[extropy-chat] visualizing nanoscale

Eugen Leitl eugen at leitl.org
Mon Nov 24 14:34:20 UTC 2003


On Mon, Nov 24, 2003 at 03:05:06PM +0100, Max M wrote:
> Eugen Leitl wrote:
> 
> >You only need 32 bit/atom, a few flags included, with current fat
> >desktops and next-generation small machines you can easily represent
> >up 10^9 atom systems, and render those in realtime, though. That's
> >at the threshold to mesoscale already.
> 
> 
> Where do you get the 32 bit/atom from?

It's a ballpark figure. An array of structs, each
being 8int (atom), 8int (x), 8int (y), 8int(z). The
coordinates are relative to (cubic) voxel edge.
The voxel size is depending on whether you include
H-H molecules, then you make diagonal <74 pm, with
C-H it's <109 pm and with C-C <154 pm (the exact
length depends on system temperature).

Resolution is about 0.25-1 pm, which is sufficient for
MD. 8 bits integer is the smallest size with direct
support in the CPU instruction set. 

If we just limit ourselves to CHNOPS, you still have
5 bits left in the 8int of the atom field. That can
be used for flags (transparence, etc). and stuff.
 
> There are many ways to make a format like that. I am just curious what 
> you imagine.
> 
> Whether it being entirly dumb, or you want it to know about the chemical 
> structure.

Of course you can make it to store bonds and other info, too.
As bonds don't reach beyond direct neighbour-voxels, and you've
only got 26 direct neighbours in cubic-primitive lattice and
valence is limited in number and state you can still stay comfortably
below 64 bit. 

3D BitBlt on 64 bit is expensive, but less so than doing
lots of FLOPs to just do some translation. That addresses dragging,
you still have to plow through entire volumetric dataset to visualize,
and ditto to do forcefield/MD calculations.

You can do ~1 fps with a 4 GByte Athon 64, I presume. Your limit
is memory bandwidth, so it would scale nicely to some 100 nodes
to do ~50 fps processing on 10^9 atom systems.

If you add more nodes/and or make nodes with better memory (like Blue Gene),
you can scale that up by several orders of magnitude while still
remaining realtime.

-- Eugen* Leitl <a href="http://leitl.org">leitl</a>
______________________________________________________________
ICBM: 48.07078, 11.61144            http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
http://moleculardevices.org         http://nanomachines.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20031124/bc334691/attachment.bin>


More information about the extropy-chat mailing list