[Gllug] machine load testing

[Nix]

On Mon, 2 Dec 2002, David Damerell uttered the following:
> On Monday, 2 Dec 2002, Adam Bower wrote:
>>I just installed a new machine today which is a hand me down from 
>>another part of the company which has been reported to "perhaps have a 
>>hardware fault" so what I would like to do is really stress test the 
>>machine RAM+CPU+disk all at the same time.
> 
> Kernel compiles are usually good.

The problem with kernel compiles is that only big problems (i.e. major
memory problems that cause GCC to hare off down a garbaged pointer) come
to light.

I've found GCC bootstrap-and-check runs to be good; run it once on a
known-good box with (nearly) the same hardware config so you have a
known-good set of test failures, then do repeated `make bootstrap; make
check' runs until it falls over.

The advantage of this is that the comparison step at the end of `make
bootstrap' spots really rather small failures in the CPU (at least in
its non-floating-point parts), and the `make check' tests if the code
built on this possibly-broken CPU actually works.

-- 
`I keep hearing about SF writers dying, but I never hear about SF
 writers being born.  So I guess eventually there'll be none left.'
                                    -- Keith F. Lynch


-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug