Several operations are tested below on all architectures on which NBS has been extensively tested. The figures given are for a noticeboard with 100 items all at the same level and are the lowest measured values of CPU microseconds:
Test # | Test name | ![]() |
DEC Alpha AXP 3000/400 | SPARCstation 10 |
0 | Define, save and find noticeboard | 820000 | 16666 | 16666 |
1 | Scalar assignment | 1 | 0.1 | - |
2 | Put scalar | 36 | 2.0 | 1.6 |
3 | Get scalar | 34 | 1.9 | 1.0 |
4 | Move 10 longwords directly | 17 | 0.3 | 0.6 |
5 | Put 10 longword array | 44 | 2.2 | 3.3 |
6 | Get 10 longword array | 43 | 2.2 | 3.2 |
7 | Move 1024 longwords directly | 591 | 16 | 87 |
8 | Put 1024 longword array | 670 | 33 | 89 |
9 | Get 1024 longword array | 646 | 32 | 89 |
10 | Find 100th item by position | 277 | 35 | 23 |
11 | Find 100th item by name | 544 | 58 | 31 |
12 | Enquire name | 42 | 1.2 | 2.5 |
13 | Enquire size | 15 | 0.7 | 1.0 |
These are fairly consistent with the times reported for the MON system. They are also pretty close to those for V2.2 of the noticeboard system and are better in some cases, presumably partly because of the greater use of inline code in V2.3.1 onwards. The speed of the UNIX machine is particularly noticeable on file related operations where the speed up is an order of magnitude greater those operations simply involved with moving data around.
NBS The Noticeboard System