#225 closed defect (wontfix)
After merge measuring branch to mainline top echoes weird load
| Reported by: | Stanislav Kozina | Owned by: | Martin Decky |
|---|---|---|---|
| Priority: | major | Milestone: | 0.4.3 |
| Component: | helenos/unspecified | Version: | mainline |
| Keywords: | Cc: | ||
| Blocker for: | Depends on: | ||
| See also: |
Description (last modified by )
Top from measuring branch echoes load between 0-1 immediately after system starts, which looks good.
After merge to mainline branch it echoes load sometimes bigger than 40, which is definitely wrong.
Change History (2)
comment:1 by , 15 years ago
| Resolution: | → wontfix |
|---|---|
| Status: | new → closed |
comment:2 by , 15 years ago
| Description: | modified (diff) |
|---|---|
| Summary: | After merge measuring branch to head top echoes weird load → After merge measuring branch to mainline top echoes weird load |
Note:
See TracTickets
for help on using tickets.
I believe that the load values are perfectly reasonably — there is a lot of work happening just during the bootstrap process. For some reason the load calculating task gets scheduled a bit later in the measuring branch and misses this huge peak (this can be easily emulated by adding thread_sleep(1) at the beginning of the load calculating thread in mainline).
The problem comes from the fact that the load value is sampled just once during the whole 5 second period in this classical algorithm. Perhaps a completely different algorithm might be more suitable for a microkernel system with a lot of context switches (like recording the maximum ready queue length in the scheduler during the whole period).
I close this ticket for now. If you really believe that there is a specific bug in the mainline (which I really can't spot), feel free to reopen it.