• Advertisement


  • Content count

  • Joined

  • Last visited

Community Reputation

4722 Excellent

About Krypt0n

  • Rank

Personal Information

  • Interests
  1. to visualize it for your eyes, you might need to map your image to sRGB.
  2. The higher your resolution, the higher the spatial frequency, hence the perceived aliasing should increase. Tho, the downsampling on PS4 Pro to 1080p is really nice for AA. Digital Foundry dissects games in this respect and some games really give you all the options to choose from, e.g. recently: https://www.youtube.com/watch?v=rDKmJB629-0
  3. if you disable depth testing and set the transparency to be additive, it should look consistent (not correct, but always consistent). that way you'd know it's just the transparency sorting.
  4.   Are you 100% sure you have a problem with the shading? Isn't it maybe some transparency sorting? "bad" is not a good explanation for a problem ;). Post maybe a screenshot of the issue.
  5. if it depends on a coherent flag (e.g. a constant for the whole draw call), you can use dynamic branching nowadays, there won't be a big performance impact, might be even faster as you don't need to switch shaders that many times.
  6. The best limitation is competition. Make the market-share a key in the demand for the product. An example from real life is e.g. Windows Phone or Black Berry, they have a low share and nobody wants these, because nobody has these. Similar e.g. if you just look at the Android market. There might be phones as good or better than Samsung's S7 ot S8 (for half the price), but it's the most wanted one. And if you ask marketing people, they'll tell you that "mind share" of users is what they all aim for. On the lower end, you need to invest more than the revenue, to get mind share. On the higher end, it's self propelling. Hence, make the player realize, one big product gonna have more demand, will lead to higher prices, bigger margins, longer lasting market success. Allow them in some way to merge products, maybe at a high cost (e.g 100%+100%==150%), but eventually worth it. This would usually also lead to fusion or buying of competing companies.
  7. Krypton Development Team

    That's a wise choice of a team name :)
  8. haven't tested it, just my brain fart: template< typename TYPE > int GetLod( const Tuple3< TYPE >& _a, const Tuple3< TYPE >& _b ) { const Tuple3< TYPE > ddx = a-b; return LeadingBit(dot(ddx,ddx)); }the idea is: usually, you would call length(ddx), and then increase the lod with every doubling of the distance. (assuming LOD0 is highest quality). taking the leading bit of the squared distance combines both, but runs with integer only. for "LeadingBit" you could exploit that intrinsic: https://msdn.microsoft.com/en-us/library/bb384809.aspx
  9. Shadow matrix calculation

    is the problem the shadow mapping or the rendering of the shadow map? Have you made a capture of your rendering and checked your shadow map? if you make a capture, check the vertexbuffer before and after the vertex shader (for shadowmap rendering). this might already show the issue. (you can use e.g. NSight)
  10. Shadow Mapping Issue

    A simple trick: offset the sampling position by the normal of the surface that receives the shadow. kind of float bias=0.001; float4 positionInLightProjection = mul(float4(thisWorldPosition+thisWorldNormal*bias,1.0f),lightProjection(ilight));
  11. General solution Virtual Textures

    I'd go for 2. it keeps everything unified. the "virtual texture" on disc is nothing more than any other texture, beside being freaking huge, but that's just numbers. you don't really care at the high level. (you might care for optimization, e.g. tiling)   for the first implementation, don't go banana with sophisticated uv-space management (bsp etc.), in theory the uv space is unlimited. you will run into some issues at some point eventually, but you might solve that in a simpler way than writing managers to solve issues you "might run into" ;)
  12. I'd give the player abilities that don't all work together, hence to be "prepared" would mean to have the right ability/mode selected. e.g. in yakuza you can choose different fighting styles. Add on top, that you only can switch styles when you're at full health, hence even a light encounter that surprises you would force you to fight in a sub-optimal mode. (you'd be dead, without admitting it).  Instead of "healthy" you could also use "fitness" instead. Hence running through the map (unprepared for fights) would render you exhausted and you wouldn't be able to switch into a fight mode. To be always prepared, you'd need to always walk slowly.
  13. I guess you decompress the position (i11:i11:i10 -> f32:f32:f32) and then add the offset of the chunk, as float. This will/would result in float precision issues. Try to use integer offsets, therefor you decompress i11:i11:i10 to i32:i32:i32, add the integer offset and just now convert to f32:f32:f32.
  14. A Bloodborne based Shooter game.

    I like the basic idea. (As I also like the difficulty and challenge of Dark Souls, yet I don't like the basic combat mechanics, it's somehow not my taste).   Wouldn't a Dark Souls as shooter boil down to pure boss fights? That is somehow appealing :). To start small, it could be an shump imo. The basic idea wouldn't change much between 3rd person 3d and top down 2d.
  15. The explained solutions were mostly for texture atlases. Mega textures work slightly different. 1. There is no hardware mip mapping, this would break quite a bit of the tech. It actually avoids these issues by handling mip filtering by the pixel shader. 2. The tiles always ( on every mip level) work the same and the border size depends solely on the filtering you use (bilinear, anisotropic,..), 1 pixel for bilinear. 3. The Mips don't include the border, the mips are only generated based on the border-less base texture. Then you divide each level into equally sized tiles, e.g. 126x126 and add the 1 pixel border. 4. While the border might sound problematic, the alternative would be to always load the surrounding 8 tiles into the texture as well, to filter properly cross border, that would be 9x the memory usage. 5. please don't mix atlas textures and mega textures. Atlas is the idea of fitting a lot of textures into one. Mega texture is actually the opposite, to have one big textures (no borders, a triangle could map it completely). The tiling is only a solution to low level memory organization. On modern hardware, you could use Partially Resident Textures (PRT), that can solve a lot of the Mega texture issues. Texture Atlases would still have bleeding issues etc.
  • Advertisement