• Advertisement
Sign in to follow this  

DX11 DX11 - Sponza Scene - Comparison of loading times

This topic is 1586 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

Hi guys! wink.png

 

Well I decided to look into the topic of global illumination, and as seen, many of those demos use the Sponza Scene (From Crytek), because of its complexity. And, well, I tried it as well, but my loading time seems quite high just for the Sponza Scene. (Loading the file, creating buffers, and creating a simple texture material).

 

Loading time for my project: 21.9 Seconds (21 for Loading the actual file, ~1 for Creating the buffers) wacko.png

 

For those people who have tried loading this scene before, what are your loading times, are mine terrible?

 

Thanks, as usual.

-MIGI0027

Share this post


Link to post
Share on other sites
Advertisement

What format are you trying to load it in?  Sounds like you might be parsing text, which is never going to be fast; instead you should preprocess the model into a binary format that can be just loaded directly into buffers with a single fread.

Share this post


Link to post
Share on other sites
My framework loads the crytek sponza in about 2-3 seconds, though preprocessing ( geometry to binary data, textures to dds, compiling hlsl... ) takes about 15-20 seconds.

Share this post


Link to post
Share on other sites

Wow, too much. Give me link to the sponza scene and I'll test it. I can load now a few MiB of meshes and about 300 MiB of textures in DDS format in 3-4 seconds.

Share this post


Link to post
Share on other sites

PS. When I mean the load time, I mean:

  • Loading the file (Assimp)
  • Parsing it (Assimp)
  • Triangulating (Assimp)
  • Copying Vertices / Normals / Tangent...
  • Calculating bounding box
  • Create shader/material
  • Create ALL textures.
  • Create buffers (v/i)

Share this post


Link to post
Share on other sites

Are you testing this in a debug or a release build?  File operations can be significantly slower in debug builds than release builds, so that might account for some of the time.  If that isn't the issue, you might want to profile each of the operations you listed above to see which one is taking the most time.  It doesn't need to be anything fancy, just dump the system time in milliseconds to the output debug area when each process finishes.  That should help you figure out what is going on.

Share this post


Link to post
Share on other sites

What format are you trying to load it in?  Sounds like you might be parsing text, which is never going to be fast; instead you should preprocess the model into a binary format that can be just loaded directly into buffers with a single fread.

Well, the thing is that I use Assimp. dry.png

You can still take your code that uses Assimp and move it into an external tool that you run ahead of time, which you can then use generate a simpler binary file that can be loaded quickly.

Share this post


Link to post
Share on other sites

 

 

What format are you trying to load it in?  Sounds like you might be parsing text, which is never going to be fast; instead you should preprocess the model into a binary format that can be just loaded directly into buffers with a single fread.

Well, the thing is that I use Assimp. dry.png

 

You can still take your code that uses Assimp and move it into an external tool that you run ahead of time, which you can then use generate a simpler binary file that can be loaded quickly.

 

 

Huh.. Don't know why I didn't think of that, but thanks, I'll give it a try.

Share this post


Link to post
Share on other sites

Sorry, I've been implementing terrain LOD and did not have time to profile my loading of the mesh. But when I do have time, this 22 MiB mesh will surely be interesting to load and render. I saw the sponza mesh everywhere. Pretty much anybody writing an engine has screenshots with it, but I never used it before. It will be a good stress test for my forward rendering of multiple point lights.

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
  • Advertisement
  • Popular Tags

  • Advertisement
  • Popular Now

  • Similar Content

    • By AxeGuywithanAxe
      I wanted to see how others are currently handling descriptor heap updates and management.
      I've read a few articles and there tends to be three major strategies :
      1 ) You split up descriptor heaps per shader stage ( i.e one for vertex shader , pixel , hull, etc)
      2) You have one descriptor heap for an entire pipeline
      3) You split up descriptor heaps for update each update frequency (i.e EResourceSet_PerInstance , EResourceSet_PerPass , EResourceSet_PerMaterial, etc)
      The benefits of the first two approaches is that it makes it easier to port current code, and descriptor / resource descriptor management and updating tends to be easier to manage, but it seems to be not as efficient.
      The benefits of the third approach seems to be that it's the most efficient because you only manage and update objects when they change.
    • By evelyn4you
      hi,
      until now i use typical vertexshader approach for skinning with a Constantbuffer containing the transform matrix for the bones and an the vertexbuffer containing bone index and bone weight.
      Now i have implemented realtime environment  probe cubemaping so i have to render my scene from many point of views and the time for skinning takes too long because it is recalculated for every side of the cubemap.
      For Info i am working on Win7 an therefore use one Shadermodel 5.0 not 5.x that have more options, or is there a way to use 5.x in Win 7
      My Graphic Card is Directx 12 compatible NVidia GTX 960
      the member turanszkij has posted a good for me understandable compute shader. ( for Info: in his engine he uses an optimized version of it )
      https://turanszkij.wordpress.com/2017/09/09/skinning-in-compute-shader/
      Now my questions
       is it possible to feed the compute shader with my orignial vertexbuffer or do i have to copy it in several ByteAdressBuffers as implemented in the following code ?
        the same question is about the constant buffer of the matrixes
       my more urgent question is how do i feed my normal pipeline with the result of the compute Shader which are 2 RWByteAddressBuffers that contain position an normal
      for example i could use 2 vertexbuffer bindings
      1 containing only the uv coordinates
      2.containing position and normal
      How do i copy from the RWByteAddressBuffers to the vertexbuffer ?
       
