[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: is there a linux-image-2.6.17-x-amd64-k8?

On Wed, Sep 13, 2006 at 04:06:57PM +0200, Francesco Pietra wrote:
> Specifically, I should add, we have detected a bug in mpqc2.31, unable - with 
> large complex molecules - to build correctly the hessian in internal 
> coordinates, reading cartesian coordinates. We turned to cartesian 
> coordinates throughout, which is less efficicient.
> Now I wonder whether instead of a bug in mpqc it could be a problem of the 
> kernel (just a naive doubt, possibly devoid of any scientific ground). I was 
> using 2.6.15.-1 k8 smp.
> No major difference to all purposes? In particular, using the system for 
> computation (mpqc2.3.1) without X, aimed at the highest possible floating 
> point.
> I am just reinstalling amd64 on dual opteron (hope the old net installer - 19 
> July - works; the newest does not, as I have reported. Therefore, I am 
> interested in replacing the generic kernel that has been just installed.

The kernel only deals with system calls, such as allocating memory and
printing, and accessing files.  The actual calculations, the kernel has
nothing to do with.  So no there really is no chance that the bug is in
the kernel instead in this case.

A k8 and an em64t kernel should have no real difference to running code.
Their instruction set is the same, the cache line size is different, but
that doesn't make any real difference.

Len Sorensen

Reply to: