Posts by Bubba Winkerbean


log in
1) Message boards : Politics : GOod GOodly MOogly.*** U.S. Embassy in Baghdad***Oh Heeba Habba GOoba Gabba (Message 1529189)
Posted 17 Jun 2014 by Bubba Winkerbean
Some countries on earth don't have rights.
They don't think it's worth any fights.
When Americans try
and some of us die,
we're still the country that bites?
2) Message boards : Politics : More on how Neo-Darwinism has it wrong again... (Message 1528793)
Posted 17 Jun 2014 by Bubba Winkerbean
To the Darwinist it's random chance,
belief in this is their firm stance.
Such complexity
from such entropy...
Some one had to start this dance.
3) Message boards : Number crunching : Step by Step, compile NVIDIA/MB CUDA app under Linux (Fedora 19) (Message 1526489)
Posted 10 Jun 2014 by Bubba Winkerbean
Right now I'm operating under the assumption that the algorithms written in the code were implemented in the most efficient manner, but that won't stop me from taking a peek at them later after I get some level of confidence about all these CFLAGS.

Tried some different Xbranch CFLAGS. First one was the most basic. Second one was Petri's list. Third one I looked through all those CFLAGS (and several other areas of the gcc manual) and picked the ones I thought might be helpful. There are a whole lot more available than I thought there would be. Third one was pretty much a guess on my part. I removed redundancy, didn't add things that didn't make sense according to what I read, and didn't add it explicitly if it said it was enabled by default. Fourth one was trying some that said might produce some incorrect output but will make it faster.

Xbranch configure #1: #./configure BOINCDIR=/home/guy/Desktop/boinc CFLAGS='-DUSE_CUDA'

Xbranch configure #2 (copied/pasted from Petri's list, BOINCDIR changed, LIBS deleted, kept redundancies):
./configure BOINCDIR=/home/guy/Desktop/boinc CFLAGS='-O3 -march=corei7-avx -Ofast -funroll-loops -mtune=corei7-avx -mpreferred-stack-boundary=8 -fexceptions -fno-rounding-math -fno-signaling-nans -frerun-cse-after-loop -fgcse -fgcse-after-reload -fgcse-lm -fgcse-las -fgcse-sm -faggressive-loop-optimizations -fexpensive-optimizations -fschedule-insns -fschedule-insns2 -fmove-loop-invariants -fselective-scheduling -fvect-cost-model -fsection-anchors -fsched-stalled-insns-dep -fsched-stalled-insns -fsched-spec-load -fsched-spec -fsched-pressure -freorder-functions -freorder-blocks -frename-registers -fregmove -fprefetch-loop-arrays -fpredictive-commoning -fpeephole -foptimize-register-move -fomit-frame-pointer -fcx-limited-range -fno-math-errno -fno-trapping-math --param inline-unit-growth=3000 -DPINNED -DNDEBUG -DHAVE_STRCASECMP -DSETI7 -DUSE_I386_OPTIMIZATIONS -DUSE_CUDA -fpeel-loops -funroll-loops -fweb'

**Lots of "warning: this target does not support ‘-fsection-anchors’ [-fsection-anchors]"
**And again, I get "Compiling entry function...sm_10, sm_11, sm_12, sm_13, sm_21, and sm_20", but nothing for sm_30 or sm_35

Xbranch configure #3: (I looked through https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html (and several other sections) and tried to follow what you guys were saying. note: -Ofast enables a bunch of optimizations. Tried this:
#./configure BOINCDIR=/home/guy/Desktop/boinc CFLAGS='-Ofast -march=corei7-avx -mtune=corei7-avx -funroll-loops -DUSE_CUDA'

Xbranch configure #4: (tried unsafe math optimization and finite math only flags. Manual said 'haswell' was valid entries for march/mtune, but compiler said it couldn't produce an executable from it. 'corei7-avx' was no where in the manual, but it compiles with them.)
#./configure BOINCDIR=/home/guy/Desktop/boinc CFLAGS='-Ofast -march=corei7-avx -mtune=corei7-avx -funroll-loops -funsafe-math-optimizations -ffinite-math-only -DUSE_CUDA'

The following are the times I got from each compiled app, using an I7 and a GTX780ti.


Configure#/runtime in seconds #1 #2 #3 #4 FG00091_V7.wu 3381 3343 3369 3378 FG00134_V7.wu 3215 3184 3199 3205 FG01307_V7.wu 1058 1014 1019 1019 FG02968_V7.wu 836 775 778 778 FG03853_V7.wu 702 646 651 649 FG04160_V7.wu 641 596 604 595 FG04221_V7.wu 643 591 594 588 FG04317_V7.wu 634 583 583 582 FG04465_V7.wu 617 567 573 577 FG09362_V7.wu 464 434 438 438 FG11753_V7.wu 362 349 350 349 FG13462_V7.wu 358 344 344 347 FG24857_V7.wu 354 342 340 340 FG53024_V7.wu 348 334 338 339 FG76516_V7.wu 346 335 336 337


So, not much difference really, but it's looking like Petri's list is what works the best.

Note on #4:

-funsafe-math-optimizations. This option is not turned on by any -O option since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions. It may, however, yield faster code for programs that do not require the guarantees of these specifications. Enables -fno-signed-zeros, -fno-trapping-math, -fassociative-math and -freciprocal-math. The default is -fno-reciprocal-math.

-ffinite-math-only. Allow optimizations for floating-point arithmetic that assume that arguments and results are not NaNs or +-Infs. This option is not turned on by any -O option since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions. It may, however, yield faster code for programs that do not require the guarantees of these specifications. The default is -fno-finite-math-only.

I compared #4 compile to the Lunatics cuda32 build and Q's remained at or above 99.92%.

Still looking for explanations of -DUSE_CUDA and the other -Dxxx_xxxx's in Petri's CFLAGS.

So, I suppose the next step in my learning process (if I want to get into this deeper) is to start looking at the code. Once I get familiar with the code, maybe some of the CFLAGS will start making better sense. I'm assuming the entry point in Xbranch is /sah_v7_opt/Xbranch/client/main.cpp?

On another note:
Finally moved the original app I compiled and tested about a week ago over to my active linux cruncher and got two MBs that immediately returned "computation error". I guess there's too much of a difference between a Fedora19/C2D(E7200)/GT620 and Fedora20/C2D(E7400)/GTX650ti. I was thinking they were going to be close enough to work but I guess I was wrong. I tried to test that app on the box that I compiled it on, but I was getting "no GPU found" for some reason. It must have had something to do with the hardware error that seemed to appear sometime while I was compiling/testing on that mobo. That mobo is now in my recycle pile.

Maybe my next step will be to reload my active linux cruncher with Fedora 19, compile an nvidia/mb app on it, and see if that works.

If there's something else any of you can think of that may be beneficial to try and report on, let me know.
4) Message boards : Number crunching : Step by Step, compile NVIDIA/MB CUDA app under Linux (Fedora 19) (Message 1525966)
Posted 9 Jun 2014 by Bubba Winkerbean
Well, I successfully did it again on an I7 machine with two GTX780ti's in it. I made changes to my original notes and changed the original post. See the first post in this thread for my more accurate notes. I almost completely re-wrote them.

Currently about 1/2 way through the full sized test units and comparing my most recently compiled app (setiathome_x41zc_x86_64-pc-linux-gnu_cuda60) on this machine with the Lunatics app (setiathome_x41g_x86_64-pc-linux-gnu_cuda32) and it's looking like I'm going to get the exact same results as I got when I did this on a Core 2 Duo a week or so ago.

