Detailed client software thread... - Page 5

Xyzzy · 2003-03-26, 04:56

http://www.mersenneforum.org/mirror/

Note that I can't promise it will be up to date, but it is better than not being able to get anything at all...

BigRed · 2003-03-26, 18:46

Quote:

What is the max that anyone has tried by the way, e.g. anyone tried 1.5gb or more for testing 33m exponent in p-1?

Stage 2 P-1 of a regular DC uses 890MB of RAM on a machine with 1 1/8 GB. It's running with E=12.

I got an E=4 P-1 on a 33m exponent on a box with !GB RAM.

db597 · 2003-04-29, 17:03

Any news about 23.3? I've not run my torture test on 23.2 yet.... keeping my fingers crossed for 23.3.

Prime95 · 2003-04-30, 03:45

There are 2 changes in this release - both with important caveats.

1) The SSE2 FFTs were changed to use a scratch area. This slightly slowed FFTs 512K and below and I think one of the P4 Celeron FFTs above 512K is slower. All other P4 FFTs should be faster. I'll post my P4 benchmark in a separate post. I don't expect any problems with this change, but let me know if you run into reproducible torture test errors.

2) P-1 stage 2 was rewritten. If you are in the middle of a P-1 run, you should complete it before trying v23.3. Also, do not install v23.3 on a dual boot machine until a v23.3 mprime is also available. This P-1 code needs more thorough testing. If you decide to use this version, stay tuned to this forum in case a bug is found that requires upgrading.

Download at:

Windows: ftp://mersenne.org/gimps/p95v233.zip

Let me know if you find any problems - especially Daran, a.k.a. Mr. P-1.

Enjoy!

Darak · 2003-04-30, 05:27

So from now on can we expect a 11 hour turn around from the point we request a new version to it being released? ;)

Keep up the good work!

Shaneamy

JC · 2003-05-01, 09:36

Let's try..

George, Any news about 23.4 ? :D

Quote:

Originally Posted by Darak

So from now on can we expect a 11 hour turn around from the point we request a new version to it being released? ;)

db597 · 2003-05-01, 16:41

Cheers for that! I'm running Linux, so I'll wait for the mprime version.

Prime95 · 2003-05-01, 21:43

Version 23.3 again. I fixed a bug where stage 2 ran even if the daytime memory setting was very low. Also, the Linux port is available.

Windows: ftp://mersenne.org/gimps/p95v233.zip
Linux: ftp://mersenne.org/gimps/mprime233.tar.gz
ftp://mersenne.org/gimps/sprime233.tar.gz

crash893 · 2003-05-01, 22:29

would it be in my intrest to change it if my day and night settings are the same?

markhl · 2003-05-02, 00:13

I just ran a benchmark for 23.3.1.

As George says, it is faster than 22.1.2: 69 vs. 75 ms for a 1024K FFT, which is the size being used for current exponents near 18M.
A lot faster for larger exponents, slightly slower for <= 448K.

It's also a lot easier to run the benchmark because it now does fewer iterations for bigger FFTs, so the benchmark takes about 40 s on my P4 1.8 GHz. See www.mersenne.org/bench.htm .

Mark

Prime95 · 2003-05-02, 01:26

Quote:

Originally Posted by crash893

would it be in my intrest to change it if my day and night settings are the same?

You do not need the newer 23.3.

2003-04-30, 03:45	#48
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 1D6D₁₆ Posts	version 23.3 There are 2 changes in this release - both with important caveats. 1) The SSE2 FFTs were changed to use a scratch area. This slightly slowed FFTs 512K and below and I think one of the P4 Celeron FFTs above 512K is slower. All other P4 FFTs should be faster. I'll post my P4 benchmark in a separate post. I don't expect any problems with this change, but let me know if you run into reproducible torture test errors. 2) P-1 stage 2 was rewritten. If you are in the middle of a P-1 run, you should complete it before trying v23.3. Also, do not install v23.3 on a dual boot machine until a v23.3 mprime is also available. This P-1 code needs more thorough testing. If you decide to use this version, stay tuned to this forum in case a bug is found that requires upgrading. Download at: Windows: ftp://mersenne.org/gimps/p95v233.zip Let me know if you find any problems - especially Daran, a.k.a. Mr. P-1. Enjoy!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Program for detailed analysis of ECM-NFS crossovers	Dubslow	Factoring	14	2016-03-18 06:01
Detailed Stats 0 to 100 Gauges	tapion64	PrimeNet	2	2014-04-12 08:26
More Detailed Exponent/Assignment Status	KingKurly	PrimeNet	8	2011-01-06 20:51
linux 25.9 client, unsafe thread output?	xorbe	Software	0	2009-04-03 04:21
"Latest client software" missing v25.6	OneOfMany	Software	1	2008-09-18 16:48

2003-03-26, 04:56	#45
Xyzzy "Mike" Aug 2002 2×23×179 Posts	http://www.mersenneforum.org/mirror/ Note that I can't promise it will be up to date, but it is better than not being able to get anything at all...

2003-04-29, 17:03	#47
db597 Jan 2003 7·29 Posts	Any news about 23.3? I've not run my torture test on 23.2 yet.... keeping my fingers crossed for 23.3.

2003-04-30, 05:27	#49
Darak Oct 2002 8₁₀ Posts	So from now on can we expect a 11 hour turn around from the point we request a new version to it being released? ;) Keep up the good work! Shaneamy

2003-05-01, 16:41	#51
db597 Jan 2003 CB₁₆ Posts	Cheers for that! I'm running Linux, so I'll wait for the mprime version.

2003-05-01, 21:43	#52
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 1110101101101₂ Posts	Version 23.3 again. I fixed a bug where stage 2 ran even if the daytime memory setting was very low. Also, the Linux port is available. Windows: ftp://mersenne.org/gimps/p95v233.zip Linux: ftp://mersenne.org/gimps/mprime233.tar.gz ftp://mersenne.org/gimps/sprime233.tar.gz

2003-05-01, 22:29	#53
crash893 Sep 2002 2³·37 Posts	would it be in my intrest to change it if my day and night settings are the same?

2003-05-02, 00:13	#54
markhl Apr 2003 California 134₈ Posts	I just ran a benchmark for 23.3.1. As George says, it is faster than 22.1.2: 69 vs. 75 ms for a 1024K FFT, which is the size being used for current exponents near 18M. A lot faster for larger exponents, slightly slower for <= 448K. It's also a lot easier to run the benchmark because it now does fewer iterations for bigger FFTs, so the benchmark takes about 40 s on my P4 1.8 GHz. See www.mersenne.org/bench.htm . Mark