Thanks to Con and everyone for making this necessary.
The contest (http://contest.kolivas.net) has been added to
the OSDL Scalable Test Platform. (http://www.osdl.org/stp)
Sample results for a 2- cpu system with 2.4.20-pre5-ac1
Test Time(sec) CPU % Ratio
noload 61.95 195% baseline
process_load 69.45 170% 1.12l
io_halfmem 67.86 179% 1.10l
io_fullmem 100.57 121% 1.62l
mem_load 96.06 129% 1.55l
Results for a single-CPU system ( 2.4.20-pre5-ac1 )
noload 104.65 99% baseline
process_load 116.80 88% 1.12l
io_halfmem 117.81 89% 1.13l
io_fullmem 128.72 82% 1.23l
mem_load 143.17 77% 1.37l
Full report: http://khack.osdl.org/stp/5235/ ( 2-cpu)
http://khack.osdl.org/stp/5234/ ( 1-cpu)
Would very much appreciate your comments. One note:
The '-j' value is adjusted = number of cpu's * 4
However, the background load is the same regardless of system size.
Should the load be scaled with the system?
( say, one copy of each load program per cpu? )
Also, our friendly PLM now automagically inhales the -ac series
(i think...i see 2.4.20-pre5-ac[1-6] plus a few more )
cliffw
On Wed, Sep 18, 2002 at 03:18:06PM -0700, Cliff White wrote:
>
> Also, our friendly PLM now automagically inhales the -ac series
> (i think...i see 2.4.20-pre5-ac[1-6] plus a few more )
>
Not yet, those are manual submissions for testing purposes.
-Nathan