| Author |
Message |
|---|
|
|
|
Hi all,
I have just posted new prpclient 5.3.1 packages to http://uwin.mine.nu/PRPNet/. This includes a fix for the crashes when processing WWWW tasks, and also includes Genefer 3.2.1. This version gives a significant speed boost over the previous, and also adds support for FMA3, so even faster if you have a Haswell chip.
As always, post any problems or questions here, or drop me a PM.
Thanks to those who provided binaries for various different platforms.
Cheers
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
Honza Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1893
ID: 352
Credit: 3,142,312,174
RAC: 0
                             
|
|
Fired up 5.3.1 on WSS, seems OK on Windows.
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 |
|
|
|
|
Fired up 5.3.1 on WSS, seems OK on Windows.
Thanks Honza, good to know.
Cheers
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
                                   
|
|
5.3.1 working well in both 64-bit and 32-bit Windows for me.
GFN65536 port working well in 64-bit and 32-bit, no speed-up noticed
MEGA port working fine in 64-bit
GFN524288 working in 64-bit using both CUDA and OCL apps. No speed up on CUDA. OCL app running 20%-25% faster on GTX 6xx cards.
Nice work Iain!
|
|
|
|
|
GFN65536 port working well in 64-bit and 32-bit, no speed-up noticed
GFN524288 working in 64-bit using both CUDA and OCL apps. No speed up on CUDA. OCL app running 20%-25% faster on GTX 6xx cards.
Nice work Iain!
This is pretty much as expected. You should observe speedups for all Genefer CPU tasks except those which use the x87 / 80-bit transform (i.e. GFN 65536 & 32768). All of the CPU code now uses a faster algorithm based on a 'Z-transform' instead of the traditional Discrete Weighted Transform. The x87 assembler code was not rewritten as this point as it is too much effort and there are better alternatives (in the future). The OpenCL code is significantly faster as it now also uses the Z-Transform, as well as some additional optimisation, but CUDA is unchanged.
Cheers
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 644
ID: 164101
Credit: 305,010,093
RAC: 0
|
|
I'm starting some tests on ATI (Radeon R9 270).
Running on platform 'AMD Accelerated Parallel Processing', device 'Pitcairn', version 'OpenCL 1.2 AMD-APP (1348.5)' and driver '1348.5 (VM)'.
geneferocl 3.2.0-0 (Windows/OpenCL/32-bit)
658332^524288+1 Time: 1.36 ms/mul. Err: 0.2266 3050541 digits
538452^1048576+1 Time: 2.74 ms/mul. Err: 0.2188 6009544 digits
440400^2097152+1 Time: 5.96 ms/mul. Err: 0.2188 11836006 digits
360204^4194304+1 Time: 11.9 ms/mul. Err: 0.1953 23305854 digits
294612^8388608+1 Time: 24.9 ms/mul. Err: 0.2070 45879398 digits
Genefer Mark = 35.
geneferocl 3.2.1-0 (Windows/OpenCL/32-bit)
658332^524288+1 Time: 1 ms/mul. Err: 0.1875 3050541 digits
538452^1048576+1 Time: 2.09 ms/mul. Err: 0.1953 6009544 digits
440400^2097152+1 Time: 4.13 ms/mul. Err: 0.1719 11836006 digits
360204^4194304+1 Time: 8.45 ms/mul. Err: 0.1719 23305854 digits
294612^8388608+1 Time: 18.1 ms/mul. Err: 0.1641 45879398 digits
Genefer Mark = 48.
It is 25%-35% faster on this card.
Yves |
|
|
|
|
|
Is the ftp site down? or is it just me? |
|
|
|
|
Is the ftp site down? or is it just me?
NOT just you!!
TheDawgz can get there either ... Posted this over in the external chat room
____________
There's someone in our head but it's not us. |
|
|
|
|
|
uwin.mine.nu is online again.
Lennart
|
|
|
 RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1137
ID: 120786
Credit: 267,535,355
RAC: 0
                    
|
|
I installed PRPNet 5.3.1. Notice default server set to ESP in master_prpclient.ini. Default should now be set to another sub-project.
