Optimal allocation of on-chip memory for multiple-API operating systems
ISCA '94 Proceedings of the 21st annual international symposium on Computer architecture
| Hi-index | 0.00 |
We have modified the X11 windowing system to use the native communication facilities of the Mach 3.0 microkernel. Our new implementation can rely on Mach''s low-overhead IPC facility as a direct replacement for sockets, or it can use shared memory as a transport between X11 clients and the server. On conventional BSD Unix systems, X11 communication is done through sockets. Because a user-level process implements Unix functionality on top of Mach 3.0, a socket-based version of X11 performs substantially worse than when running on a monolithic Unix kernel. Using Mach IPC as the transport between X11 clients and the server, X11 performance is slightly better than that of a monolithic system in which sockets are implemented inside the kernel as opposed to within a user level process. Using Mach''s shared memory facilities as the transport, we have measured performance improvements of over 40%.