Windows v2.3.0 CRITICAL UPGRADE

[Guilherme]

I think that SoB work units should be smaller. Maybe 300 or 400 blocks on each work unit. 1700 or 1900 blocks take too long to finish on my PC.

[vjs]

If the current tests are taking too long to finish on your PC you can always run double checks they are much smaller.

Simply enter the user name supersecret (for very small tests) or garbage for tests about half the size.

The amount of bandwidth required to do what you are suggesting would be enormous. Currently the project uses very little bandwidth about 100k per k/n test.

If they were to break the test appart into 5 pieces of 300 to 500 blocks the amount of bandwidth required would be well over 5mb per test. Inaddition to this the storage space on the server would also need to be trippled if not more since intermediate files would require storage, etc, etc, etc. Part of the problem with these tests is they must start at the begining 1 and continue to 1700 since the results from 1 to 1700 are cumulitive. You can't start a test at 1500 and continue to the end 1700, unless you have the files from 0-1500.

Also if there were any benifit and possibility to split test it would probably be done already.

Its just not possible, beside don't you plan on sticking with the project for more than a month or two anyways? It's an interesting project and here, there is actually something to learn.

[Frodo42]

OK I have these old Linux machines that appearently has problems with the new 2.3.0 client (I haven't been testing 2.2.0 on them).
It appears that these computers have a very old version of glibc ... the error I get when trying to start the client is:

./sb: /lib/libc.so.6: version `GLIBC_2.3' not found (required by ./sb)

I don't have any root access to these computers, so I'm not sure I can upgrade glibc (if anyone knows how I could do that I would be very pleased. The machines are running Red Hat 6.1 ... I'm used to Gentoo)
If the client could be made compatible to old versions of glibc that would be great, but if it means speed loss to the other clients then it's not worth it for these puny machines.
Maybe a special version of the 2.2.3 clent could be compiled on an old machine?

Anyways it's just 7 P3 400 Mhz machines ... not like it's going to mean the end of the world if they don't get updated to v.2.2.3.

[Death]

Originally posted by Frodo42
Anyways it's just 7 P3 400 Mhz machines ... not like it's going to mean the end of the world if they don't get updated to v.2.2.3. [/B]

like a 2.8 GHz pentium3 =))

[Mystwalker]

Originally posted by Frodo42
I don't have any root access to these computers, so I'm not sure I can upgrade glibc (if anyone knows how I could do that I would be very pleased. The machines are running Red Hat 6.1 ... I'm used to Gentoo)

I don't know about glibc, but usually (I'm not that deep into Linux, though), libaries can be put to e.g. usr/local - if it's possible to tell Linux resp. the SoB client to use this glibc instead of the general one.

[CanSpice]

Originally posted by Mystwalker
I don't know about glibc, but usually (I'm not that deep into Linux, though), libaries can be put to e.g. usr/local - if it's possible to tell Linux resp. the SoB client to use this glibc instead of the general one.

You need root access to put anything in /usr/local. However, it is possible to create a directory in your home space (say, ~/local/) and install things in there. You may need to set your LD_LIBRARY_PATH environment variable to ~/local/lib to allow the SoB client to find your local version of glibc though.

[Alien88]

I will be making a new tar/gz of the client later today, but for now http://www.seventeenorbust.com/download/sb.tar.gz - that's just the binary. that'll work, it's statically compiled. The last one was compiled dynamically instead of statically, hence the glib problems (that was a mistake on my part).

[Frodo42]

Thank you.
I won't be able to check the statically linked client until I get back to the computers in a week or so, but I'm sure it will work.

[redtech116]

I still don't think the new 2.3.0 code is using the SSE2 on Pentium M in my new Thinkpad .
Only getting about 500Kc/m's from a 1.5Ghz
the data sheet on the CPU says it has SSE2

[Alien88]

Originally posted by Frodo42
Thank you.
I won't be able to check the statically linked client until I get back to the computers in a week or so, but I'm sure it will work.

I've updated the download page. There is now also a .tar.gz that contains only the binary, so you do not have to worry about your config being over-written.

[Mystwalker]

Originally posted by redtech116
I still don't think the new 2.3.0 code is using the SSE2 on Pentium M in my new Thinkpad .
Only getting about 500Kc/m's from a 1.5Ghz
the data sheet on the CPU says it has SSE2

The Pentium M definitely has SSE2 - what does the client tell you at startup?

btw. (just to make sure):
Did you enter into a proper energy profile? With max battery settings, the CPU won't be at full speed...

[Ken_g6[TA]]

The Pentium M has SSE2, but performs worse on SB when using SSE2. here's why.

Can SSE2 be turned off for V2.3 on the Pentium M?

[redtech116]

Yes, I turn all power savings off even when not connected to ac power
the client....Intel(R) Pentium(R) M processor 1500MHz detected. Enabling cpu specific optimizations.

[tqft]

As promised report -
V2.2 report
[Sat Dec 18 16:55:05 2004] Intel Pentium II or Pentium II Xeon processor detected. Enabling cpu specific optimizations.
[Sat Dec 18 16:55:10 2004] restarting proth test from cache (k=19249, n=7213406) [100.0%]
[Sat Dec 18 17:29:16 2004] residue: 62ED1D6B3DF3228E
[Sat Dec 18 17:29:16 2004] completed proth test(k=19249, n=7213406): result 3


v2.3 report
[Sat Jan 01 21:56:06 2005] Intel Pentium II or Pentium II Xeon processor detected. Enabling cpu specific optimizations.
[Sat Jan 01 21:56:11 2005] restarting proth test from cache (k=19249, n=7213406) [100.0%]
[Sat Jan 01 21:58:54 2005] residue: D5DFAE5C0F6555A2
[Sat Jan 01 21:58:54 2005] completed proth test(k=19249, n=7213406): result 3

[Haranaka]

Are there still 2.2 clients around sending results? And are the results being blocked already?

[vaughan]

I have one v2.2 that is at 71.4% and also a v1.2.5 that is at 56.1%

[Keroberts1]

i don't think the 2.2 client will beallowedto submit it results I believe the server stopped accepting them. If at any time it has been restarted it will generate an incorect result.

[vaughan]

My v2.2.0 client finished its task uneventfully. Seemed to behave normally from my end, not sure if it was accepted or not.
The P166mmx is still chugging away its v1.2.5 and is at 58.5% It takes 2.25 hours per block. My P4 3.2 Prescott takes 2.25 minutes running v2.3.0!

[meckano]

If wcpuid program can check for sse2, why can't another programmer?
anywho.

I just installed, and config shows setting as idle, but runs as normal priority.hmmm
anywho.

Just here to give AMD a boost, saw they were down to 10th.
- oops, that's daily stats.

Wowa! I stopped it and lost all data?
- what is that log size thing?
- needing minimum setup info page!
- Guess i'll be writing another eh? lol

Yeah, give me info and I'll make a page for it.
- for starters, how to not lose info when hitting stop.

- and why did i start a new test? I expired original right away though.
-- yes, i had closed 17 program

Maybe you want to see some of my work before spending time on me, ok, here:
http://clubplus.net/~dbrown/

I read something in this post about z files, I have 2, each relating to the 'n' I am working on. I wiped the original I was working on.