Is there a restriction on the size of a block of memory I can allocate using a single cudaMalloc / cuMemAlloc under Windows 7?
Cuda 2.3 release notes [link] state the following under 'Known Issues for Windows Vista and Server 2008 :
There is no such mention referring to Windows 7. However, I ran an experiment that looped through various block sizes beginning from the memory size reported by the cudaGetDeviceProperties reducing each successive attempt by 100 MB. The maximum memory I could allocate on a Tesla C1060 (4GB memory size) was about 895 MB.
Is there a way I can get around this under Windows 7?
Thanks in anticipation!