| 1 |
Hello PaulNM, |
| 2 |
|
| 3 |
|
| 4 |
PaulNM wrote: |
| 5 |
> Hey folks, |
| 6 |
> |
| 7 |
> I'm wondering if anyone has experimented with different values for |
| 8 |
> MAKEOPTS="-jx" on Athlon 64 X2's. The guides usually suggest # of |
| 9 |
> cores/processors plus one. Looking at "man make.conf" doesn't help |
| 10 |
> much, it just says "Suggested settings are between CPUs+1 and |
| 11 |
> 2*CPUs+1". The only thing "man make" mentions about the "-j" option |
| 12 |
> is "If the -j option is given without an argument, make will not |
| 13 |
> limit the number of jobs that can run simultaneously." I've also seen |
| 14 |
> some mentions of -j9 by people with dual core systems elsewhere, |
| 15 |
> indicating it drastically speed up some emerges. |
| 16 |
> |
| 17 |
> So this leaves me with several possibilities. There's -j3 (current |
| 18 |
> setting), -j5, -j (no limit), -j9, or some other random number. Any |
| 19 |
> suggestions/warnings/links would be appreciated. |
| 20 |
> |
| 21 |
> Thanks, |
| 22 |
> PaulNM |
| 23 |
> |
| 24 |
|
| 25 |
On my 4x Dual-Core Opteron system, -j9 actually worked as expected, but |
| 26 |
not as much on the other system with two CPUs. |
| 27 |
|
| 28 |
Is that what you want? |
| 29 |
|
| 30 |
Here is the story. |
| 31 |
The value of MAKEOPTS including -jx is passed to the `make' command so |
| 32 |
that x kid jobs at most can be executed. If 1+#CPU is given as the |
| 33 |
parameter, each CPU is assigned a single kid process because make |
| 34 |
command itself does have much to do while the kids are working hard. |
| 35 |
Giving the larger number just let them do extra job to switch among kids |
| 36 |
in expense of performance down. |
| 37 |
|
| 38 |
te |
| 39 |
-- |
| 40 |
gentoo-amd64@g.o mailing list |
Archives