx[1] = ind[i + 1];
edges[i + 1].inx[0] = ind[i + 2];

edges[i + 2].p[1]   = verts[ind[i + 2]].pos;
edges[i + 2].p[0]   = verts[ind[i + 0]].pos;
edges[i + 2].inx[1] = ind[i + 2];
edges[i + 2].inx[0] = ind[i + 0];
}

// compare
for (std::size_t i = 0, j = 0; i < out.size(); i += 9, j += 3)
{
Edge e;

// edge 0
e.p[0]   = verts[out[i + 0]].pos;
e.p[1]   = verts[out[i + 1]].pos;
e.inx[0] = out[i + 0];
e.inx[1] = out[i + 1];

if (e == edges[j + 0])
{
out[i + 0] = edges[j + 0].inx[0];
out[i + 1] = edges[j + 0].inx[1];
}

// edge 1
e.p[0]   = verts[out[i + 1]].pos;
e.p[1]   = verts[out[i + 2]].pos;
e.inx[0] = out[i + 1];
e.inx[1] = out[i + 2];

if (e == edges[j + 0])
{
out[i + 1] = edges[j + 0].inx[0];
out[i + 2] = edges[j + 0].inx[1];
}

// edge 2
e.p[0]   = verts[out[i + 2]].pos;
e.p[1]   = verts[out[i + 3]].pos;
e.inx[0] = out[i + 2];
e.inx[1] = out[i + 3];

if (e == edges[j + 0])
{
out[i + 2] = edges[j + 0].inx[0];
out[i + 3] = edges[j + 0].inx[1];
}

// edge 3
e.p[0] = verts[out[i + 3]].pos;
e.p[1] = verts[out[i + 4]].pos;
e.inx[0] = out[i + 3];
e.inx[1] = out[i + 4];

if (e == edges[j + 1])
{
out[i + 3] = edges[j + 1].inx[0];
out[i + 4] = edges[j + 1].inx[1];
}

// edge 4
e.p[0] = verts[out[i + 4]].pos;
e.p[1] = verts[out[i + 5]].pos;
e.inx[0] = out[i + 4];
e.inx[1] = out[i + 5];

if (e == edges[j + 1])
{
out[i + 4] = edges[j + 1].inx[0];
out[i + 5] = edges[j + 1].inx[1];
}

// edge 5
e.p[0] = verts[out[i + 5]].pos;
e.p[1] = verts[out[i + 6]].pos;
e.inx[0] = out[i + 5];
e.inx[1] = out[i + 6];

if (e == edges[j + 1])
{
out[i + 5] = edges[j + 1].inx[0];
out[i + 6] = edges[j + 1].inx[1];
}

// edge 6
e.p[0] = verts[out[i + 6]].pos;
e.p[1] = verts[out[i + 7]].pos;
e.inx[0] = out[i + 6];
e.inx[1] = out[i + 7];

if (e == edges[j + 2])
{
out[i + 6] = edges[j + 2].inx[0];
out[i + 7] = edges[j + 2].inx[1];
}

// edge 7
e.p[0]   = verts[out[i + 7]].pos;
e.p[1]   = verts[out[i + 8]].pos;
e.inx[0] = out[i + 7];
e.inx[1] = out[i + 8];

if (e == edges[j + 2])
{
out[i + 7] = edges[j + 2].inx[0];
out[i + 8] = edges[j + 2].inx[1];
}

// edge 8
e.p[0] = verts[out[i + 8]].pos;
e.p[1] = verts[out[i + 0]].pos;
e.inx[0] = out[i + 8];
e.inx[1] = out[i + 0];

if (e == edges[j + 2])
{
out[i + 8] = edges[j + 2].inx[0];
out[i + 0] = edges[j + 2].inx[1];
}
}
}

Edited by belfegor

#### Share this post

##### Share on other sites

I don't suppose there's any video / audio recording of the presentation using those slides available somewhere, possibly for a nominal fee?

Not that I know of, sorry.

On a more off-topic note, I recognize that avatar of yours but have been unable to remember the name of the show (or was it possibly a book?) it featured in, care to enlighten me?

It's Rocko!

#### Share this post

##### Share on other sites

I think i had some mistakes in code, having hard time to understand what exactly i need to do.

