C'mon! There's gotta be somebody with access to the rosetta code who has the opportunity to compile it on a Linux/PPC distribution.

The Rosetta code isn't open source, so it's not available for "somebody", unless the same "somebody" is part of the Rosetta project.

There is a version of Rosetta for MacOS X/PPC, but not for Linux PPC.

I suspect the reason for this is the fact that the number of reported processors in Boincstats that fall under "PowerPC" are 111 - I may not have been looking in the right place, but IBM PowerPC doesn't exist. There are a number of "PowerMac" listed, but I presume these are all running MacOS of some sort...

And it's not just a case of compiling it, that "somebody" would also have to make sure it calculates correctly, including potentially corner-cases [i.e. those cases where you didn't get the "expected" range of results].

--
Mats

---

The Rosetta code isn't open source, so it's not available for "somebody", unless the same "somebody" is part of the Rosetta project.

That's what I mean.

And it's not just a case of compiling it, that "somebody" would also have to make sure it calculates correctly, including potentially corner-cases [i.e. those cases where you didn't get the "expected" range of results].

Well, I'm simply comparing to SETI@Home here. Since their client is Open Source, I suspect that they have solved the problem of computing correct results (at least from the point of view of platform-specific numerical peculiarities) for a number of platforms far greater than Rosetta@Home. Perhaps this "somebody" could glean some insight from SETI's experience.

Also, is it not the case that incorrect results can be identified as such ? If so, those results suffering from such corner cases could simply be discarded.

Furthermore, as long as the code is compiled with gcc for all platforms, there is a good chance that platform-specific floating point treatment will be taken care of.

Conclusion: The task of porting Rosetta@Home to Linux/PPC may not be as daunting as it first appears and, IMHO, it is worth some small effort. If, after such effort the resulting client turns out to produce incorrect results, I suppose we'll never have Rosetta@Home on Linux/PPC :o(

Either way, I lament the fact that now, since SETI@Home isn't distributing WUs, my Linux/PPC processor is sitting idly by.

---

The Rosetta code isn't open source, so it's not available for "somebody", unless the same "somebody" is part of the Rosetta project.

That's what I mean.

And it's not just a case of compiling it, that "somebody" would also have to make sure it calculates correctly, including potentially corner-cases [i.e. those cases where you didn't get the "expected" range of results].

Well, I'm simply comparing to SETI@Home here. Since their client is Open Source, I suspect that they have solved the problem of computing correct results (at least from the point of view of platform-specific numerical peculiarities) for a number of platforms far greater than Rosetta@Home. Perhaps this "somebody" could glean some insight from SETI's experience.

Also, is it not the case that incorrect results can be identified as such ? If so, those results suffering from such corner cases could simply be discarded.

Furthermore, as long as the code is compiled with gcc for all platforms, there is a good chance that platform-specific floating point treatment will be taken care of.

Conclusion: The task of porting Rosetta@Home to Linux/PPC may not be as daunting as it first appears and, IMHO, it is worth some small effort. If, after such effort the resulting client turns out to produce incorrect results, I suppose we'll never have Rosetta@Home on Linux/PPC :o(

Either way, I lament the fact that now, since SETI@Home isn't distributing WUs, my Linux/PPC processor is sitting idly by.

SETI solves the problem of "correctness" by calculating every result multiple times and comparing the results. This is a good way, except it takes a number of times more machines to solve the same problem (because every calculation is done multiple times). Rosetta relies on correct calculation, but does a particular WU only once.

I agree that GCC _SHOULD_ generate the same results on multiple platforms. But without a verification of the result (at least a trivial one), it wouldn't be proper to say that it does work.

I can't really speak for the project as such, as I don't know what the motivation is for NOT producing a Linux/PPC version. They do a MacOS/PPC version, so presumably no bugs in the PPC processor that causes problems... But I would think the "count" of PPC/Linux machines is one of the points here. There are MILLIONS of Windows/Linux x86 machines out on the market, and those will give millions of results every day, whilst the comparatively small number of Linux/PPC is "not worth the effort". That's just my speculation tho'.

--
Mats

---

I can provide access (ssh login) to the following platforms:

powerpc linux
hppa linux
ia64 linux
alpha linux
sparc linux

I can also help with porting to said platforms (I've been involved in porting SETI to hppa linux back in the days the code wasn't open)

HTH