Hi everybody, I see a very common problem to all the owners of a AMD64 processor: the boinc client tries to download a 64 bit app, but it's not available.

Yes, I know all the anonymous plataform crap, but, why cannot just download the 32 bit version? it's compatible with the amd64 processors. It should be a fallback option if the 64 bit client isn't available. Or just an option in the general config file where you put the platform you wish to use.

This thing is driving me crazy. I'm trying to download the 32 bit client for myself, but it's not available in the webpage. What kind of obscure hacking am I supposed to do? It should be easier. Please, take account on this.

---

Are you running a 64bit OS and 64bit Boinc manager? (I'm running 32 bit WinXP and a standard 32bit Boinc Manager here, and it downloads the 32 bit Rosetta apps for my Athlon 64 system.)

---

I have at least two AMD64 computers (my computers are open to public viewing)

Both use WinXp Pro for the OS and neither gave me a speck of trouble downloading the Rosetta client.

I do not have any 64 bit applications.

Please take a look at my computers and tell me if any of them are even close to the system that is giving you trouble.

---

Rosie, Rosie, she's our gal,
If she can't do it, no one shall!

Are you running a 64bit OS and 64bit Boinc manager? (I'm running 32 bit WinXP and a standard 32bit Boinc Manager here, and it downloads the 32 bit Rosetta apps for my Athlon 64 system.)

Yes, I'm running a 64 bit linux and 64 bit boinc client. I think it's problem of the boinc client, it should download the best app from each projet. If there is no 64 bit app it should fallback to 32 bit in amd64 systems.

Where can I submit a bug to the boinc program?

---

I think at the moment you have to use the anonymous crap. BOINC does not support Windows64 yet and although a 32bit application would run fin in Windows x64 it refuses to download such a platform. I think the Win64 support needs to be first introduced by BOINC and not Rosetta. Rosetta can only specify which app is for which platform. Perhaps you post this over on the BOINC boards.

---

This thing is driving me crazy. I'm trying to download the 32 bit client for myself, but it's not available in the webpage.

The 32 bit Linux rosetta binary should be in the directory: https://boinc.bakerlab.org/rosetta/download/
The current version is the file named: rosetta_5.16_i686-pc-linux-gnu

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

---

miceliux
Given there are only 4 official BOINC clients, Windows (32bit), Linux (32bit) and MacOS(PPC & Intel) then any problem you have compiling it yourself and the use there after is afaik officialy down to you.

You could ask in the BOINC forumms or post to the development lists (there are various ones see the boinc website) or just build the client to report itself as the 32bit version, though maybe not ideal for other projects.

Though Augustine has the possible Rosetta serverside workaround for now, of course add the Windows/Vista 64 platforms as well ;-)
Though it'll not help you with any other projects you want to join along the way.


EDIT: didn't see the post dates of the original posts before the last one.... <grr>

---

Team mauisun.org

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too.

---

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too.

HashClash now supports AMD64 with a 32-bit Linux application...

---

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too.

HashClash now supports AMD64 with a 32-bit Linux application...

HashClash now supports AMD64 on Windows with a 32-bit application...

---

Augustine,

so what ?

This is rosetta@home....

---

Team mauisun.org

FWIW, running two instances of the client, one the 32-bit client, the other, the 64-bit client, on the same 4-core system, but limiting each client to 2 cores, I can compare the relative performance of 32-bit and 64-bit SIMAP's HMMER: the 64-bit version is about 7% faster.

By enabling vectorization (supported by default on AMD64), the SIMAP developers observed other 8% improvement.

Bottom line: porting the project application to AMD64 has the potential to improve performance by 15%!

---

I liked the developer's comments about how they were still looking into corruption issues with Windows on AMD cpus. Were they using an Intel only math library?

---

I liked the developer's comments about how they were still looking into corruption issues with Windows on AMD cpus. Were they using an Intel only math library?

Unless such an Intel only library makes use of some instruction not supported by AMD, I doubt that this would be the reason (of course, all recent Intel libraries are blocked from being used by non-Intel processors anyways, as opposed to for example ACML from AMD).

I didn't see anything on this issue in the thread linked below.

Can I also point out that a 15% improvement may not be the case for Rosetta [as discussed in several other threads re 64-bit Rosetta]. The fact that the current source-code + gcc 4.x -m64 doesn't generate correct results [or in some cases even a binary that runs correctly] for any optimisation above -O1 doesn't quite encourage work on 64-bit until the compiler works better... Of what I've seen of the source-code, it doesn't lend itself to 64-bit performance improvements. In the performance intensive code-path, almost none of the code is calling small functions (which is often where 64-bit improves because of register passed arguments), and almost all of the performance limits are FPU-related.

Being able to use -ftree-vectorize MAY help, but a lot of calculations aren't trivially vectorized [in one of the most time-consuming functions, it's because the inner loop is very short...]

--
Mats

---

I liked the developer's comments about how they were still looking into corruption issues with Windows on AMD cpus. Were they using an Intel only math library?

Unless such an Intel only library makes use of some instruction not supported by AMD, I doubt that this would be the reason (of course, all recent Intel libraries are blocked from being used by non-Intel processors anyways, as opposed to for example ACML from AMD).

I didn't see anything on this issue in the thread linked below.

Can I also point out that a 15% improvement may not be the case for Rosetta [as discussed in several other threads re 64-bit Rosetta]. The fact that the current source-code + gcc 4.x -m64 doesn't generate correct results [or in some cases even a binary that runs correctly] for any optimisation above -O1 doesn't quite encourage work on 64-bit until the compiler works better... Of what I've seen of the source-code, it doesn't lend itself to 64-bit performance improvements. In the performance intensive code-path, almost none of the code is calling small functions (which is often where 64-bit improves because of register passed arguments), and almost all of the performance limits are FPU-related.

Being able to use -ftree-vectorize MAY help, but a lot of calculations aren't trivially vectorized [in one of the most time-consuming functions, it's because the inner loop is very short...]

--
Mats

Mats, I think the Intel compilers have had that feature (I mean bug ;-) fixed now.

---

Team mauisun.org

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too.

HashClash now supports AMD64 with a 32-bit Linux application...

HashClash now supports AMD64 on Windows with a 32-bit application...

Guess what? Leiden Classical now supports AMD64 on Linux with a 32-bit application. ;-)

---

Augustine the AMD64 awareness BOINC spammer ;-)

---

Team mauisun.org

So far, two projects have added support for x86-64: SETI sends the x86 application and SIMAP, a true x86-64 application. Would you consider supporting the platform x86_64-pc-linux-gnu or x86_64-unknown-linux-gnu similarly?

Chess960 now supports AMD64 with a native Linux application too.

HashClash now supports AMD64 with a 32-bit Linux application...

HashClash now supports AMD64 on Windows with a 32-bit application...

Guess what? Leiden Classical now supports AMD64 on Linux with a 32-bit application. ;-)

Yay! Malaria now supports AMD64 on Linux with a 32-bit application too.

---

All projects do if you use dumas777's Linux x86_64 compiled client
http://boese.dnsalias.com:6969/

RE: https://boinc.bakerlab.org/rosetta/forum_thread.php?id=2549

---

Team mauisun.org