The connection machine
ACM SIGARCH Computer Architecture News
| Hi-index | 0.00 |
The operation that causes the greatest slow-up in the execution of computer code on systems with cache is branching. We propose a simple solution to the problem of doing look-ahead branch prediction. This method can be implemented totally in hardware.We also suggest that cache be partitioned to reduce idle CPU cycles during update of cache memory. The ideal cache would consist of at least 4 partitions. Two for data and two for instructions.