I am trying to check if the GPU can access the memory allocated by the system allocator (e.g malloc).
I am using Cuda 10.2 and the documentation mentions that the device should support pageableMemoryAccessUsesHostPageTables. but when I checked it on my machine, it returns 0.
I want to know if this feature is supported on x86?
It's not available on x86. It is available on Power9. You may wish to investigate managed memory (perhaps with oversubscription), or pinned host allocation using cudaHostAlloc