Re: Suggestion for Pentium Error Stress Testing

Rogier Wolff (R.E.Wolff@BitWizard.nl)
Fri, 14 Nov 1997 13:05:22 +0100 (MET)

Billy Harvey wrote:
>
> Perhaps Richard or someone could whip up a program similar to the
> bovine effort, that linux folks could extract a keyspace from and run
> a sequence of bits through their system, to see if it will crash. The
> code would have to write a (flushed and synced) log before and after
> each execution effort so that we would have a trace to see which (if
> any, but I'm sure there are probably more) bit patterns cause
> problems, and on which processors.

To speed this up a LOT you should log just one in 1000 tries. Once you
find one that hangs your machine you just need to repeat the last 1000
tries. Either you test "random" numbers, and save the random seed at
the checkpoint, or you use something predictable (like "opcode++").

My machine does around 9000 forks per second, so you should be able to
do around 5000 opcodes per second. If you want to scan 4G of opcodes
you need 1M seconds, or about two weeks.

Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** +31-15-2137555 ** http://www.BitWizard.nl/ **
Florida -- A 39 year old construction worker woke up this morning when a
109-car freight train drove over him. According to the police the man was 
drunk. The man himself claims he slipped while walking the dog. 080897