I have new Windows 7 OS installed. When I tried to run with only server GFN262144 uncommented I got error:
"wwwwcl.exe...VCOMP110.DLL is missing..."
I commented out wwwwexe=wwwwcl.exe, (which shouldn't be needed for GFN262144 anyway), and its now working.
I have AMD X6 1100T.
Supported genefer 3.2.1-0 transform implementations: default x87 sse2
In 3.2.0 previously for me sse3 was also supported.
I seem to have downloaded WUs from nearer the trailing edge.
Lets see how I go. |
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 644
ID: 164101
Credit: 305,010,093
RAC: 0
|
Supported genefer 3.2.1-0 transform implementations: default x87 sse2
In 3.2.0 previously for me sse3 was also supported.
The specific implementations of the new transform are sse2, sse4, avx and fma. Because the new sse2 is faster than the previous sse3, the sse3 was removed.
With the new design (data ordering), sse3 would not be faster than sse2.
Yves |
|
|
|
|
I installed PRPNet 5.3.1. Notice default server set to ESP in master_prpclient.ini. Default should now be set to another sub-project.
Thanks Roger, I'll make that change in the next release packages.
Cheers
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
|
|
|
I see in the .ini file that www is said to be..
"Available ONLY to 64bit OS"
- but there are both wwwcl.exe and wwwcl64.exe programs available.
Does this mean I can run www OCL tasks under 32-bit OS as well?
Perhaps we should add a specific '32' or '64' to the program names to clarify this..?
____________
 |
|
|
|
|
|
I've been crunching WU in the challenge and came home to this today
[2014-06-14 23:23:51 MS] GFN262144: PRPNet server is version 5.3.0
Hi! Welcome to PrimeGrid's GFN 262144 Prime Search.
geneferocl 3.2.1-0 (Windows/OpenCL/32-bit)
Copyright 2001-2014, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Iain Bethune, Michael Goetz, Ronald Schneider
Command line: geneferocl.exe work_GFN262144.in
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 770', version 'OpenCL 1.1
CUDA' and driver '337.88'.
Start test of file 'work_GFN262144.in' - 23:23:51
Testing 1080936^262144+1...
Using OCL transform
Starting initialization...
Initialization complete (1.581 seconds).
Estimated time remaining for 1080936^262144+1 is 0:33:26
Testing 1080936^262144+1... 2359296 steps to go (0:13:57 remaining)
maxErr exceeded for 1080936^262144+1, 0.4531 > 0.4500
maxErr exceeded by all available transform implementations
PRP: 1080936^262144+1 955000/5254374
First i've thought it may be a transient error so i restarted the client and it began to calculate the same WU from the start, when i returned to my computer some time later the same error showed up again
geneferocl 3.2.1-0 (Windows/OpenCL/32-bit)
Copyright 2001-2014, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Iain Bethune, Michael Goetz, Ronald Schneider
Command line: geneferocl.exe work_GFN262144.in
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 770', version 'OpenCL 1.1
CUDA' and driver '337.88'.
Start test of file 'work_GFN262144.in' - 09:05:16
Testing 1080936^262144+1...
Using OCL transform
Starting initialization...
Initialization complete (1.506 seconds).
Estimated time remaining for 1080936^262144+1 is 0:35:22
Testing 1080936^262144+1... 2359296 steps to go (0:13:58 remaining)
maxErr exceeded for 1080936^262144+1, 0.4531 > 0.4500
maxErr exceeded by all available transform implementations
Is there a bug within the genefer app, or is this algorithm not capable of doing this test?
And second question, how can i throw away that i can continue to crunch WU for the challange? (I've completed over 50WU succesfully prior to this, and my GPU ran stable on GNF short task for the last 2 weeks, so i guess it's not the cause of a very unstable GPU) |
|
|
|
|
|
I'm getting the same thing on ALL my GPUs...
Command line: geneferocl.exe -d 1 work_GFN262144.in
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 750 Ti', version 'OpenCL
1.1 CUDA' and driver '335.23'.
Start test of file 'work_GFN262144.in' - 08:37:38
Testing 1084450^262144+1...
Using OCL transform
Starting initialization...
Initialization complete (2.547 seconds).
Estimated time remaining for 1084450^262144+1 is 1:35:43
Testing 1084450^262144+1... 2883584 steps to go (0:50:25 remaining)
maxErr exceeded for 1084450^262144+1, 0.4688 > 0.4500
maxErr exceeded by all available transform implementations
PRP: 1084450^262144+1 162500/5255602
I think these Genefer tasks have grown too big, and 'hit the wall' as far as GPUs are concerned - and crunching from now on will be CPUs only..!?!
(not sure what this additional PRP: thing its doing..? - but it looks like it'll take a very long time! - so I trashed it...!)
btw: if u want to 'trash' a task, just set:
startoption=1
or
stopoption=2
or
stopasapoption=2
in that prpclient.ini file - then close (<Ctrl-C>) and restart that prpclient.exe to abandon task and get more...
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
                 
|
PRP: 1084450^262144+1 162500/5255602
(not sure what this additional PRP: thing its doing..? - but it looks like it'll take a very long time! - so I trashed it...!)
If you have PFGW enabled in PRPNet, it will try that after Genefer fails. It has a sequence of programs it will use, from fastest to slowest.
That looks like the output from PFGW.
____________
My lucky number is 75898524288+1 |
|
|
|
|
If you have PFGW enabled in PRPNet, it will try that after Genefer fails. It has a sequence of programs it will use, from fastest to slowest.
That looks like the output from PFGW.
Indeed a pfgw64 process runs on my computer.
If it's true that GPU can't do any more tasks i'll have to stop participating. CPU is reserved for other projects at the moment. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
If it's true that GPU can't do any more tasks i'll have to stop participating. CPU is reserved for other projects at the moment.
That's not entirely true.
The b limit isn't a hard limit. When you get near it, some will fail, while others will work. What this means is that the GPU app is going to fail on some of the numbers, but will succeed on others.
As b gets higher, a larger percentage of the tests will fail.
PRPNet should automatically switch to the CPU version of Genefer, if you have that enabled. It will, of course, be slower. Or you could just let the tasks fail and let your GPU move on to the next one.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Having a curious problem on one of my clients with the new version:
Running 4 Tasks, 2 on CPU (all is fine) and 2 on GPU with geneferopencl. One GPU running on gpuaffinity=0 in prpclient.ini, the other on gpuaffinity=1. But one of the gpu tasks is problematic with the new prpclient, the other have no problems:
014-06-15 10:56:42 MS] Total Time: 0:49:24 Total Work Units: 1 Special Results Found: 0
[2014-06-15 10:56:42 MS] GFN262144: Returning work to server prpnet.primegrid.com at port 11002
[2014-06-15 10:56:43 MS] GFN262144: INFO: Test for 1078146^262144+1 was accepted
[2014-06-15 10:56:43 MS] GFN262144: INFO: All 1 test results were accepted
[2014-06-15 10:56:44 MS] GFN262144: Getting work from server prpnet.primegrid.com at port 11002
[2014-06-15 10:56:45 MS] GFN262144: PRPNet server is version 5.3.0
[2014-06-15 10:59:34 MS] GFN262144: Could not open file [lresults.txt] for reading. Assuming user stopped with ^C
[2014-06-15 10:59:34 MS] Total Time: 0:52:16 Total Work Units: 1 Special Results Found: 0
[2014-06-15 10:59:34 MS] Client shutdown complete
But I did not have a lresults.txt, also not in the other gpu directory where all is running fine.
Also at this time, i got an entry in memleak.txt:
Dump the changes that occurred between two memory checkpoints
-858993460 bytes in -858993460 Free Blocks.
-858993460 bytes in -858993460 Normal Blocks.
-858998593 bytes in -858993503 CRT Blocks.
-858993460 bytes in -858993460 Ignore Blocks.
-858993460 bytes in -858993460 Client Blocks.
Largest number used: -858998593 bytes.
Total allocations: -859002510 bytes.
Detected memory leaks!
Dumping objects ->
{562} normal block at 0x0069A160, 8 bytes long.
Data: < h > B4 CC 68 00 00 00 00 00
{211} normal block at 0x00699FD8, 8 bytes long.
Data: <, i > 2C E7 69 00 00 00 00 00
{210} normal block at 0x0069A080, 8 bytes long.
Data: <0Uj > 30 55 6A 00 00 00 00 00
Worker.cpp(19) : {208} normal block at 0x006A5528, 36 bytes long.
Data: < @h &h i GFN2> 10 40 68 00 E8 26 68 00 80 A0 69 00 47 46 4E 32
{207} normal block at 0x0069A010, 8 bytes long.
Data: < i > FC E6 69 00 00 00 00 00
ServerHandler.cpp(269) : {204} normal block at 0x0069E6E8, 100 bytes long.
Data: < CJ @h #j &h > 0C 43 4A 00 10 40 68 00 80 23 6A 00 E8 26 68 00
Object dump complete.
Any ideas?
Regards Odi
____________
 |
|
|
|
|
|
I seem to have the same issue as Odicin
Have captured this output in the console window
Start test of file 'work_GFN262144.in' - 17:16:20
Testing 1087346^262144+1...
Using OCL transform
Starting initialization...
Initialization complete (1.827 seconds).
Estimated time remaining for 1087346^262144+1 is 0:34:35
Testing 1087346^262144+1... 4259840 steps to go (0:26:16 remaining)
maxErr exceeded for 1087346^262144+1, 0.4688 > 0.4500
maxErr exceeded by all available transform implementations
Base prime factor(s) taken : 3217
not support
[2014-06-15 17:23:55 MS] GFN262144: Could not open file [lresults.txt] for readi
ng. Assuming user stopped with ^C
[2014-06-15 17:23:55 MS] Total Time: 0:07:35 Total Work Units: 0 Special Resu
lts Found: 0
[2014-06-15 17:23:55 MS] Client shutdown complete
The lines :
Base prime factor(s) taken : 3217
not support
seemed to come from llrCUDA.exe |
|
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1137
ID: 120786
Credit: 267,535,355
RAC: 0
|
|
The prpclient.exe in Windows 5.3.1 version appears to be a 64 bit app. I tried to install on Windows XP 32 bit OS and it complained prpclient.exe wasn't 32 bit. Same with 5.3.0. Older version 5.0.8 worked fine although the server requires 5.3.0 or greater client.
If 32 bit OS support has been dropped then the line in the PRPNet thread needs to be updated:
prpclient-5.3.1-windows (32 bit prpclient and 32/64bit apps) |
|
|
|
|
The prpclient.exe in Windows 5.3.1 version appears to be a 64 bit app. I tried to install on Windows XP 32 bit OS and it complained prpclient.exe wasn't 32 bit. Same with 5.3.0. Older version 5.0.8 worked fine although the server requires 5.3.0 or greater client.
If 32 bit OS support has been dropped then the line in the PRPNet thread needs to be updated:
prpclient-5.3.1-windows (32 bit prpclient and 32/64bit apps)
It could be 32bit but if with visual studio compiled, the xp platform was not included. |
|
|
|
|
|
I'm noticing an issue with my client. I'm currently running 100% Wieferich with the following setting:
server=WIEFERICH:100:25:prpnet.primegrid.com:13000
To the best of my knowledge this should mean I get 25 work units, process 25 workunits and then return 25 workunits. However, what happens is the server (thinks it) sends 25, I get 24, I process 24 and I return 24. The 25th just disappears into thin air. You can check pending work and see that I have WU's pending with an interval of about 30 minutes.
I'm running a fully patched Win7 installation.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
 |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1893
ID: 352
Credit: 3,142,312,174
RAC: 0
|
|
Charley, I used to have similar issue with 50 work units cache - half of those limbo WUs were in oblivion until expiration. Lowering to 25 solved the issue.
Funny part was that it was not happending all the time. Sometimes running PRPClient showed the issue, sometimes not.
What happens if you restrart your client?
And if you change cache size?
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 |
|
|
|
|
|
Restarting doesn't make a blind bit of difference. I'll change cache size and report back in a bit.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
|
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1893
ID: 352
Credit: 3,142,312,174
RAC: 0
|
|
On a side note, PRPClient being funny.
And for the record, client was terminated using Crtl+C.
[2014-09-09 12:18:20 SE(e] User name Honza at email address is ?@volny.cz
It appears that the PRPNet client was aborted without completing
the workunits asigned by server WIEFERICH. What do you want to do with them?
...
4 = Complete in-progress, abort the rest, report them, then get more work
...
Choose option: 4
Huh? The test did not complete, yet you didn't terminate it.
[2014-09-09 12:18:31 SE(e] Total Time: 0:00:11 Total Work
Units: 0 Special Results Found: 0
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 |
|
|
|
|
|
Workaround (not a "fix") is to just grab fewer WUs. I'm doing 10 at a time and it seems fine. Imperfect, I know. But I think I remember this issue being reported a while ago regarding requesting exactly the max allowed number of units at a time... one unit always went missing.
--Gary |
|
|
|
|
I'm noticing an issue with my client. I'm currently running 100% Wieferich with the following setting:
server=WIEFERICH:100:25:prpnet.primegrid.com:13000
To the best of my knowledge this should mean I get 25 work units, process 25 workunits and then return 25 workunits. However, what happens is the server (thinks it) sends 25, I get 24, I process 24 and I return 24. The 25th just disappears into thin air. You can check pending work and see that I have WU's pending with an interval of about 30 minutes.
I'm running a fully patched Win7 installation.
Yes, it's something that's been around a bit. I run 20 with no issues.
See here: http://www.primegrid.com/forum_thread.php?id=5157&nowrap=true#67531 |
|
|
|
|
|
As it seems a lower amount solves the issue. The server and my PC agree that my latest batch is 15 WU's in size.
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)
|
|
|
|
|
|
When prpclient 5.3.1 logs a prime from genefer with "Residue: PRP" in test_results.log, it reports the time as "0 seconds". This should be fixed. |
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 834
ID: 301928
Credit: 488,476,972
RAC: 0
            
|
|
Is it possible to recompile wwwcl with older version of Visual Studio? It does not run on XP anymore. I even can do it myself, is the source available? (I tried to get some PSA credits on a old system with GPU, but geneferocl falls back to CPU mode due to maxerr, which is not fun at all, and wwwcl does not run on XP).
The same question applies to PrpClient itself. Right now I'm running executable from 5.2.8 but I could miss something important.
|
|
|
|
|
|
I have just recieved a 32-bit XP compatible prpclient build. When it has been tested I'll post an updated client package and you can all have a go.
Cheers
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
Is it possible to recompile wwwcl with older version of Visual Studio? It does not run on XP anymore. I even can do it myself, is the source available? (I tried to get some PSA credits on a old system with GPU, but geneferocl falls back to CPU mode due to maxerr, which is not fun at all, and wwwcl does not run on XP).
The same question applies to PrpClient itself. Right now I'm running executable from 5.2.8 but I could miss something important.
It's a toolset switch setting -- the current VS is capable of creating XP compatible code, but it's not the default.
____________
My lucky number is 75898524288+1 |
|
|
|
|
Is it possible to recompile wwwcl with older version of Visual Studio? It does not run on XP anymore. I even can do it myself, is the source available? (I tried to get some PSA credits on a old system with GPU, but geneferocl falls back to CPU mode due to maxerr, which is not fun at all, and wwwcl does not run on XP).
The same question applies to PrpClient itself. Right now I'm running executable from 5.2.8 but I could miss something important.
It's a toolset switch setting -- the current VS is capable of creating XP compatible code, but it's not the default.
http://www.bc-team.org/downloads.php?cat=7
XP compatible
|
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 834
ID: 301928
Credit: 488,476,972
RAC: 0
|
http://www.bc-team.org/downloads.php?cat=7
XP compatible
No, they are not. I found this page early, but no luck. Both "prpnet 5.3.2 -update only -32bit -r31" and "wwwwcl.exe 2.2.5 -win -32bit" from this page have following executable header:
OS version 6.00
Subsystem version 6.00
It must be 5.00 to run on XP and 4.00 for W2K.
Waiting for a new build from Iain.
|
|
|
|
|
http://www.bc-team.org/downloads.php?cat=7
XP compatible
No, they are not. I found this page early, but no luck. Both "prpnet 5.3.2 -update only -32bit -r31" and "wwwwcl.exe 2.2.5 -win -32bit" from this page have following executable header:
OS version 6.00
Subsystem version 6.00
It must be 5.00 to run on XP and 4.00 for W2K.
Waiting for a new build from Iain.
Are you sure? It was compiled with xp toolset in visual studio 2012. You only need the visual c++ redistributable. |
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 834
ID: 301928
Credit: 488,476,972
RAC: 0
|
http://www.bc-team.org/downloads.php?cat=7
XP compatible
No, they are not. I found this page early, but no luck. Both "prpnet 5.3.2 -update only -32bit -r31" and "wwwwcl.exe 2.2.5 -win -32bit" from this page have following executable header:
OS version 6.00
Subsystem version 6.00
It must be 5.00 to run on XP and 4.00 for W2K.
Waiting for a new build from Iain.
Are you sure? It was compiled with xp toolset in visual studio 2012. You only need the visual c++ redistributable.
Absolutely sure. If you have OS version field set to 6, you'll get "This is not a Win32 executable" message on WinXP. The good news are that I've patched PE header, changed these fields to 5.01 and both programs run, unlike old versions! (Old versions really required some APIs which didn't exist in WinXP, so after patching a header I got "NewSuperDuperFunctionEx not found in KERNEL32.DLL" error). I run small self-test on wwwcl, seems to work, and will do some real workunits a bit later, when CPU-genefer finishes (ouch...)
So please check your project settings about OS version / Subsystem version fields. This information is written on linker stage, and as a last resort could be overridden by additional linker command-line options. To check header, use "dumpbin /headers myprog.exe" from MSVC command-line prompt.
BTW, prpclient.exe does not require msvc redistributable at all (it references no additional DLLs). It's required only for wwwcl which refers to VCOMP110.DLL. I wonder if static version of this library is available, so this reference could be eliminated too. |
|
|
|
|
|
I have posted a new build of prpclient here http://www.epcc.ed.ac.uk/~ibethune/files/prpclient.xex.zip At least it was built with the v110_xp toolset and according to the executable header it the OS version is 5.01.
Can you please download and try it out (unzip and change from .xex to .exe), then let me know if it works.
Cheers
- iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
I have posted a new build of prpclient here http://www.epcc.ed.ac.uk/~ibethune/files/prpclient.xex.zip At least it was built with the v110_xp toolset and according to the executable header it the OS version is 5.01.
Can you please download and try it out (unzip and change from .xex to .exe), then let me know if it works.
Cheers
- iain
I can verify that your prpclient.exe build runs on a 32 bit XP system.
EDIT: I mean that I *have* verified that it does indeed run on XP 32 bit Windows.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Thanks Mike, I'll see about getting the windows package updated.
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
|
|
let me know if it works.
I can verify that your prpclient.exe build runs on a 32 bit XP system.
EDIT: I mean that I *have* verified that it does indeed run on XP 32 bit Windows.
I being the one asking for this recently also have verified it is working on XP 32 bit. Thank you again for your time in this issue. PSA updated badge hunting here I come. |
|
|
|
|
|
Can someone put a link that has the 64 bit and 32 bit downloads for 5.3.1 as there are several listed on this thread but am not sure if I'm would be getting the correct ones. Sorry for my confusion.
Cheers Rick |
|
|
|
|
|
Hi Rick, for now unless you are running Windows XP on a 32-bit machine, use the official versions in the release packages. If you are, then use http://www.epcc.ed.ac.uk/~ibethune/files/prpclient.xex.zip
I will be updating the packages to 5.3.2 soon, with executables that should work on all platforms.
Cheers
- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime! |
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 834
ID: 301928
Credit: 488,476,972
RAC: 0
|
|
Thanks everybody for help, finally got wwwwcl running Wieferich on XP.
1) Used new prpclient.exe posted by Iain;
2) wwwwcl itself was taken from current PrpNet package, but version fields in the header were patched (file offsets 0x140 and 0x148, change 6 to 5). Really, it is already XP-compatible (does not references new Vista+ Win32 APIs) and runs just fine after this change. Only have to install MSVC 2012 Redistributable to get VCOMP11.DLL on my system.
3) Have to upgrade old NVIDIA drivers from 280.26 to latest 347.09, otherwise wwwwcl reported wrong sign on some near-hits (e.g. "22084519 is a special instance (+1 +23 p", it should be -23 according to output of CPU version).
4) It almost does not worked at all without tweaking of "thread" and "blocks" parameters. First workunit going to take almost 10 hours to complete with 0% of CPU and GPU usage... (still crunching :-). Have to benchmark it in different combinations, got minimum execution and 90% GPU load for Wieferich at -t 8 -b 2048 (with 100% CPU load on two cores, but nobody cares). With this setting, this old GT440 runs 330 seconds per WU. Not blazing fast, but good enough to get some PSA credits "in background" on old currently unused computer.
Also, when I compared output from CPU and GPU version on a small range (-p 3 -P 1e8), I noticed few differences in results:
1) GPU version could not detect cases when both "+xx" and "-xx" results exist, for example:
173 is a special instance (-1 +85 p)
173 is a special instance (-1 -88 p)
Only second result will be reported by wwwwcl.
2) wwwwcl does not detect number "7" at all, although CPU version says:
7 is a special instance (+1 +1 p)
7 is a special instance (+1 -6 p)
Are these differences known/acceptable?
|
|
|
|
|
Hi Rick, for now unless you are running Windows XP on a 32-bit machine, use the official versions in the release packages. If you are, then use http://www.epcc.ed.ac.uk/~ibethune/files/prpclient.xex.zip
I will be updating the packages to 5.3.2 soon, with executables that should work on all platforms.
Cheers
- Iain
Thanks - If you are releasing a new version soon I will wait unless it's after the WFS challenge and the download is the link in the very first post of this thread correct?
Thanks man |
|
|
|
|
Also, when I compared output from CPU and GPU version on a small range (-p 3 -P 1e8), I noticed few differences in results:
1) GPU version could not detect cases when both "+xx" and "-xx" results exist, for example:
173 is a special instance (-1 +85 p)
173 is a special instance (-1 -88 p)
Only second result will be reported by wwwwcl.
2) wwwwcl does not detect number "7" at all, although CPU version says:
7 is a special instance (+1 +1 p)
7 is a special instance (+1 -6 p)
Are these differences known/acceptable?
Do remember what 173 is a special instance (-1 +85 p) means. See this post for example. It means that:
2^(172/2) == -1 + 85*173 (mod 173^2)
and we denote by A the number 85. Since this is a congruence modulo 173^2, adding or subtracting a multiple of 173^2 = 173*173 on the right-hand side is permissable. So A could be any of the numbers:
A = ..., -434, -261, -88, 85, 258, 431, ...
Now, we are only interested in "small" A values, |A| <= 1000. But since your prime is so tiny, 173, we see that many of the A values satisfy this.
Personally I would think it would be preferrable if the value with the lowest absolute value was returned here. But this is really a silly "problem". For realistic numbers, such as the prime 338526053161351871, only if we are very lucky will the A be closer to 0 than 1000.
As soon as p >= 2003, at most one A "representative" can be numerically less than 1000. Did you see any discrepancies between CPU and GPU version for primes over 2000?
This post assumes you ran your tests with -s to set the threshold for "special" A to 1000.
/JeppeSN |
|
|
|
|
|
Dear Iain, please, can you add info about WinXP32 patch to the message in the thread http://www.primegrid.com/forum_thread.php?id=1215&nowrap=true#80880 ?
I have WinXP32, and I just had trouble during running a prpnetclient just after install.
There will be a prpnet challenge in January, so I think this info will be usefull. |
|
|
