Ray-projection polynuclear parallel body drawing method

Abstract

The invention discloses a ray-projection polynuclear parallel body drawing method based on Cell B. E., comprising the steps of reading and preprocessing three-dimensional data, dividing task for a screen image space according to rows through a control processing unit, allocating the task to a coprocessor unit course sequentially, setting a three-dimensional scene and drawing attributes, drawing each pixel on the screen and displaying a result image. The ray-projection polynuclear parallel body drawing method based on Cell B. E. provided by the invention firstly adopts an improvement of task division of PPE and SPE, and then adopts a performance optimization of an SPE program, carries out SIMD operation for the most important tasks such as intersection test, coordinate conversion, light and shade account, color synthesis, and the like in the SPE drawing procedure and adopts an infiltration technology to effectively solve irregular accessing-memory problem in the ray-projection method, thereby finally improving the performance of the program.

Description

technical field [0001] The present invention relates to a scientific computing visualization processing method, in particular to an improvement of a ray-casting parallel volume rendering method based on the CellB.E. architecture. Background technique [0002] The scientific visualization technology represented by volume rendering technology transforms abstract data into an intuitive form that is easy to be accepted and understood - graphics, which provides a powerful tool for understanding and discovering various phenomena and laws in the process of scientific computing. The application fields of scientific computing visualization are very extensive, involving almost all natural sciences and engineering fields. . [0003] In addition, scientific computing visualization can also be applied to fields such as space exploration, astrophysics, and mathematics. Volume rendering technology can generate an overall image of a three-dimensional data field, including a large number o...

AI Technical Summary

Problems solved by technology

[0006] First, the ray-casting volume rendering method needs to perform rendering calculations for each pixel on the screen, and the rendering calculations for each pixel include time-consuming calculations such as sampling, interpolation, shading calculations, and color synthesis; when the viewing direction changes, the data in the data field The front-back relationship between the sampling points also changes, and all pixels on the screen have to be redrawn, so the amount of calculation is extremely large

[0007]Secondly, when the ray-casting volume rendering method draws the volume data, it needs to draw the volume data from different perspectives, and carry out light along the line of sight in different directions Projection, sampling, interpolation, etc.; when performing cubic linear interpolation on sampling points, it is necessary to read discontinuous voxel data, so on the whole, there is no fixed rule for reading voxel data, due to a large number of irregular The existence of memory access leads to a lack of obvious locality in memory access. When running on a common platform, the hit rate of Cache and TLB is extremely low, which leads to low memory access efficiency and ultimately leads to poor overall performance.

[0011] However, the design and implementation of the algorithm level cannot fully exploit the performance of the Cell B.E. system, and it is necessary to combine the characteristics of the system level such as the heterogeneous multi-core structure, computing mechanism, memory access mechanism, and communication mechanism of Cell B.E. algorithm for optimization

The limitation of SPE's local memory capacity, the limitation of DMA data transmission, the synchronization and communication between PPE and SPE also bring new challenges to the ray-casting volume rendering method

Images