[Chimera-users] v1.2105 for amd64 (x86_64)

Greg Couch gregc at cgl.ucsf.edu
Mon Apr 18 14:54:19 PDT 2005


On Mon, 18 Apr 2005, Sabuj Pattanayek wrote:

> I don't know if this has been asked before (couldn't find anything on
> searches) but will there be a native amd64/emt64 (x86_64) build of
> chimera v1.2105?

You didn't say if you wanted a Linux or Windows64 version.

The short answer is no.  Partially because we don't have any of those 
systems in our lab, but also because we haven't yet encountered a scenario 
where it is useful.  There two reasons for 64-bit executables: the ability 
to address more memory to handle larger problems, and to use new 
instructions that speed up floating point computations.  The address space 
issue is lurking, but doesn't appear to be a problem today (please correct 
me if I'm wrong) because chimera really wants all of its data to fit in 
RAM and few systems have more than 4GB of memory in them.  The floating 
point speed issue is more compelling, but chimera doesn't do much floating 
point except in a few cases (volume and surface calculations) and just 
switching to Pentium4 instructions instead of Pentium, might speed things 
up on more systems -- we're planning to do that experiment later this year 
as part of a larger optimization effort.

So to summerize, when we get 64-bit Opteron/Xeon system, it will be 
possible to create 64-bit executables because we already have 64-bit 
executables for Alpha processors.  But even on 64-bit architectures, it is 
often better to run a 32-bit executable because the same size problem uses 
less memory (32-bit pointers are half the size of 64-bit pointers), in 
fact that is exactly what we do for SGI IRIX systems.

(There is one Linux 64-bit binary of an old version of chimera on our 
download web page that was made because older versions the 64-bit NVidia 
graphics driver didn't provide a 32-bit interface.  My understanding is 
that has been fixed, so it is no longer necessary.)

 	Hope this helps,

 	Greg Couch
 	UCSF Computer Graphics Lab
 	gregc at cgl.ucsf.edu


More information about the Chimera-users mailing list