> There are however a few tricks to get such a program successfully
> running. The only solution I see for now is to copy all the code and data
> to video memory and execute from there. This type of program must be
> somehow like memtest86: stand-alone booting from floppy (no OS would
> survive since it will end corrupting memory for sure).
On P6 CPUs (maybe others too?) you can 'freeze' the on-CPU cache, thus you
can loose physical memory and you can still run code.
this way you can print out the results on the screen before dying. Or did
you want to run in video memory and save results in the CMOS? ;)