 emil28092005
|
e6e305dfa9
|
feat: shadow mapping — depth-only render pass from light POV + PCF
- VulkanShadowMap.cs: 2048x2048 D32_SFLOAT image, sampler (clamp-to-border),
separate shadow pipeline (depth-only, no color, depth bias enabled, CULL_NONE)
- Shadow shaders: shadow.vert (lightViewProj * model * pos), shadow.frag (empty)
- Main shaders updated: fragLightSpacePos output from vertex, PCF 3x3 in fragment
- Push constants expanded to 160B: model(64) + lightPos(16) + lightColor(16) + lightViewProj(64)
- 3 push constant ranges: Vertex(model), Fragment(light), Vertex|Fragment(lightViewProj)
- Descriptor set: 2 bindings — UBO(vp) + CombinedImageSampler(shadowMap)
- Shadow pass: before main pass, renders scene depth from light position
- Image transitions: shadow map UNDEFINED→DEPTH→SHADER_READ_ONLY each frame
- vkCmdSetDepthBias(1.25, 0, 1.75) for acne prevention
- Light VP: CreateLookAt(lightPos, origin, up) * Perspective(60°, 1.0, 0.1, 60)
- 0 validation errors on build
|
2026-06-18 20:09:32 +03:00 |
|