So now my question is, what about all those CFLAGS (you use Petri) when ./configuring Xbranch? Is there some place I can find a list of them specific to configuring this MB app? Do I look through the source code for answers to specifying CFLAGS before configuring? Where did you get them Petri? I believe tuning through the use of those CFLAGS is key to compiling the fastest app.
5) Questions and Answers : Unix/Linux : Does any 64bit fedora linux work without additional libs (Message 1525666)
Posted 7 Jun 2014 by Bubba Winkerbean
http://setiathome.berkeley.edu/forum_thread.php?id=74094#1477422

This worked for me.

Edit: Step 1, then steps 16-26 if you want to crunch CPU only.
6) Message boards : Number crunching : Step by Step, compile NVIDIA/MB CUDA app under Linux (Fedora 19) (Message 1525558)
Posted 7 Jun 2014 by Bubba Winkerbean
The machine I origninally did this on cooked. Something on the mobo--no more ethernet port, inconsistent USB port--looks a lot like the south bridge cooked. Off to the recycle center it goes.

So I'm running through my cleaned up, step by step notes again on one of my newer machines and making adjustment. Some typos like "/.bashrc", not "/bashrc", and consolidating some steps. I'll repost them again once I make it through them again.

I'm currently stuck at step 31. (Permission denied? -- I'll figure it out and update my notes)
7) Message boards : Politics : Windows8: The Beginning of The End? Or... Win9 v soon!? (Message 1525147)
Posted 6 Jun 2014 by Bubba Winkerbean
Linux Mint has been mentioned in these threads in the recent past so I downloaded Cinnamon. It recognized the wireless NIC that other distributions do not, it comes with Office Libre and found my network printer/installed a basic driver for it, made it easy to download/install the NVIDIA GPU driver from the GPU it detected, and it was surprisingly intuitive to use. I didn't try to set it up to crunch, but my guess is it would have been relatively easy. I think it's a good 1st Linux distribution to try if you've never tried Linux before.

It's currently the most popular according to http://distrowatch.com/dwres.php?resource=major

http://distrowatch.com/ is also a good place to watch the new distributions as they come out and scroll down your screen.

*If you only have *one* computer and you rely on it for your finances and communications, I think it would be best for most people to stick with M$ for the time being.
8) Message boards : Politics : So now Obama has negated a 200 year policy of not negotiating with terrorists (Message 1524558)
Posted 5 Jun 2014 by Bubba Winkerbean
--
9) Message boards : Number crunching : Step by Step, compile NVIDIA/MB CUDA app under Linux (Fedora 19) (Message 1524105)
Posted 3 Jun 2014 by Bubba Winkerbean
*Some changes made since originally posting -- I recently determined the PC I originally did this on has some sort of hardware error and I stepped through this again on another machine. Verified, consolidated some steps, added some missing steps and fixed a typo or two.

Step by Step, compile NVIDIA/MB CUDA app under Linux (Fedora 19).
*The Cuda SDK/Toolkit is not supported by Fedora 20, but it is supported by Fedora 19.
*These are my cleaned up notes, second try ("mistake free" is almost guaranteed!)
*This is what I did and it seemed to work
*Purpose of this: maybe help some others get some linux boxes up, running and crunching MBs on their NVIDIA GPUs
*I know there are some short cuts I could have taken in many of my commands, but I made most of them explicit to try to help the nubes a bit more.

Notes for nubes:
*Fedora uses the "yum" command to install packages and take care of dependencies. Other Linux distributions such as Ubuntu use "apt-get"
*"vi" is a text editor
--arrow keys to move around
--'i' to insert
--'esc' to exit insert
--':' to enter command mode
--'x' to exit and save changes
--'q' to exit without saving changes
*To change from runlevel 5 (graphical) to runlevel 3 (text) in Fedora, hit 'e' soon after post, at the kernel selection screen, then add a "3" to the end of the "linux=" line using the arrow keys and then F10 to continue. "#shutdown now" to reboot to runlevel 5
* "$" is from user, "#" is from root login
* /home/guy will be different for you--/home/"WhateverYourLoginNameIs"

