Jump to content

klepto2

Developers
  • Posts

    929
  • Joined

  • Last visited

Everything posted by klepto2

  1. I promote this suggestion because: In the latest weeks you made a great progress with the editor, but while testing the editor with lots of files i found some things which could be made easier if a small clear cache function is integrated: Also this function is more or less a very easy one to implemnt. In the latest updates you suggested multiple times to delete the thumbnail folder The overall cache can grow extremely in size and to easily clean it up from the editor would be handy the clear cache should be optionally global or per project
  2. As a small feature of my upcoming atmospheric scattering integration, i am working on a realtime ibl sampler based on the Khronos Ibl Sampler. After some early experiments, i was able to come up with a small tool which currently takes an input cubemap and generates the proper textures via compute pipelines. I showed some early progress here as i found some small issues. Planned features are similar to the Khronos one, convert panorama images to cubemaps, etc. also i will add some cli, but this is something i need to figure out if this is possible. I will also possibly provide a lib with it which can be used to perform realtime updates (at lower resolutions of course). the nicest thing in the tool is the usage of host synchrinsation, which means, that i can download purely gpu generated data back to the CPU.
  3. Fixed the blurriness: The color is not as bright as the one from the ibl sampler as it uses dds or tex files to gnerate the filtered cubemap, which is different to the hdr format. Maybe i can find a soltuin for that as well.
  4. The code is the same as the one from Khronos. What might be, is that I have one step missed. But the mipmaps are generated the same way. And before downloading they look smooth. But will take a closer look.
  5. this might be due the type of host synchronization i do. I currently use vkCmdCopyImage which doesn't support linaer filtering for the download. I will later try to change it to staging buffers, which might be better in this case.
  6. The Zone-map with the provided zone-sky only. the generated specular and diffuse maps in action this is a current screen of the tool with the settings used: textures_generated.zip
  7. It is saving the dds file with the wrong mipmap count, but it will not load in ultraEngine, while other tools load it, but display a black screen.
  8. Ok, found the error. I generated to many mipmaps. BCH6 has a blocksize of 4 which i missed, which means the lowest mipmap size is also 4 and not 1. Thanks for the point in the right direction.
  9. The saved cubemap is black, the pixmaps in the mipchain passed to the save texture function are valid. They are saved correctly, when the pixmap save method is used. Just the provided way in the blog sample and the sample itself is not working. If you still need some data i will provide it later. The cubemap data is calculated in a computeshader and the result is mapped back to the host. For the later atmosphere shader I just need the calculations, but for the engine I thought I could provide a small tool which also saves the generated cubemaps. I know iblsampler, and I use some shader code from it.
  10. The reason why cubemaps are not shown in the asset browser is, that the pbr shaders don't support cubemaps as diffuse texture. You can try it yourself and create a material and try to assign a cubemap to the diffuse slot. In this case SaveTexture just saves black color instead of the actual pixel values. I know that it worked some time ago as i have tried the code mentioned in the blog before. Maybe Josh just hasn't applied some internal changes which were made lately. Btw: yes i tried them in the Editor as well
  11. the noise texture was ok, the other noise texture i use is closer to the original. The artifacts came from this line: p.x *= gl_FragCoord.x/buffersize.y; this: should have been: p.x *= buffersize.x/buffersize.y; this line calculates the aspect ratio and if you use gl_FragCoord.x it is dependend on the pixel position which leads to the artifacts you have seen.
  12. I am working on a small tool to generate the specular and diffuse maps based on an input cubemap: While the generation itself works flawlessly, I can't get the SaveTexture feature to work with cubemaps. I download the image data from the gpu and if i save the pixmaps (each face and mipmap) seperately this works. But with a mipchain i only get a black cube texture. (Tried it with various viewers, renderdoc, nvidia texture tools). Then i tried the sample you provided here: and followed the same workflow as you. The result is the same, the code in this blog also results in a black texture
  13. Try this: i have added the diffuse layer, so the background is visible. 1. I changed the noise texture to the one used by shadertoy (using other textures might lead to artifacts) 2. I corrected the aspect ratio calculation you had p.x*= gl_FragCoord.x / buffersize.y; where is should be p.x*= buffersize.x / buffersize.y; Effect.zip
  14. You can upload the pfx and textures here, and i can look at it if you want.
  15. yeah, that is what most users do with shadertoy shaders. But gpu noise is far too expansive to be used in realtime scenarios (espacially when it is a game or so) It is always better to use a noise texture than to calculate the noise with the various available noise methods for glsl. 1. Performance: noise calculation on the gpu is much (and i mean very much) slower than a simple texture lookup. 2. low end hardware: most noise mehtods have something like this: float hash( vec2 p ) { float h = dot(p,vec2(127.1,311.7)); return fract(sin(h)*43758.5453123); } but this will result in unexpected results on low end hardware (espacially integrated gfx)
  16. But i agree, with the proper background the rain would be visible without the need of muliplication.
  17. While this is true, it isn't needed in this case. The shader on shadertoy also just uses a black background, which might be what he currently wants. The blending with the actual background is not yet the problem. The problem is, that the calculated value for 'f' is just to low to be recognized in the output. a multiplication by 2.0 or 4.0 will move the rain into a more visible spectrum. Another problem might be the pfx lua script. This would be nice to see as well to figure out if the noise texture is correctly bound.
  18. Read the comment on shadertoy: change this line fragData0=vec4(col, 1.0 ); to fragData0=vec4(col * 4.0, 1.0 ); and you should be able to see the effect.
  19. One way is to use a post effect, I can't fully reply currently, but this is a starting point: DayNightEffect.zip You attach the pfx to your scene, and if you have a driectional light in the scene it will automatically find it and use the correct light vector.
  20. Maybe just simple: provide all 3 and just make a default setting in the editor. This way the user can choose which he prefers.
  21. Another option (which would be IMHO more convenient) would be to add Import and Eport menues. This way you can later support multiple map formats without breaking the original save and save as workflow. Load/Save/Save as -> only for UltraEngine files Import/Export -> reserved for plugins or fixed imports/exports in the engine.
  22. Also some of the objects are not exported correctly, some are scaled or positioned wrong.
  23. If you load a leadwerks map with a terrain (lets say the-zone.map) , resave it as a .ultra map and reload the saved ultra scene, then the terrain layers are there, but all shown as empty. I assume that might be as the materials for these layers are generated on the fly by the leadwerks map loader (you can see that, if you try to edit a layer material in the imported leadwerks map) and the new scene can't find the materials.
  24. If i load the 'The-Zone"-map the performance for navigating in the Perspective viewport is fast. But when i select the terrain and try to navigate, the controls begin to feel laggy. If i select any other element in the scene, the camera-navigation is smooth again.
×
×
  • Create New...