CUDAMemoryCopyToHost
CUDAMemoryCopyToHost[mem]
force copies CUDAMemory from the GPU to the CPU.
更多信息和选项
- The CUDALink application must be loaded using Needs["CUDALink`"].
- CUDAMemory is returned by CUDAMemoryLoad or CUDAMemoryAllocate.
范例
基本范例 (1)
First, load the CUDALink application:
Needs["CUDALink`"]This loads a Range[10] list into CUDALink:
mem = CUDAMemoryLoad[Range[10]]This gives information about the returned memory. Notice how the "DeviceStatus" is "Uninitialized":
CUDAMemoryInformation[mem]This forces a copy from the host memory (CPU) to the device memory (GPU):
CUDAMemoryCopyToDevice[mem]This gives information about the returned memory. Notice how both "HostStatus" and "DeviceStatus" are "Synchronized":
CUDAMemoryInformation[mem]This will perform an unnecessary copy from the device to the host:
CUDAMemoryCopyToHost[mem]The information is not changed:
CUDAMemoryInformation[mem]文本
Wolfram Research (2010)，CUDAMemoryCopyToHost，Wolfram 语言函数，https://reference.wolfram.com/language/CUDALink/ref/CUDAMemoryCopyToHost.html.
CMS
Wolfram 语言. 2010. "CUDAMemoryCopyToHost." Wolfram 语言与系统参考资料中心. Wolfram Research. https://reference.wolfram.com/language/CUDALink/ref/CUDAMemoryCopyToHost.html.
APA
Wolfram 语言. (2010). CUDAMemoryCopyToHost. Wolfram 语言与系统参考资料中心. 追溯自 https://reference.wolfram.com/language/CUDALink/ref/CUDAMemoryCopyToHost.html 年