Major steps:
-Fresh install of Fedora 19
-Install Cuda SDK, Compiler, NVIDIA driver, test to see if it works
-Compile NVIDIA/MB app
-Test compiled app

1. Fresh install of Fedora 19, #yum update

2. #yum install gcc-c++ kernel-devel mesa-libGLU-devel.x86_64 libXi-devel.x86_64 libXmu-devel.x86_64 freeglut-devel.x86_64

3. Using Firefox, download cuda_6.0.37_Linux_64.run from developer.nvidia.com/cuda-downloads, move to /home/guy/Desktop
(cuda_6.0.37_Linux_64.run installs NVIDIA driver Linux-x86_64 331.62, Cuda Toolkit, and Samples)

4. #chmod +x cuda_6.0.37_Linux_64.run // (make it executable)

5. #mkdir /home/guy/Desktop/temp // (not enough space in partition with /tmp, need 2.1GB of free space)

6. #chmod 777 /home/guy/Desktop/temp // make it readable, writeable, executable

Note: I use /Desktop because it's already there and it's not used for anything else

7. #vi /etc/modprobe.d/blacklist.conf // append 'blacklist nouveau' (creates blacklist, blacklists nouveau)

8. #vi /etc/sysconfig/grub // append 'rd.driver.blacklist=nouveau' to end of 'GRUB_CMDLINE_LINUX="...about 3 lines of stuff..."' (be careful with the double quotes)

9. #grub2-mkconfig -o /boot/grub2/grub.cfg // (to pick up the above changes)

10. #yum remove xorg-x11-drv-nouveau.x86_64 // (deletes nouveau because it prevents the NVIDIA driver from working)

11. #mv /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r)-nouveau.img // (Backup old initramfs nouveau image)

12. #dracut /boot/initramfs-$(uname -r).img $(uname -r) // (Create new initramfs image without nouveau)

13. You don't have 'init 5' (graphic user interface) capability at this point because you just deleted nouveau. Make sure you catch the screen shortly after POST and hit 'e' to append a '3' using your arrow keys to the end of the "linux" line so when you hit F10 and continue, you enter text mode (init 3). If you miss this, hit reset and try again. Reboot to init 3

14. Login as root and cd to directory where you downloaded cuda_6.0.37_Linux_64.run (/home/guy/Desktop in my case)

15. #./cuda_6.0.37_Linux_64.run --override -tmpdir /home/guy/Desktop/temp //(--override is to force use of installed compiler)(^c to skip EULA or page down and read it if you want)

Accept EULA? accept Install Linux-x86_64 NVIDIA driver 331.62? y Install CUDA 6.0 toolkit? y Enter Toolkit location: <enter> for default Install symbolic link? y Install CUDA 6.0 samples? y CUDA 6.0 Samples location: <enter> for default At the end: Driver: Installed Toolkit: Installed in /usr/local/cuda-6.0 Samples: Installed in /root/NVIDIA_CUDA-6.0_Samples Notes from install: PATH /usr/local/cuda-6.0/bin LD_LIBRARY_PATH /usr/local/cuda-6.0/lib64, or, add /usr/local/cuda-6.0/lib64 to /etc/ld.so.conf and run ldconfig as root


16. reboot to init 5 (in my case, it hangs booting to init 5, I installed a later NVIDIA driver, NVIDIA-Linux-x86_64-331.79, and it then booted to init 5)

17. Need some $PATHs now.
Temporary: (applies now)
#export PATH=/usr/local/cuda-6.0/bin:$PATH
#export LD_LIBRARY_PATH=/usr/local/cuda-6.0/lib64:$LD_LIBRARY_PATH

or Permanent: (applies next reboot)
#vi ~/.bashrc
add a section:

#Paths
export PATH=/usr/local/cuda-6.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-6.0/lib64:$LD_LIBRARY_PATH

