Jump to content

New Texture Compression Formats

Josh

438 views

The Vulkan renderer now supports new texture compression formats that can be loaded from DDS files. I've updated the DDS loader to support newer versions of the format with new features.

BC5 is a format ATI invented (originally called ATI2 or 3Dc) which is a two-channel compressed format specifically designed for storing normal maps. This gives you better quality normals than what DXT compression (even with the DXT5n swizzle hack) can provide.

BC7 is interesting because it uses the same size as DXT5 images but provides much higher quality results. The compression algorithm is also very long, sometimes taking ten minutes to compress a single texture!  Intel claims to have a fast-ish compressor for it but I have not tried it yet. Protip: You can open DDS files in newer versions of Visual Studio and select the compression format there.

Here is a grayscale gradient showing uncompressed, DXT5, BC7 UNORM, and BC7 SNORM formats. You can see BC7 UNORM and SNORM have much less artifacts than DXT, but is not quite the same as the original image.

uncompressed.thumb.png.9759cc5f5afc7c94947bad97a914cdee.png

dxt1.thumb.png.af7bf97f7c1cf6eaaf5b239cfa3577db.png

dxt5.thumb.png.0a450dcd60ac364d839e4c5ee0ed1d83.png

bc7_unorm.thumb.png.d3c5339ff07e4c68477070c854019c12.png

bc7_snorm.thumb.png.c9f5006d09352e3eba2d890bb1911100.png

The original image is 256 x 256, giving the following file sizes:

  • Uncompressed: 341 KB
  • DXT1: 42.8 KB (12.6% compression)
  • DXT5, BC7: 85.5 KB (25% compression)

I was curious what would happen if I zipped up some of the files, although this is only a minor concern. I guess that BC7 would not work with ZIP compression as well, since it is a more complicated algorithm.

  • DXT5: 16.6 KB
  • BC7 UNORM: 34 KB
  • BC7 SNORM: 42.4 KB

Based on the results above, I would probably still use uncompressed images for skyboxes and gradients, but anything else can benefit from this format. DXT compression looks like a blocky green mess by comparison.

I was curious to see how much of a difference the BC5 format made for normal maps so I made some similar renders for normals. You can see below that the benefits for normal maps are even more extreme. The BC5 compressed image is indistinguishable from the original while the DXT5n image has clear artifacts.

norm_uncompressed.thumb.png.641ec4ad732fe481db4cd58e4ae35499.png

norm_dxt5n.thumb.png.251b778e5633a48c0a75aa70da98c7f8.png

norm_bc5_unorm.thumb.png.e6d267d8e39dc97a09c7b806123f2ec9.png

In conclusion, these new formats in the Vulkan renderer, when used properly, will provide compression without visible artifacts.

  • Like 4


3 Comments


Recommended Comments

I replaced our own texture format constants like TEXTURE_RGBA with the Vulkan texture constants and added support for pretty much all the DDS formats that correspond to Vulkan formats.

  • Like 1

Share this comment


Link to comment

Also got Intel’s compressor working, it will make a good converter.

  • Like 2

Share this comment


Link to comment

Here's a nice shot showing the quality BC5 brings to normal maps:

Image1.png

  • Like 1

Share this comment


Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Add a comment...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Blog Entries

    • By Josh in Josh's Dev Blog 10
      Current generation graphics hardware only supports up to a 32-bit floating point depth buffer, and that isn't adequate for large-scale rendering because there isn't enough precision to make objects appear in the correct order and prevent z-fighting.

      After trying out a few different approaches I found that the best way to support large-scale rendering is to allow the user to create several cameras. The first camera should have a range of 0.1-1000 meters, the second would use the same near / far ratio and start where the first one left off, with a depth range of 1000-10,000 meters. Because the ratio of near to far ranges is what matters, not the actual distance, the numbers can get very big very fast. A third camera could be added with a range out to 100,000 kilometers!
      The trick is to set the new Camera::SetClearMode() command to make it so only the furthest-range camera clears the color buffer. Additional cameras clear the depth buffer and then render on top of the previous draw. You can use the new Camera::SetOrder() command to ensure that they are drawn in the order you want.
      auto camera1 = CreateCamera(world); camera1->SetRange(0.1,1000); camera1->SetClearMode(CLEAR_DEPTH); camera1->SetOrder(1); auto camera2 = CreateCamera(world); camera2->SetRange(1000,10000); camera2->SetClearMode(CLEAR_DEPTH); camera2->SetOrder(2); auto camera3 = CreateCamera(world); camera3->SetRange(10000,100000000); camera3->SetClearMode(CLEAR_COLOR | CLEAR_DEPTH); camera3->SetOrder(3); Using this technique I was able to render the Earth, sun, and moon to-scale. The three objects are actually sized correctly, at the correct distance. You can see that from Earth orbit the sun and moon appear roughly the same size. The sun is much bigger, but also much further away, so this is exactly what we would expect.

      You can also use these features to render several cameras in one pass to show different views. For example, we can create a rear-view mirror easily with a second camera:
      auto mirrorcam = CreateCamera(world); mirrorcam->SetParent(maincamera); mirrorcam->SetRotation(0,180,0); mirrorcam=>SetClearMode(CLEAR_COLOR | CLEAR_DEPTH); //Set the camera viewport to only render to a small rectangle at the top of the screen: mirrorcam->SetViewport(framebuffer->GetSize().x/2-200,10,400,50); This creates a "picture-in-picture" effect like what is shown in the image below:

      Want to render some 3D HUD elements on top of your scene? This can be done with an orthographic camera:
      auto uicam = CreateCamera(world); uicam=>SetClearMode(CLEAR_DEPTH); uicam->SetProjectionMode(PROJECTION_ORTHOGRAPHIC); This will make 3D elements appear on top of your scene without clearing the previous render result. You would probably want to move the UI camera far away from the scene so only your HUD elements appear in the last pass.
    • By tipforeveryone in tipforeveryone's Blog 12
      I spent a whole week for learning UE4 with cpp, yep, UE4 is a great engine for sure, but I found out that my mind could not understand the way UE4 works easily. It is too complex and made me tired. Then I returned to my Leadwerks project and felt so familiar. Soooo... sweet, everything is simple as it is
      It felt like I have had a long trip to UE city then return to my hometown. I miss Leadwerks indeed.
      Last year, I thought I could only use Leadwerks with LUA and never touch its CPP side. But I tried my best, learned Cpp for 8 months. Now I am not a cpp pro but I am confident in using this language. At least I can rewrite my whole project in CPP instead. this 3-years project helped me to understand my potential and interest in gamedev.
      I wish Josh be successful in progress of making Turbo, a new hope for much better Leadwerks.
      To all people who are using Leadwerks and help me these years, love you.
      ...
      Peace!
    • By Josh in Josh's Dev Blog 2
      Still a lot of things left to do. Now that I have very large-scale rendering working, people want to fill it up with very big terrains. A special system will be required to handle this, which adds another layer to the terrain system. Also, I want to resume work on the voxel GI system, as I feel these results are much better than the performance penalty of ray-tracing. There are a few odds and ends like AI navigation and cascaded shadow maps to finish up.
      I am planning to have the engine more or less finished in the spring, and begin work on the new editor. Our workflow isn't going to change much. The new editor is just going to be a more refined version of what we already have, although it is a complete new program written from scratch, this time in C++.
      It's kind of overwhelming but I have confidence in the whole direction and strategy of this new product.
×
×
  • Create New...