Here is corrected version but still wrong results.

void Test::calcPNAENIndices(const std::vector<USHORT>& ind, const std::vector<VERTEX>& verts, std::vector<USHORT>& out)
{
struct Edge
{
float3 p[2];
USHORT inx[2];

bool operator == (const Edge& o) const
{
if (inx[0] == o.inx[0] && inx[1] == o.inx[1])
return true;

if (Equal(p[0].x, o.p[0].x) && Equal(p[0].y, o.p[0].y) && Equal(p[0].z, o.p[0].z))
{
if (Equal(p[1].x, o.p[1].x) && Equal(p[1].y, o.p[1].y) && Equal(p[1].z, o.p[1].z))
{
return true;
}
}

return false;
}
};

std::vector<Edge> edges(ind.size());
out.resize(ind.size() * 3);

for (std::size_t i = 0; i < ind.size(); i += 3)
{
// initial values
out[3 * i + 0] = ind[i + 0];
out[3 * i + 1] = ind[i + 1];
out[3 * i + 2] = ind[i + 2];

out[3 * i + 3] = ind[i + 0];
out[3 * i + 4] = ind[i + 1];
out[3 * i + 5] = ind[i + 1];

out[3 * i + 6] = ind[i + 2];
out[3 * i + 7] = ind[i + 2];
out[3 * i + 8] = ind[i + 0];

// store reversed
edges[i + 0].p[1]   = verts[ind[i + 0]].pos;
edges[i + 0].p[0]   = verts[ind[i + 1]].pos;
edges[i + 0].inx[1] = ind[i + 0];
edges[i + 0].inx[0] = ind[i + 1];

edges[i + 1].p[1]   = verts[ind[i + 1]].pos;
edges[i + 1].p[0]   = verts[ind[i + 2]].pos;
edges[i + 1].inx[1] = ind[i + 1];
edges[i + 1].inx[0] = ind[i + 2];

edges[i + 2].p[1]   = verts[ind[i + 2]].pos;
edges[i + 2].p[0]   = verts[ind[i + 0]].pos;
edges[i + 2].inx[1] = ind[i + 2];
edges[i + 2].inx[0] = ind[i + 0];
}

for (std::size_t i = 0; i < out.size(); i += 9)
{
// i think i should skip first 3 indices as they point to triangle and i need to check edges
for (std::size_t j = 3; j < 9; ++j)
{
std::size_t first  = j;
std::size_t second = j + 1;
if (second == 9)
second = 3;

Edge e;
e.p[0]   = verts[out[i + first]].pos;
e.p[1]   = verts[out[i + second]].pos;
e.inx[0] = out[i + first];
e.inx[1] = out[i + second];

for (std::size_t k = 0; k < edges.size(); ++k)
{
if (e == edges[k])
{
out[i + first]  = edges[k].inx[0];
out[i + second] = edges[k].inx[1];
}
}
}
}
}


Can someone take a look please and decipher instructions given for this to work:

1.  Create an output IB that is 3 times the size of input IB.

2.  For each input Triangle in IB, with indices i0, i1 and i2:
a.  Write out an initial output entry of:  i0, i1, i2, i0, i1, i1, i2, i2, i0, which sets edges to

initially be neighbors of themselves. This would produce identical results to PN
Triangles.
b.  Lookup the positions p0, p1, and p2, using i0, i1 and i2 to perform a lookup for
position of the associated vertex in VB.
c.  Define 3 Edges, which consist of the two indices and two positions that make up
the corresponding Edge. An Edge should consist of the origin index, the
destination index, the origin position and the destination position.
d.  For each edge, store the reverse of that edge in an easily searchable data structure
for the next step. The reference implementation uses an stdext::hash_map<Edge,
Edge> for this purpose. Reverse simply flips the sense of the edge (originating at
the destination position and index and heading to the origin position and index).

3.  Walk the output index buffer (OB) constructed in step 2. For each patch of 9 indices:
a.  For each Edge in the current Patch, perform a lookup into Edge->Edge mapping
created in step 2d.
b.  If found, replace the current indices with the indices found in the map. Note that
two edges should be considered matching if their "from" and "to" indices match,
OR if their "from" and "to" positions match.
c.  If not, continue to use the existing indices.

Upon completion of this algorithm, a buffer suitable for usage with PN-AEN will be
available.

