Message boards : Number crunching : If Boinc ever DID make a 64-bit version...
Message board moderation
Author | Message |
---|---|
Send message Joined: 21 Mar 05 Posts: 33 Credit: 63,483 RAC: 0 |
How much work would it be to optimize the program currently running on Boinc? I mean, do we have programmers still helping us with this project, or are we totally out of resources in that department? If we do have some, what are they working on right now? I\'m not exactly unusual with my AMD64 processor. They are quite cheap, which means that everyone buying a computer from about 3 months ago onward, will be moving to a 64-bit processor. I think there has to be a growing demand for a 64-bit client. Matthew |
Send message Joined: 5 Aug 04 Posts: 1496 Credit: 95,522,203 RAC: 0 |
Matthew, A 64-bit client was tested in Sulfur Alpha and it is significantly faster than the 32-bit version. There were issues with it, though. Given all the fires Tolu and Neill have to fight, it's anyone's guess where it is on the priority list. HTH. ___________________________________________ |
Send message Joined: 21 Mar 05 Posts: 33 Credit: 63,483 RAC: 0 |
You can see how cheap AMD64's are, even right now... http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_609,00.html You can get a discounted motherboard with AMD64 2800+ for about $200 or $250 - not a huge outlay of cash for most people. That just gives you some idea of how soon it will be before most people are sitting there with 64-bit CPUs, just waiting for a 64-bit OS (and Boinc). Matthew |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
My opinion is that they will have to do it. With the longer, more computationally intensive experiments upcoming, they are going to have to take advantage of the huge speed increase seen with 64 bits (~30% in sulphur alpha under Linux). The question, I believe, is when. With another IT person joining Tolu and Neill, it become more likely earlier that it otherwise would have been, but when is earlier? Time will tell. |
Send message Joined: 23 Aug 04 Posts: 49 Credit: 183,611 RAC: 0 |
> My opinion is that they will have to do it. With the longer, more > computationally intensive experiments upcoming, they are going to have to take > advantage of the huge speed increase seen with 64 bits (~30% in sulphur alpha > under Linux). The question, I believe, is when. With another IT person > joining Tolu and Neill, it become more likely earlier that it otherwise would > have been, but when is earlier? Time will tell. The model was originally coded for 64-bit machines, so it shouldn't be too big a deal to get it running on them. It will be some time off, though. It's on our "nice to do someday" list rather than our "fundamental to the experiment" list. Top of that list is (1) getting our data portal sorted out; (2) getting the coupled model (hindcast and forecast) prepared. [(2) depends on (1).] Both of these entail a lot of labor, unfortunately. Dave |
Send message Joined: 21 Mar 05 Posts: 33 Credit: 63,483 RAC: 0 |
Is there a real shortage of programming talent? What I'm thinking is, if someone (like me) was anxious for 64-bit AND had the ability to program, perhaps that is the solution. If someone could jump-start me on how to get a development environment set up for Win64 (and how to compile Boinc sourcecode) I might try it myself. I found the page on "how to compile Boinc" and it looks like there are several libraries I have to install and compile. That is the kind of complexity I have to work through. According to Microsoft, anything compiled for .NET will work automatically at the right bit-level (32 or 64) because the runtime libraries are all that will need to be changed. Other things will require a recompile. Matthew |
Send message Joined: 23 Aug 04 Posts: 49 Credit: 183,611 RAC: 0 |
> Is there a real shortage of programming talent? What I'm thinking is, if > someone (like me) was anxious for 64-bit AND had the ability to program, > perhaps that is the solution. > > If someone could jump-start me on how to get a development environment set up > for Win64 (and how to compile Boinc sourcecode) I might try it myself. > > I found the page on "how to compile Boinc" and it looks like there are several > libraries I have to install and compile. That is the kind of complexity I have > to work through. Ahhh... I was talking about the model itself. Unfortunately the Met Office's model isn't open source (because it has considerable commercial value). Many climate models are open source, and we have tried (unsuccessfully, so far) to get these involved in the experiment. Each is something like half a million lines of code, written over a couple of decades, so they tend to be quite unwieldy things (and not easily optimised). Dave |
©2024 cpdn.org