cpu并行算法和gpu并行

In parallel programming, an embarrassingly parallel algorithm is one that requires no communication or dependency between the processes. Unlike distributed computing problems that need communication between tasks—especially on intermediate results, embarrassingly parallel algorithms are easy to perform on server farms that lack the special infrastructure used in a true supercomputer cluster.

在并行编程中，一种令人尴尬的并行算法是一种不需要进程之间通信或依赖性的算法。 与需要任务之间进行通信的分布式计算问题不同(特别是在中间结果上)，令人尴尬的并行算法易于在缺少真正的超级计算机集群中使用的特殊基础结构的服务器场上执行。

Due to the nature of embarrassingly parallel algorithms, they are well suited to large, internet-based distributed platforms, and do not suffer from parallel slowdown. The opposite of embarrassingly parallel problems are inherently serial problems, which cannot be parallelized at all.

由于令人尴尬的并行算法的性质，它们非常适合大型的，基于Internet的分布式平台，并且不会受到并行速度的影响。 令人尴尬的并行问题的本质是固有的串行问题，它们根本无法并行化。

The ideal case of embarrassingly parallel algorithms can be summarized as following:

令人尴尬的并行算法的理想情况可以总结如下：

• All the sub-problems or tasks are defined before the computations begin.在计算开始之前，所有子问题或任务都已定义。
• All the sub-solutions are stored in independent memory locations (variables, array elements).所有子解决方案都存储在独立的存储位置(变量，数组元素)中。
• Thus, the computation of the sub-solutions is completely independent.因此，子解的计算是完全独立的。
• If the computations require some initial or final communication, then we call it nearly embarrassingly parallel.如果计算需要一些初始或最终通信，那么我们几乎尴尬地称其为并行。

Many may wonder the etymology of the term “embarrassingly”. In this case, embarrassingly has nothing to do with embarrassment; in fact, it means an overabundance—here referring to parallelization problems which are “embarrassingly easy”.

许多人可能想知道“尴尬”一词的词源。 在这种情况下，尴尬与尴尬无关。 实际上，这意味着过多，这里指的是“非常容易”的并行化问题。

A common example of an embarrassingly parallel problem is 3d video rendering handled by a graphics processing unit, where each frame or pixel can be handled with no interdependency. Some other examples would be protein folding software that can run on any computer with each machine doing a small piece of the work, generation of all subsets, random numbers, and Monte Carlo simulations.

令人尴尬的并行问题的一个常见示例是由图形处理单元处理的3d视频渲染，其中每个帧或像素都可以互不依赖地进行处理。 其他一些示例是蛋白质折叠软件，该软件可以在任何计算机上运行，​​每台计算机只需完成一小部分工作，生成所有子集，随机数以及进行蒙特卡洛模拟。

翻译自: https://www.freecodecamp.org/news/embarrassingly-parallel-algorithms-explained-with-examples/

cpu并行算法和gpu并行