      (Code from turanszkij )
      Here is my shader implementation for skinning a mesh in a compute shader:
      1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 struct Bone { float4x4 pose; }; StructuredBuffer<Bone> boneBuffer;   ByteAddressBuffer vertexBuffer_POS; // T-Pose pos ByteAddressBuffer vertexBuffer_NOR; // T-Pose normal ByteAddressBuffer vertexBuffer_WEI; // bone weights ByteAddressBuffer vertexBuffer_BON; // bone indices   RWByteAddressBuffer streamoutBuffer_POS; // skinned pos RWByteAddressBuffer streamoutBuffer_NOR; // skinned normal RWByteAddressBuffer streamoutBuffer_PRE; // previous frame skinned pos   inline void Skinning(inout float4 pos, inout float4 nor, in float4 inBon, in float4 inWei) {  float4 p = 0, pp = 0;  float3 n = 0;  float4x4 m;  float3x3 m3;  float weisum = 0;   // force loop to reduce register pressure  // though this way we can not interleave TEX - ALU operations  [loop]  for (uint i = 0; ((i &lt; 4) &amp;&amp; (weisum&lt;1.0f)); ++i)  {  m = boneBuffer[(uint)inBon].pose;  m3 = (float3x3)m;   p += mul(float4(pos.xyz, 1), m)*inWei;  n += mul(nor.xyz, m3)*inWei;   weisum += inWei;  }   bool w = any(inWei);  pos.xyz = w ? p.xyz : pos.xyz;  nor.xyz = w ? n : nor.xyz; }   [numthreads(1024, 1, 1)] void main( uint3 DTid : SV_DispatchThreadID ) {  const uint fetchAddress = DTid.x * 16; // stride is 16 bytes for each vertex buffer now...   uint4 pos_u = vertexBuffer_POS.Load4(fetchAddress);  uint4 nor_u = vertexBuffer_NOR.Load4(fetchAddress);  uint4 wei_u = vertexBuffer_WEI.Load4(fetchAddress);  uint4 bon_u = vertexBuffer_BON.Load4(fetchAddress);   float4 pos = asfloat(pos_u);  float4 nor = asfloat(nor_u);  float4 wei = asfloat(wei_u);  float4 bon = asfloat(bon_u);   Skinning(pos, nor, bon, wei);   pos_u = asuint(pos);  nor_u = asuint(nor);   // copy prev frame current pos to current frame prev pos streamoutBuffer_PRE.Store4(fetchAddress, streamoutBuffer_POS.Load4(fetchAddress)); // write out skinned props:  streamoutBuffer_POS.Store4(fetchAddress, pos_u);  streamoutBuffer_NOR.Store4(fetchAddress, nor_u); }  
    • By mister345
      Hi, can someone please explain why this is giving an assertion EyePosition!=0 exception?
       
      _lightBufferVS->viewMatrix = DirectX::XMMatrixLookAtLH(XMLoadFloat3(&_lightBufferVS->position), XMLoadFloat3(&_lookAt), XMLoadFloat3(&up));
      It looks like DirectX doesnt want the 2nd parameter to be a zero vector in the assertion, but I passed in a zero vector with this exact same code in another program and it ran just fine. (Here is the version of the code that worked - note XMLoadFloat3(&m_lookAt) parameter value is (0,0,0) at runtime - I debugged it - but it throws no exceptions.
          m_viewMatrix = DirectX::XMMatrixLookAtLH(XMLoadFloat3(&m_position), XMLoadFloat3(&m_lookAt), XMLoadFloat3(&up)); Here is the repo for the broken code (See LightClass) https://github.com/mister51213/DirectX11Engine/blob/master/DirectX11Engine/LightClass.cpp
      and here is the repo with the alternative version of the code that is working with a value of (0,0,0) for the second parameter.
      https://github.com/mister51213/DX11Port_SoftShadows/blob/master/Engine/lightclass.cpp
    • By mister345
      Hi, can somebody please tell me in clear simple steps how to debug and step through an hlsl shader file?
      I already did Debug > Start Graphics Debugging > then captured some frames from Visual Studio and
      double clicked on the frame to open it, but no idea where to go from there.
       
      I've been searching for hours and there's no information on this, not even on the Microsoft Website!
      They say "open the  Graphics Pixel History window" but there is no such window!
      Then they say, in the "Pipeline Stages choose Start Debugging"  but the Start Debugging option is nowhere to be found in the whole interface.
      Also, how do I even open the hlsl file that I want to set a break point in from inside the Graphics Debugger?
       
      All I want to do is set a break point in a specific hlsl file, step thru it, and see the data, but this is so unbelievably complicated
      and Microsoft's instructions are horrible! Somebody please, please help.
       
       
       

    • By mister345
      I finally ported Rastertek's tutorial # 42 on soft shadows and blur shading. This tutorial has a ton of really useful effects and there's no working version anywhere online.
      Unfortunately it just draws a black screen. Not sure what's causing it. I'm guessing the camera or ortho matrix transforms are wrong, light directions, or maybe texture resources not being properly initialized.  I didnt change any of the variables though, only upgraded all types and functions DirectX3DVector3 to XMFLOAT3, and used DirectXTK for texture loading. If anyone is willing to take a look at what might be causing the black screen, maybe something pops out to you, let me know, thanks.
      https://github.com/mister51213/DX11Port_SoftShadows
       
      Also, for reference, here's tutorial #40 which has normal shadows but no blur, which I also ported, and it works perfectly.
      https://github.com/mister51213/DX11Port_ShadowMapping
       
  • Advertisement