[net.unix-wizards] memory access benchmark

cnrdean@ucbtopaz.BERKELEY.EDU (10/10/85)

I am looking for a benchmark program which will test for the 
maximum main memory size that a program can have in a computer.
That is, maybe a computer manufacturer says that a program can be 
64K of data and 64K of instructions.  But, I would like to test to 
make sure that this is true.  

My guess is that to test the data size, I would simply keep 'alloc'ing
until a failure happens.  I would like, not only to declare a
large array, but to test that values are accurately stored.

But, I am not sure how to write a program which keeps incrementing
its instruction size.  Is this possible?

Any help would be appreciated.  If you can send some source, so much 
the better.

Sam Scalise
UC Berkeley

sdo@u1100a.UUCP (Scott Orshan) (10/15/85)

In article <89@ucbjade.BERKELEY.EDU> cnrdean@ucbtopaz.BERKELEY.EDU () writes:
>
>My guess is that to test the data size, I would simply keep 'alloc'ing
>until a failure happens.  I would like, not only to declare a
>large array, but to test that values are accurately stored.

This wouldn't work on the Sperry 1100 versions of the UNIX system.
The stack and data share a memory area.  There is a user defined
address at which the stack starts growing downward and the heap
starts growing upward.  This address is an option to the loader
and it allows stack to be traded for heap.

0400000  <----- Stack Growth ----| Data |------ sbrk area ---> 0777777
				 ^
				 |
			This can be adjusted
-- 

			Scott Orshan
			Bell Communications Research
			201-981-3064
			{ihnp4,bellcore,pyuxww}!u1100a!sdo