max memory limits ???

hi

can some1 explain the memory limits on the 2.4 kernel

- what is the maximum amount of physical memory it supports ?
- what is the limit for user-space apps ?
- what is the limit for kernel ?
- are there any limits on malloc, static-memory, stack-memory ?
- does using HIGHMEM results with performance penalty ?
- anything else ?

please CC me for any answer

regards
Benhanokh Gabriel

-----------------------------------------------------------------------------
"If you think C++ is not overly complicated, just what is a protected abstract virtual base class with a pure virtual private destructor, and when was the last time you needed one?"

        -- Tom Cargil, C++ Journal, Fall 1990. --

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

There are some Alpha systems with 256 GB memory where Linux is
sometimes tested. Those processors support up to several
terabytes of memory (nobody built such maximum machines, though.)

For intel machines limits are different, of course. (PAE36 supports up to 16*4 GB = 64 GB physical address space,
within which there must be some holes to support e.g. PCI bus address spaces, and system boot rom, to name a few.)

For intel ia64 -- that is a 64 bit machine, thus it propably follows Alpha and UltraSPARC model in this regard.

At 32 bit systems: 3.5 GB with extreme tricks, 3 GB for more usual. At 64 bit systems -- 2^60 or some such semi-meaningless value.

Whatever the hardware supports.

userspace rules.

Of course, it causes extra mapping operations on machines needing it to support larger memory (intel PAE36 featured hardware). User processes can access all what is mapped into them at the same time -- All programs, kernel included are limited to 32 bit addresses, but kernel can juggle maps to reach areas not mapped in its address space at some moment.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Hi,

The other performance implications are:

HIGHMEM currently results in extra copies if you are on Intel and using a 4G or 64G config, whenever you perform disk IO to memory above 1GB. If you are using the 64G config (PAE36), then the page tables double in size, occupying more of your cache. Those are likely to be the two biggest performance costs in enabling highmem in an i686 kernel.

Cheers,
Stephen
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

What are those tricks? Do they involve changing TASK_SIZE ?

And why do programs get mapped at 0x08000000 instead of 0x00001000 ? This is wasting another 5% of the addressing space.

Please CC me on your reply.

Thanks,
-Raymond
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/