Edited by belfegor

#### Share this post

##### Share on other sites

That is that. Thank you very much.

With "initial" IB (notice cracks):

PNAEN

Now i would like to displace position along normal using heightmap but i dont know how to get average normal so i can pass it to domain shader. Any pointers?

#### Share this post

##### Share on other sites
Now the real challenge starts. Here the sample goes the easy way and just passes the normals from the inner triangle and interpolates linearly. "Correct" PN triangles uses a quadratic bezier patch for the normals, examples of which you can find in the (June 2010 SDK) samples or in the Hieroglyph 3 engine (there's also full chapter in the Practical Rendering book).
There's also a displacement tesselation sample using decals in the SDK which might be worth a peek.

Either way it sounds complex as the presentation MJP linked to shows. Can't help you any further now, I haven't done displacement mapping (not counting terrain). For a simple start maybe you can get away averaging normals at corners (not using a bezier quad patch and taking averaged normals from all adjacent triangles). But that's really just an idea.

Anyway, looks like it also needs special care on the content creation side (citing above presentation).

Speaking of which (and out of curiosity): Why does that pig/boar generate cracks ? I wouldn't expect them on organic surfaces (the initial index buffer results in usual PN triangles). You got a link to that mesh ?

#### Share this post

##### Share on other sites

• Advertisement
• Advertisement

• ### Popular Now

• 10
• 18
• 14
• 18
• 15
• Advertisement
• ### Similar Content

• By chiffre
Introduction:
In general my questions pertain to the differences between floating- and fixed-point data. Additionally I would like to understand when it can be advantageous to prefer fixed-point representation over floating-point representation in the context of vertex data and how the hardware deals with the different data-types. I believe I should be able to reduce the amount of data (bytes) necessary per vertex by choosing the most opportune representations for my vertex attributes. Thanks ahead of time if you, the reader, are considering the effort of reading this and helping me.
I found an old topic that shows this is possible in principal, but I am not sure I understand what the pitfalls are when using fixed-point representation and whether there are any hardware-based performance advantages/disadvantages.
(TLDR at bottom)
The Actual Post:
To my understanding HLSL/D3D11 offers not just the traditional floating point model in half-,single-, and double-precision, but also the fixed-point model in form of signed/unsigned normalized integers in 8-,10-,16-,24-, and 32-bit variants. Both models offer a finite sequence of "grid-points". The obvious difference between the two models is that the fixed-point model offers a constant spacing between values in the normalized range of [0,1] or [-1,1], while the floating point model allows for smaller "deltas" as you get closer to 0, and larger "deltas" the further you are away from 0.
To add some context, let me define a struct as an example:
struct VertexData { float[3] position; //3x32-bits float[2] texCoord; //2x32-bits float[3] normals; //3x32-bits } //Total of 32 bytes Every vertex gets a position, a coordinate on my texture, and a normal to do some light calculations. In this case we have 8x32=256bits per vertex. Since the texture coordinates lie in the interval [0,1] and the normal vector components are in the interval [-1,1] it would seem useful to use normalized representation as suggested in the topic linked at the top of the post. The texture coordinates might as well be represented in a fixed-point model, because it seems most useful to be able to sample the texture in a uniform manner, as the pixels don't get any "denser" as we get closer to 0. In other words the "delta" does not need to become any smaller as the texture coordinates approach (0,0). A similar argument can be made for the normal-vector, as a normal vector should be normalized anyway, and we want as many points as possible on the sphere around (0,0,0) with a radius of 1, and we don't care about precision around the origin. Even if we have large textures such as 4k by 4k (or the maximum allowed by D3D11, 16k by 16k) we only need as many grid-points on one axis, as there are pixels on one axis. An unsigned normalized 14 bit integer would be ideal, but because it is both unsupported and impractical, we will stick to an unsigned normalized 16 bit integer. The same type should take care of the normal vector coordinates, and might even be a bit overkill.
struct VertexData { float[3] position; //3x32-bits uint16_t[2] texCoord; //2x16bits uint16_t[3] normals; //3x16bits } //Total of 22 bytes Seems like a good start, and we might even be able to take it further, but before we pursue that path, here is my first question: can the GPU even work with the data in this format, or is all I have accomplished minimizing CPU-side RAM usage? Does the GPU have to convert the texture coordinates back to a floating-point model when I hand them over to the sampler in my pixel shader? I have looked up the data types for HLSL and I am not sure I even comprehend how to declare the vertex input type in HLSL. Would the following work?
struct VertexInputType { float3 pos; //this one is obvious unorm half2 tex; //half corresponds to a 16-bit float, so I assume this is wrong, but this the only 16-bit type I found on the linked MSDN site snorm half3 normal; //same as above } I assume this is possible somehow, as I have found input element formats such as: DXGI_FORMAT_R16G16B16A16_SNORM and DXGI_FORMAT_R16G16B16A16_UNORM (also available with a different number of components, as well as different component lengths). I might have to avoid 3-component vectors because there is no 3-component 16-bit input element format, but that is the least of my worries. The next question would be: what happens with my normals if I try to do lighting calculations with them in such a normalized-fixed-point format? Is there no issue as long as I take care not to mix floating- and fixed-point data? Or would that work as well? In general this gives rise to the question: how does the GPU handle fixed-point arithmetic? Is it the same as integer-arithmetic, and/or is it faster/slower than floating-point arithmetic?
Assuming that we still have a valid and useful VertexData format, how far could I take this while remaining on the sensible side of what could be called optimization? Theoretically I could use the an input element format such as DXGI_FORMAT_R10G10B10A2_UNORM to pack my normal coordinates into a 10-bit fixed-point format, and my verticies (in object space) might even be representable in a 16-bit unsigned normalized fixed-point format. That way I could end up with something like the following struct:
struct VertexData { uint16_t[3] pos; //3x16bits uint16_t[2] texCoord; //2x16bits uint32_t packedNormals; //10+10+10+2bits } //Total of 14 bytes Could I use a vertex structure like this without too much performance-loss on the GPU-side? If the GPU has to execute some sort of unpacking algorithm in the background I might as well let it be. In the end I have a functioning deferred renderer, but I would like to reduce the memory footprint of the huge amount of vertecies involved in rendering my landscape.
TLDR: I have a lot of vertices that I need to render and I want to reduce the RAM-usage without introducing crazy compression/decompression algorithms to the CPU or GPU. I am hoping to find a solution by involving fixed-point data-types, but I am not exactly sure how how that would work.
• By cozzie
Hi all,
I was wondering it it matters in which order you draw 2D and 3D items, looking at the BeginDraw/EndDraw calls on a D2D rendertarget.
The order in which you do the actual draw calls is clear, 3D first then 2D, means the 2D (DrawText in this case) is in front of the 3D scene.
The question is mainly about when to call the BeginDraw and EndDraw.
Note that I'm drawing D2D stuff through a DXGI surface linked to the 3D RT.
Option 1:
A - Begin frame, clear D3D RT
B - Draw 3D
C - BeginDraw D2D RT
D - Draw 2D
E - EndDraw D2D RT
F - Present
Option 2:
A - Begin frame, clear D3D RT + BeginDraw D2D RT
B - Draw 3D
C - Draw 2D
D - EndDraw D2D RT
E- Present
Would there be a difference (performance/issue?) in using option 2? (versus 1)
Any input is appreciated.

• Do you know any papers that cover custom data structures like lists or binary trees implemented in hlsl without CUDA that work perfectly fine no matter how many threads try to use them at any given time?
• By cozzie
Hi all,
Last week I noticed that when I run my test application(s) in Renderdoc, it crashes when it enable my code that uses D2D/DirectWrite. In Visual Studio no issues occur (debug or release), but when I run the same executable in Renderdoc, it crashes somehow (assert of D2D rendertarget or without any information). Before I spend hours on debugging/ figuring it out, does someone have experience with this symptom and/or know if Renderdoc has known issues with D2D? (if so, that would be bad news for debugging my application in the future );
I can also post some more information on what happens, code and which code commented out, eliminates the problems (when running in RenderDoc).
Any input is appreciated.

• Hi Guys,
I understand how to create input layouts etc... But I am wondering is it at all possible to derive an input layout from a shader and create the input layout directly from this? (Rather than manually specifying the input layout format?)
Thanks in advance :)

• Advertisement
×

## Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

Participate in the game development conversation and more when you create an account on GameDev.net!

Sign me up!