On 13-May-2000 Rajagopal Ananthanarayanan wrote:
> Kanoj Sarcar wrote:
>>
>> The mips64 SMP kernel completed a mips kernel compile on a 8 cpu
>> o2000, with approx 128 Mb memory.
>>
>> Here are some timings:
>>
>> make -j8:
>> 46.34user 32.34system 6:25.91elapsed 20%CPU (0avgtext+0avgdata
>> 0maxresident)k
>> 0inputs+0outputs (497100major+1814481minor)pagefaults 0swaps
>>
>> make -j4
>> 12.56user 16.34system 6:37.93elapsed 7%CPU (0avgtext+0avgdata
>> 0maxresident)k
>> 0inputs+0outputs (497100major+1807168minor)pagefaults 0swaps
>>
>> make -j32
>> 44.00user 29.25system 6:35.93elapsed 18%CPU (0avgtext+0avgdata
>> 0maxresident)k
>> 0inputs+0outputs (497174major+1826803minor)pagefaults 0swaps
>
>
> Awesome! And the resultant kernel boots ok? ;-)
>
> I wonder if we can get base-line stats on a 4-CPU X86 system,
> say for -j4 & -j8 and compare that with the O2K system.
> What version of the kernel are you using, 2.3.99preX?
First, we need to agree on what exactly we will be compiling. I have
numbers from 4P x86 -j4 but for a different kernel configuration. I must
say though that the CPU utilizations above are extremely low for parallel
make. For 4P I get utilizations above 350%. On the other hand, the number
of page faults is comparable (given the x86 box has more memory) so I'm
curious what the CPUs on the o2k are doing.
--
Dimitris Michailidis dimitris@xxxxxxxxxxxx
|