> On Sat, 28 Mar 1998, Samuli Kaski wrote:
>
> > Insipired by compiler tests conducted here I did the same thing. The
> > used tree was 2.1.91.
> >
> > make dep; make clean; make zImage took on
> >
> > 2.0.34-pre4 10 min 36 sec
> > 2.1.91 9 min
>
> > Nice, not mindblasting though, improvement I would say.
>
> given that the compiler has not changed between those two compiles, and
> only 20-30% of the compilation time is 'spent' in the kernel, it is
> something like a 50% perceived kernel speed improvement ...
Are you sure of that?
I just tried the same test on a 2.0.29 kernel and got:
PII 233 + 64 MB + NCR UW SCSI (same controller used for previous results).
2 x Atlas II 2.2 GB.
212.75 user 17.57 system 4:03.92 elapsed 94% CPU.
Kernel time is about 8% of the elapsed time for my test.
Elapsed time improvement observed in the previous report is about 18%.
> ['spent' is the sum of "spent executing" and "spent waiting"]
And we probably spent time on benchmark results that were just garbage in,
unless,
- the kernel speed improvement is greater than infinite.
- 2.0.34-pre4 is a lot broken.
- there is much user time spent in the kernel.
- I missed something important.
Gerard.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu