Hi,
I have encountered a bug with Rosetta 4.77 running on Linux with BOINC 4.43. Sometimes when the scheduling process of BOINC gives control to the Rosetta application, the CPU gets idle, but Rosetta is visible via ps. I also had two computation errors which i very rarely had with other projects.
Greetings, Benno

---

I too have been having problems and only on my Linux machines. On my quad server was only using approximately 75% of its processing capability. Checking it this morning, It's down to about 30%. Even after changing the the nice level to -20.(overkill I know) I can't even get one out of four processors to run at 100%. I do plan on putting Microsoft on and see if that makes a difference. but I think the problem is more with BOINC than with Rosetta. Even on other projects, I could not get it to run at 100%.

---

I think it is with the Rosetta linux client. I am running 3 or 4 distributed projects on four different linux boxes and they all run at 99%. One has dual AMD mp 2600's and both cpu's run at 99%. Currently I am only running Rosetta on Windows XP64 with an AMD 3000 cpu.

---

well, I just installed server 2003 started Rosetta and all four processors went to 100%. I will run Windows on this project as long as I can. But I will need to go back to Linux. I see there were asking for assistance in their code, I wish I knew more C++ and what was going on behind Linux to help with their code. As I plan on staying with this project as long as I can.

---

I haven't seen this sort of problem (although more detail about your troubles would be illuminating). I ran R@H with 64 threads on a 32 node Linux cluster w/o error and with sustained load averages of 1.0 per thread just last night....We've run the app on a wide variety of Linux platforms, could you provide kernel and distro information please?

---

I am running a IBM Netfinity 5500. The only change I have made is the OS.

The systems are:

ID: 899 SuSE 10.0 beta (2.6.13-9-smp)

Measured floating point speed 284.71 million ops/sec
Measured integer speed 498.57 million ops/sec
--------------------------------------------------------

ID: 4461 Enterprise Server 2003 (no updates)

Measured floating point speed 493.17 million ops/sec
Measured integer speed 784.71 million ops/sec
--------------------------------------------------------

Yes, If you like Microsoft. This is excellent
But it is almost opposite to all the other benchmarks that I have seen. (well maybe not that good)

---

I haven't seen this sort of problem (although more detail about your troubles would be illuminating). I ran R@H with 64 threads on a 32 node Linux cluster w/o error and with sustained load averages of 1.0 per thread just last night....We've run the app on a wide variety of Linux platforms, could you provide kernel and distro information please?

BOINC 4.43
Distribution: Debian testing
Kernel: 2.6.11.10

The only thing I saw was that CPU load dropped to 0 when BOINC scheduled the time from other applications to Rosetta, but three Rosetta processes remained in memory. It worked again after restarting BOINC

---

BOINC 4.43
Distribution: Debian testing
Kernel: 2.6.11.10

The only thing I saw was that CPU load dropped to 0 when BOINC scheduled the time from other applications to Rosetta, but three Rosetta processes remained in memory. It worked again after restarting BOINC

Thanks for the information. When David Kim gets back from holiday I'll see that we look into this.

---

Any updates on this ?

---

Vist http://usalug.org
https://boinc.bakerlab.org/rosetta/team_display.php?teamid=593