On Tue, 20 Oct 2015 16:46:23 -0700, Ethan A Merritt <[log in to unmask]> wrote:
>On Tuesday, 20 October, 2015 22:35:35 Kay Diederichs wrote:
>> On Tue, 20 Oct 2015 12:19:13 -0700, Ethan A Merritt <[log in to unmask]> wrote:
>> ...
>> >my primary lab machines are still running a 32-bit user enviroment.
>> >I think it is too soon to drop 32-bit support.
>> >
>>
>> Hi Ethan,
>>
>> the workaround is installing a 64bit (guest) virtual machine in your 32bit primary lab machine. That allows you to run any 64bit software.
>
>I'm pretty sure that anyone still running a 32bit linux installation is
>doing so because of limited memory. Running a virtualization layer would
>make the problem worse, not better.
XDS needs little memory to run. The amount of memory it needs is (t+k+x)*i + c where i is the size of one frame (in pixels, each stored in 4 bytes), t is the number of threads (which could be 1) and k is a small number (2 or so) representing storage of auxiliary arrays. x is the number of frames in the frame cache, which is _the_ new feature of the latest version, and it may be set to 0. c is a constant on the order of 50 MB. That means, a 1GB virtual machine suffices for XDS (if t=1 and x=0).
>
>> Actually I'm surprised that your primary lab machine is 32bit. What does "primary" mean; what is it used for? How much RAM does it have? I often have Phenix and Phaser jobs with higher memory requirements than 4G. What do you do in such cases? Are there really no 64bit machines that you can use?
>
>Sure, go ahead and laugh. Frugal soul that I am,
>I'm typing this on a 7-year-old machine with nice graphics but
>limited memory that still serves 90% of my desktop computational
>needs just fine. Yeah, of course we have newer + bigger machines
>running 64bit. But so far I have not been sufficiently motivated
>to buy a replacement for my desktop and move all my stuff onto it.
>That and the fact that I do have a few programs that I use regularly
>which don't seem to work on a 64bit installation (xxdiff,
>crossover/wine+MSWord2K).
Come on, this whole discussion was about use of a machine for XDS. You make it sound like discussing of 32bit for use in general. You can use your 32bit machines for what you want and as long as you want - but for an increasing number of purposes it may not be the best choice, and you concede that you have been using your 64bit machines for computing anyway. In so far I don't see how your situation is relevant.
What I really would like to know is the number of people who have no 64bit Linux with easy access, and for whom it constitutes hardship if the 32bit version is not provided. I have created a Doodle poll at http://doodle.com/poll/ibm9enu55p693y5w for those people - please use the comment part at the bottom.
>
>I know I'm a dinosaur. I have neither a smart phone nor a Facebook page.
>My point was that if someone like me is still getting along fine
>with a 32-bit machine, surely there are other people out there who are
>doing likewise and who have fewer options than I do for transferring
>operations to a newer machine. To me it seems several years too soon
>to drop support for these older machines, unless of course the support
>cost is prohibitive. Is there more to it than running a parallel build
>with a different set of compiler flags?
yes. It is a) testing and b) support. It boils down to the question: do we (WK and I) want to spend our time on providing the 32bit package, for a few users who cannot be bothered to install or log into a 64bit machine, or do we use it for development of better algorithms? Or is there a real need for the 32bit version?
If you think about it, the development of better algorithms, resulting in better data, may be more effective for the community and the field. In so far it may be in your own interest that the 32bit version is phased out.
best,
Kay
>
>
> cheers,
>
> Ethan
>
>
>> Does it really constitute hardship to people like you to transition to 64bit, or to have at least one 64bit machine? I cannot believe this. Our students here all have notebooks with 4G RAM and more. They mostly run 64bit Windows.
>>
>> best,
>>
>> Kay
>--
>Ethan A Merritt
>Biomolecular Structure Center, K-428 Health Sciences Bldg
>MS 357742, University of Washington, Seattle 98195-7742
|