|
|
|
| Table 10.14 Multiprocessor area. |
| | Integer ALU | | FP multiplier | 20.3 | | Integer register | | Divide supprot | 3 | | Shifter | | FP unit | 37.8 | | Incrementor | | 2x FP unit | 75.6 | | PC unit | | | | | 2 TLBs | | Latches | 10.68 | | Decode + control | | Bus | 42.72 | | Cache controller | | | | | Bus logic | | Total multiprocessors | 160.20 | | Stored buffer + bypass | | | | | Load/store | | Die area | 230 | | Clock generator | | Area available | 184 | | Integer unit | | Area in ''A" | 335.11 | | Added branch adder | | | | | | Remaining area | 174.91 | | 2x integer unit | | 10% aspect ratio | 17.49 | | Remove one TLB | | Cache area | 157.41 | | Total integer | | | | | FP RegFile | | Cache bits | 388,289.96 | | FP adder | | Cache KB | 47.40 |
|
|
|
|
|
|
The in-line reference traffic is calculated as follows: |
|
|
|
 |
|
|
|
|
Iin-line = (1/m) * IF access time * instr/IF, |
|
|
|
|
|
|
|
|
wherem is the number of cycles between instruction issues, IF access time is the number of cycles required per access, and instructions per IF is the width of the interface between the I-cache and the I-buffer. For the baseline case: |
|
|
|
|
| | |
|
|
|
|
1 (maximum of 1 instruction decoded per cycle). |
|
|
|
| | | | | | |
|
|
|
|
1 or 2 (interface width of 4 or 8 bytes). |
|
|
|
|
|
|
| Instr/IF | I (in-line) | | 1 | 1 | | 2 | 0.5 |
|
|
|