AMP FFT库产生不正确的结果 [英] AMP FFT library produces incorrect results

问题描述

你好.我终于有了AMP FFT库，并且已经针对我自己的多核CPU实现用3种不同的图像大小(1024 x 1024、2048 x 2048、4096 x 4096)对其进行了测试.对于前两个图像尺寸，它会产生正确的结果，但是对于 它运行的最后一个(4096 x 4096)，但返回由全零组成的结果.有人可以复制吗?

Hello. I've finally got the AMP FFT library working and I've tested it with 3 different image sizes (1024 x 1024, 2048 x 2048, 4096 x 4096) against my own multicore CPU implementation. For the first two image sizes it produces the correct results, but for the last one (4096 x 4096) it runs but returns a result consisting of all zeros. Can anyone else reproduce this?

-L

推荐答案

感谢您报告此问题.我们已经能够在最终复制它.如果您 f ，您将切换到调试"您应该在Visual Studio中进行配置 看到4096 x 4096的异常，表示Dispatch调用中的线程组数超出了X维度的限制.我们正在努力查看是否可以取消此
限制.

Thanks for reporting the issue. We have been able to reproduce it on our end. If you would switch to "Debug" configuration in Visual Studio you should see an exception for 4096 x 4096 saying that the number of thread groups in Dispatch call exceeded the limit in X dimension. We are working to see if we can lift this
limitation.

感谢您的耐心等候，

这篇关于AMP FFT库产生不正确的结果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！