T. Jaeger et al. Achieved IPC performance. In 6th Workshop on Hot Topics in Operating Systems (HotOS-VI), 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....minimizing the cache footprint. This leaves 167 instructions for the core cost of a PPC call and return, which compares favorably to the corresponding cost of 158 instructions for two one way calls for one of the fastest uniprocessor IPC systems running on the same (MIPS R4000 based) processor [20] . In addition, up to 8KB of data can be exchanged between client and server (in both directions) through remapping a region for an additional cost of only 82 instructions. The final component of the PPC system, remote PPCs, are like regular PPCs with a pair of remote interrupts on the call and ....

....IPC overhead has become largely irrelevant [2] 20 Although many results have been reported 20 We do not agree with these arguments. over the years on a number of different platforms, the core cost for a call return pair (with similar functionality) is usually between 100 and 200 instructions [11, 12, 17, 20] . However, the Tornado PPC facility goes beyond the standard techniques used to optimize IPC performance, by optimizing for the multiprocessor case by eliminating locks and shared data accesses, and by providing concurrency to the servers. The key previous work done in multiprocessor IPC was by ....

T. Jaeger et al. Achieved IPC performance. In 6th Workshop on Hot Topics in Operating Systems (HotOS-VI), 1997.

....minimizing the cache footprint. This leaves 167 instructions for the core cost of a PPC call and return, which compares favorably to the corresponding cost of 158 instructions for two one way calls for one of the fastest uniprocessor IPC systems running on the same (MIPS R4000 based) processor [19] . In addition, up to 8KB of data can be exchanged between client and server (in both directions) through remapping a region for an additional cost of only 82 instructions. The final component of the PPC system, remote PPCs, are like regular PPCs with a pair of remote interrupts on the call and ....

.... point where it has been argued that the IPC overhead has become largely irrelevant [2] 20 Although many results have been reported over the years on a number of different platforms, the core cost for a call return pair (with similar functionality) is usually between 100 and 200 instructions [10, 11, 16, 19] . However, the Tornado PPC facility goes beyond the standard techniques used to optimize IPC performance, by optimizing for the multiprocessor case by eliminating locks and shared data accesses, and by providing concurrency to the servers. The key previous work done in multiprocessor IPC was by ....

T. Jaeger et al. Achieved IPC performance. In 6th Workshop on Hot Topics in Operating Systems (HotOS-VI), 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC