Communication between threads in the shared address space model is primarily done using shared variables that act as locks or as data. How is the mapping of each thread's virtual address space determined as to ensure both common and private portions of physical location?
This comment was marked helpful 0 times.
edwardzh
@rutwikparikh Normally, the operating system has a tree to find the physical memory regions that contain a given virtual address, and the tree's nodes contain information on what is private and what is shared. Likewise, the kernel virtual memory needs to be concurrent to support this.
NVIDIA Cuda uses option 2 with keyword shared.
This comment was marked helpful 0 times.
Communication between threads in the shared address space model is primarily done using shared variables that act as locks or as data. How is the mapping of each thread's virtual address space determined as to ensure both common and private portions of physical location?
This comment was marked helpful 0 times.
@rutwikparikh Normally, the operating system has a tree to find the physical memory regions that contain a given virtual address, and the tree's nodes contain information on what is private and what is shared. Likewise, the kernel virtual memory needs to be concurrent to support this.
This comment was marked helpful 0 times.