AMD introduces HIP RT, allowing a ray tracing structure for developers that is straightforward and simpler to use with several options
HIP RT is built in a scarcely diverse way by stopping the necessity to comprehend numerous next-gen kernel types. HIP RT presents new object types like hiprtGeometry and hiprtScene. When the geometric data is transmitted to HIP RT, the procedure creates the data structure and then is enacted to the HIP kernel. This will permit the device-side library API to complete an intersection test. AMD’s HIP RT is created to permit developers to utilize the Ray Accelerators’ complete advantages for ray tracing in AMD RDNA 2 GPUs. AMD RDNA 2 architecture-based graphics processors support hardware ray tracing acceleration to optimize render times. Before, applications that support HIP had not had access to operating this hardware acceleration. AMD has introduced a few basic features necessary for ray-tracing capabilities since this is the first release for the new ray-tracing acceleration support. Features such as:
In general, scenes are comprised of many meshes. The triangle mesh is the basic primitive of HIP RT from which we build hiprtGeometry. Developers can utilize hiprtScene support on top of triangular meshes to construct an acceleration structure to locate a mesh more effectively, especially when a ray is bisecting multiple meshes. Developers will find it easier to locate the closest hit of a ray or capture all hits. This process can be helpful for transparencies and shadows. HIP RT builds an acceleration structure in the library to create the ray-tracing acceleration. The new acceleration application delivers numerous options, allowing users to utilize a specific format they require, allowing for more customization. Users can choose a high-quality build that takes longer to produce, resulting in a high-quality data structure that makes individual ray tracing queries more efficient. Advanced users can employ HIP RT to load individually constructed bounding volume hierarchies (BVH), which the HIP RT can use to complete the intersection on the GPU. This is specific to HIP RT and cannot be found on other ray tracing APIs. Acceleration builders to access triangles from axis-aligned bounding boxes (AABBs), set for dynamically changing objects when processed by the GPU. With HIP RT, users can complete the geometry update logic located on the GPU, write the new AABBs to a specific buffer, and transmit it to HIP RT. The core data stays with the GPU when the buffer is transferred with AABBs. Ray-tracing supports primitives that are not triangular by offering the ability to input a custom intersection function in the formula, allowing for the code to become registered in HIP RT and called by the process in the future. AMD’s motion blur technology is more straightforward for the developer to program for finite virtual camera exposure times.
Developers interested in utilizing AMD HIP RT can find the application on GPUOpen. News Source: Phoronix
