ray-tracing | Ray TracingPhong Model Phong Shading | Graphics library

To preface this, I wanted an interface, and figured that C++20 had an interface mechanic. I have never used C++20, and found concepts about 2 hours ago. So the mistake here could be something really simple.

Suppose I have an imperative 3rd party library I would like to abstract away because imperative code is messy.

So, I define some very specific function that accepts the 3rd party library class:

I'm currently looking at some resource on how to enable X-ray for my stepfunction statemachine, from this tutorial: https://docs.aws.amazon.com/step-functions/latest/dg/concepts-xray-tracing.html#xray-concept-create

"When you enable X-Ray for an existing state machine, you must ensure that you have an IAM policy that grants sufficient permissions for X-Ray to perform traces. You can either add one manually, or generate one. For more information, see the IAM policy section for X-Ray."

There are so many permissions for X-ray, I wonder which ones I need? I'd like to add them to my step function role manually, but I don't want to add all of them including the unneeded ones.

In threejs, I successfully import and display a glb file of a house (the house has two floors and various objects on each floor, among which chairs, some tables, a kitchen etc, such as a real house). My problem is that I am able to make the house and its objects reflect the environmental light but I am not able to make the objects of the house reflect themselves. My ultimate goal would be to implement some real time ray-tracing, but at this point I would be happy to only generate some real time reflections which in addition to the environmental light also reflect the other objects of the house. Same with shadows.

I have not found anything online about it this type of reflections. Does anyone know a good place where to start? Or if you have faced a similar challange, how did you solve it?

I define a class. I want to add some texture with my ray-tracing code with CUDA. And I use new when call the constructor.

I have an application (based on the vulkan-tutorial.com) in which I use the titular raytracing extension for vulkan. In it, an acceleration structure is created for some geometry. This geometry then changes (vertices are displaced dynamically, per frame), and thus I update the appropriate BLAS by calling vkCmdBuildAccelerationStructureKHR with VkAccelerationStructureBuildGeometryInfoKHR::update = VK_TRUE. This works fine (although the update ignores my changing the maxPrimitiveCount and similar parameters - It uses as many primitives as I specified during the first build; somewhat makes sense to me and is not part of my question).

I've researched a bit and came across some best practices here: https://developer.nvidia.com/blog/best-practices-using-nvidia-rtx-ray-tracing/ In there, they mention this: "Consider using only rebuilds [of the BLAS] with unpredictable deformations." This seems like something I want to try out, however, I can't find any sample code for rebuilding BLAS, and if I simply set update to VK_FALSE, I get massive amounts of validation layer errors and no image on screen. Specifically, I get a lot of "X was destroyed/freed but was still in use" where X is command buffers, VkBuffers, memory, fences, semaphores... My guess is the rebuild is trying to free the BLAS while it's still in use.

My question is therefore: How do you properly perform a "rebuild" of a BLAS, as mentioned in the above article?

I was considering using some std::shared_ptr to keep track of the BLAS being still in use by a given swapchain image but that seems excessively complicated and somewhat unclean, besides, I would need as many BLAS as I have swapchain images, multiplying required graphics memory by the swapchain size... that can't be practical in real life applications, right?

I don't know why, but ray-triangle intersection algorithm (Möller-Trumbore and Watertight) works with 2 directions.

from origin to end - how it must work

from end to origin - the problem

white cross - an unnecessary intersection

screenshot

I tried to use 3 component vectors, so I have removed all W values from code.

Ray

My original code was -

I would like to get the intersection point of a line (defined by a vector and origin) on a triangle. My engine use right handed coordinate system, so X pointing forward, Y pointing left and Z pointing up.

---- Edit ----

With Antares's help, I convert my points to engine space with:

In my raytracer all surfaces are centered at the origin and oriented on Y axis. Displacement, rotations and resizing are obtained through transformation matrix applied on rays.

I recently rendered a torus in my ray-tracing using its Cartesian equation:

(x^2 + y^2 + z^2)^2 - 2 * (r1^2 + r2^2) * (x^2 + y^2 + z^2) + 4 * r1^2 * y^2 + (r1^2 - r2^2)^2

to which I replaced every point with the ray equation:

ex: X = Ray.ori.x + T * Ray.dir.x;

With the ray components replaced in the equation, I got the 5 coefficients of my quartic function which can be used to find the equation roots (the T intersections) with a 4th degree polynomial solver algorithm.

I was wondering if a mobius strip can be rendered the same way. My research did not bring up much, I found some Raytracing codes using cubic equations but copying the 4 coefficients led me to incomprehensible forms and artifacts.

Could you help me to render it? Also advice to render it with another method is welcome.

Thanks!