Re: [Lhms-devel] Re: 150 nonlinear

["Mika Penttilä" <mika.penttila@kolumbus.fi>]

Dave Hansen wrote:

>On Tue, 2004-10-26 at 14:20, Mika Penttila wrote:
>  
>
>>"There are two problems that are being solved: having a sparse layout
>>requiring splitting up mem_map (solved by discontigmem and your
>>nonlinear), and supporting non-linear phys to virt relationships (Dave
>>M's implentation which does the mem_map split as well)."
>>
>>
>>so what's the split?
>>    
>>
>
>So, mem_map is normally laid out so that, if you have 1GB of memory, the
>memory for 0x00000000 is at mem_map[0], and the memory for the last page
>(at 1GB - 1 page) is at mem_map[1<<30 / PAGE_SIZE - 1].  
>
>That's fine and dandy for most systems.  But, imagine that you have some
>memory on a funky machine where you have 2GB of memory, but it is laid
>out like this:
>
>    0-1 GB - first 1 GB
>  1-100 GB - empty
>100-101 GB - second 1 GB
>
>Then, you'd need to have mem_map sized the same as a 101GB system on
>your dinky 2GB system (disregard the ia64 implementation).
>
>The split I'm referring to is cutting mem_map[] up into pieces for each
>contiguous section of memory.  
>
>Make sense?
>
>-- Dave
>
>
>  
>
Yes, I see Dave M's approarch is doing this, but isn't Andy's as well? 
What's the key differences between these two?

--Mika


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>