Re: -mmX 4G patches feedback [numbers: how much performance impact]

[Eric Whiting]

Ingo Molnar wrote:
> 
> [***] non-PAE 4:4 kernels are being used too - there are a fair number
>       of users who run simulation code using 4GB of physical RAM and a
>       pure 4:4 kernel with no highmem features required. For these 4:4
>       users the overhead on number-crunching is even smaller, only
>       0.03%.

Ingo, 

What you describe above is a situation that matches our environment. 

We do have servers with more than 4G RAM that need PAE, but we have more
workstations with 4G (or 3G) of RAM that do not need PAE, but do need the 4:4
VM. In terms of a large simulation (> 3G of userspace VM) running to completion
-- a 3G RAM box with 4:4 VM is better than a 32G box with 3:1 VM -- even though
the box with 3G RAM may swap a little/lot.

4:4 is not the 'real' long term solution, but it will let us run more jobs on
32bit hardware until the x86_64 HW/SW support stabilizes a little more.  We
currently run jobs on solaris when single process VM requirement exceeds 4G. 

eric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/