应该重写什么样的应用程序才能使用 OpenCL？
Mac OS X 10.6 附带了 OpenCL，但是如果将它们重写为使用 OpenCL，有多少应用程序可以获得更好的性能呢？应该重写什么样的应用程序才能使用 OpenCL？
Mac OS X 10.6 comes with OpenCL, but how many applications could have better performances if they would be rewritten to use OpenCL? What kind of applications should be rewritten to use OpenCL?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（3）
我的公司正在使用 OpenCL 进行科学计算。虽然我不是使用它的开发团队的一员，但我可以告诉您他们使用它做的一些事情：图像的大规模数据处理、想象重建以及先前编写的科学代码的大规模并行化。
基本上，OpenCL 相当前沿。除非你想/必须处理它，否则最好留给那些需要它的人。我并不是想劝阻你，但它是一款专为利基市场设计的利基产品。为了潜在的就业目的，学习其他领域的编程会更好。
My company is using OpenCL for scientific calculations. While I'm not part of the development team which is using it I can tell you a few things they're using it to do: large scale data processing of images, imagine reconstruction, and massively parallelization of previously written scientific code.
Basically, OpenCL is rather cutting edge. Unless you want to/have to deal with it, it's best left for those that need it. I'm not trying to discourage you but it is a niche product designed for a niche market. You'd be better served studying other areas of programming for potential employment purposes.
OpenCL 并不是一个神奇的“提高性能”按钮。为了获得性能提升，需要满足几个先决条件，第一个是适当的硬件（Tesla GPU 会有所帮助）。至于可以从中受益的软件，它是任何进行大量并行计算的软件。这排除了大多数“日常”软件，并留下图形处理应用程序（渲染器、图像处理程序）、操作可能并行的大型数据处理应用程序以及科学应用程序。
通常，如果您编写一个可以从 OpenCL 中受益的应用程序，您就知道您已经可以从中受益了；）。
OpenCL isn't a magical "increase performance" button. There are several prerequisites that need to be satisfied in order to gain a performance boost, the first one is proper hardware (a Tesla GPU helps). As for software that could benefit from it, it's any software that does a lot of parallel computing. This excludes most "daily" software, and leaves graphics processing applications (renderers, image manipulation programs), large data-processing applications where operation may be parallel, and scientific applications.
Usually if you write a application that can benefit from OpenCL you know that you'll benefit from it already ;).
首先，您无法使用 OpenCL“重写”整个应用程序。您仍然需要一种用于 GUI 等的“正常”语言。
使用 OpenCL，您可以加速非常大的数据并行问题。大多数问题不属于这一类。
最有前途的候选人是图形/视频和声音处理。但每个需要对大量数据进行相同操作的问题都是合适的。
First of all you can't "rewrite" an entire application using OpenCL. You still need a "normal" language for the GUI etc.
With OpenCL you can accelerate very large data parallel problems. Most problems don't fall into this category.
The most promising candidates are graphics/video and sound processing. But every problem that requires the same operation on a very large set of data will be suitable.