Re: kernel compile time comparison (2.0 vs 2.1 with 64MB)

Peter T. Breuer (ptb@it.uc3m.es)
Mon, 3 Aug 1998 12:05:54 +0200 (MET DST)


"A month of sundays ago Andrew Derrick Balsa wrote:"
>
> I am repeating myself, but getting correct benchmarking numbers means
> one must first identify the performance bottleneck, and then devise or
> use the available tools to measure the actual bottleneck. Otherwise we
> begin mixing variables, and the numbers loose any meaning.
>
> Perhaps I should have been more explicit in my previous post, so:
> *provided one has enough memory to avoid swapping*, kernel compilation
> is basically CPU-bound (integer and cache/memory bandwidth).

Interesting. I just set up a Celeron-based system for a friend. By
coincidence, it has exactly the same configuration (apart from cpu and
motherboard, and type of memory) as one of my servers .. a PP200.

Running the same kernel (2.0.25) on both gives numbers of 195s at kernel
compilation for the Celeron running at 412MHz. The PP200 does 254s.
The Celeron ran at 93% cpu. The PP at 94%.

I don't understand those numbers, thinking about them - what was the
Celeron doing to occupy it fully at twice the clock rate for 4/5 of the
time on the same task? Memory i/o must be being included in the cpu
figures, otherwise that's impossible.

Both machines with 128M ram (PC100 for the Celeron). Adaptec 2940 fast but
not wide (not-integrated) controllers to IBM disks. The motherboard for
the Celeron is a Chaintech BX (6BTM). The PP has a 3 year old Microstar
dual board with one slot occupied. I forget the model number. I can
try it as a dual.

On the same kernel compilation, my own P188mmx (75MHz bus, TX board)
with ide takes about 310s, also at full cpu usage. My P133 (T1 board)
takes 510s. My P188mmx is a tad faster than a classic P200.

I almost feel like varying the parameters one at a time :-).

Kernel compilation seems cpu and/or memory bound, and the measurements
seem to indicate that the measurements that I make don't distinguish the
two.

> In benchmarking, the devil is in the details. :-)

Well, three years of hardware development ought to make a difference
that is noticable with your eyes shut. I have to say, no, not really.
I still run a 486 dx 33MHz quite happily as an X terminal, with 32M
memory. I can't say it's slower at anything except kernel compilation.
It benchmarks about the same at that as a couple of old sparc's I have.
They have wonderful disk i/o. But putting ram in the 486 levels it out.

> Cheers,
> --
> Andrew
>

Peter

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html