I've used YellowDog in the past, and I think it is a good dist. I wouldn't mind seeing a PPC Linux version of Chimera, either.
Geoff.
On Dec 7, 2005, at 10:50 AM, A. Zelter wrote:
I think Conrad is saying that on the x86 architecture they can compile for RedHat and it will run on most distributions of x86 Linux. He's asking if there's a similar situation on PPC Linux, and which distribution of PPC Linux would provide the best prospects for such coverage. It seems to me that this is to minimize the occurrence of someone saying "I see you have the binaries for XYZ PPC Linux, but they don't run on my distribution of ABC PPC Linux. Could you please compile it for ABC PPC Linux?"
Ahhh thanks. I knew I didn't get something. Well, I'm no expert on that kind of thing, but Fedora Core 4 has a ppc version. I think this is a little more modern than Red Hat Linux 7.1 - it runs kernel 2.6.11. However, RedHat did make a PPC version of their 7.1 distro (Red Hat Linux 7.1 for pSeries). I suppose that would be equivalent. Another good bet might be yellowdog linux. The latest version is 4.0.1, and it uses kernel 2.6.10. Yellowdog is based on redhat but built specifically for Macs. This is probably what I would try using. Alex
Here's an an overview of possible new platforms for Chimera distributions.
Currently we distribute Chimera binaries for:
Windows (x86, 32-bit app) Mac OS X (PPC, requires X windows) Linux (x86, 32-bit) SGI Irix Alpha Tru64
Additional platforms that we do not currently distribute for are:
Mac OS X Aqua (PPC, no X windows) Mac OS X, Intel Linux 64-bit kernel Windows 64-bit Linux, PPC
There are two obstacles to supporting a new platform. We want to have the target system in our lab for building and testing. This is the case for all our current distributions. And we have to put in work to build the following 28 third-party packages that Chimera uses:
TclTk Togl Tix Pmw PyOpenGL zlib Python Numeric jpeg tiff freetype Imaging netcdf Scientific MMTK HappyDoc tr openssl ftgl omni msms otf autostereo swish-e FFmpeg al2co expat qhull
And we will update versions of these third party packages and will have to get them to work on all distributed platforms in the future.
Currently it is not possible for outsiders to compile Chimera. One reason is that one package, msms molecular surface calculation, cannot be distributed. But the real practical obstacle is that it would take some debugging to successfully compile all the third-party packages on a new platform. Chimera has makefiles to build all the third-party packages but a few of them are likely to fail to build on any new platform and some building expertise will be needed to resolve the problems.
Here are a few comments on the possible new platforms.
We have built Mac OS X Aqua versions of Chimera. Because of many problems with the Tk window toolkit we did not distribute it. That was 2 years ago and many of the problems may be fixed.
We have built 64-bit Linux versions of Chimera. The 32-bit linux Chimera will run on those machines. The reason for building a 64-bit version is to handle very large data sets (> 4 Gbytes). The 64-bit pointers are needed to have an address space bigger than 4 Gbytes. The large data sets are primarily density maps.
The Mac OS X Intel platform will become important in under a year and we will have to support it.
The Linux PPC platform seems likely to become less common with Apple no longer producing PPC machines in the near future. I'm not sure there are enough potential Linux PPC Chimera users to make supporting this platform worthwhile.
In summary, I think Mac Aqua and Mac Intel are the next 2 platforms we are likely to support.
Tom
There are two obstacles to supporting a new platform. We want to have the target system in our lab for building and testing. This is the case for all our current distributions. And we have to put in work to build the following 28 third-party packages that Chimera uses:
Hi Tom, Thanks for the information. That sounds fair enough - ppc architecture will be less important soon and even now only a small proportion of linux users use that architecture. I have one more thought about support for different platforms though. Could you have a binary just for the msms molecular surface calculation package, and then the other packages as source plus makefiles. That would probably be much less work for you, and then whoever wanted to spend the time trying to get it to work could do, but you wouldn't have to be involved or support it in any way? You might even find others identify and fix problems on these new platforms - saving work for the Chimera team. Just a thought. Thanks for your consideration. Alex
Hi Alex,
We have a problem where the msms molecular surface code crashes and takes down Chimera with it, so moving it out of Chimera as a separate binary might be helpful for that problem too. We'll keep in mind the desirability of making Chimera compilable by outsiders while we figure out how to fix the molecular surface problems.
Thanks for the suggestion.
Tom
participants (3)
-
A. Zelter -
Geoffrey H Wathen -
Thomas Goddard