使用OpenGL，结果是否取决于安装的硬件或驱动程序? [英] Using OpenGL, are the results dependent on hardware or drivers installed?

问题描述

我正在按照此处的说明使用OpenGL遮挡进行渲染:(https://developer.download.nvidia.com/books/HTML/gpugems/gpugems_ch29.html )

I am using OpenGL occlusion for a render as instructed here: (https://developer.download.nvidia.com/books/HTML/gpugems/gpugems_ch29.html)

我在Jupyter上运行相同的代码(它使用"Cirrus Logic GD 5446"作为GPU，在VM架构师JupyterHub上仿真了25年的PCI图形芯片)和本地计算机(使用GeForce 840M)).

I am running the same code on Jupyter (it uses "Cirrus Logic GD 5446" as a GPU, a 25 year old PCI graphics chip that is emulated on the VM architecturer JupyterHub) and on a local machine (with GeForce 840M).

该代码在两种环境下均可工作，但是为每个构面生成的像素数量略有不同.本地执行更加精确，并且运行速度快了两倍.

The code works on both environments, however, there is a slight difference in number of pixels that are generated for each facet. The local execution is a bit more accurate and also runs over two time faster.

是否有可能在所有机器上实现类似的性能和结果，或者这不是我们手中?

Is it possible to achieve similar performance and results on all machines or this is something not in our hands?

推荐答案

OpenGL规范对可重复性(不变性)非常明确:

The OpenGL specification is very clear about reproducibility (Invariance):

https://www.khronos.org/registry/OpenGL/specs/gl/glspec46.core.pdf 附录A

仅对于同一实现上的相同操作序列，才保证得到相同的结果.仅更改一项操作，否则实现和结果可能会有所不同.

Identical results are warrented only for identical sequence of operations on the same implementation. Change only one operation, or the implementations and results may vary.

这篇关于使用OpenGL，结果是否取决于安装的硬件或驱动程序?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！