18. Test compiler (make files are included)
# cd /root/NVIDIA_CUDA-6.0_Samples/5_Simulations/fluidsGL
# make // should create binary 'fluidsGL'
# ./fluidsGL // should bring up an interesting graphic window and if you point, click and move your cursur, it'll make some interesting swirls)
(from /root/NVIDIA_CUDA-6.0_Samples, typing #make should compile all samples and put them in /bin/x86_64/linux/release according to the instructions)

Now that you have a working compiler, you can compile an NVIDIA/MB app. First need a few more things. The following are all files needed in accordance with http://boinc.berkeley.edu/trac/wiki/SoftwarePrereqsUnix and Claggy's list provided in another thread.

19. #yum install rapidsvn.x86_64 git m4 libtool mingw64-pkg-config.x86_64 libcurl-devel.x86_64 libsqlite3x-devel.x86_64 docbook2X.x86_64 libxml2-devel.x86_64 libsmbclient-devel.x86_64 libsmi-devel.x86_64 libnotify-devel.x86_64 libgtkhotkey-devel.x86_64 libjpeg-turbo-devel.x86_64 openssl-devel.x86_64

20. #cd /home/guy/Desktop

21. #svn checkout https://setisvn.ssl.berkeley.edu/svn/branches/sah_v7_opt
(creates /sah_v7_opt directory, Certificate: accept (p)ermanently, this copies all application source code to your computer)
-checked out revision 2392, yours may be a later revision

22. #git clone git://boinc.berkeley.edu/boinc-v2.git boinc // (creates boinc directory, this copies boinc source code to your computer)

23. #cd /home/guy/Desktop/boinc //need to compile boinc first, before you can compile the NVIDIA/MB app

24. #./_autosetup

25. #./configure --disable-server LDFLAGS=-static-libgcc BOINCDIR=/home/guy/Desktop/boinc

26. #make clean

27. #make //boinc now compiles with some warnings, but no hard errors

Now, to get an MB app to compile, I had to add a few things and comment out a few things. (Thanks Jason, Petri, Claggy for your help!)

28. #cd /home/guy/Desktop/sah_v7_opt/Xbranch/client/cuda

29. #vi cudaAcc_utilities.h
and add a line at the top of cudaAcc_utilities.h: #include "driver_types.h"

NOTE: With the following steps (30 and 31) I'm still getting inconsistent results. Not sure which files are created the first time you try to autosetup, configure, and make. All I can say for the time being is try to comment out and add entries to the following files if they exist. Then try to autosetup, configure and make. If it errors out, run through the list again. This is what I did on my second try from a fresh install on another computer.

30. If you have *not* tried to compile yet, comment out the following includes:

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/m4/sah_largefile_breaks_cxx.m4 -- sah_cxx_includes=`echo "#include \"$BOINCDIR/lib/std_fixes.h\"" ; echo $sah_cxx_includes`

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/m4/sah_header_stdcxx.m4 - #include \"$BOINCDIR/lib/std_fixes.h\"

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/configure.ac -- #include "std_fixes.h"

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/sah_config.h -- #include "std_fixes.h"

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/configure.ac.public -- #include "std_fixes.h"

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/db/xml_util.cpp -- #include "std_fixes.h"

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/client/analyzeFuncs.cpp -- *add* an #include "unistd.h" at the top

*grep returned these also, but since I'm not building for a MAC, I didn't comment them out.
/home/guy/Desktop/sah_v7_opt/Xbranch/mac_build/config-i386.h -- #include "std_fixes.h"
/home/guy/Desktop/sah_v7_opt/Xbranch/mac_build/config-ppc.h -- #include "std_fixes.h"

31. If you *have* tried to compile and it has errored out saying things like variables defined but not within this scope in files std_fixes.h and/or utils.h, comment out the following #includes:

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/client/cuda/db/xml_util.cpp -- #include std_fixes.h and #include utils.h

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/client/cuda/m4/sah_header_stdcxx.m4 -- #include \"BOINCDIR/lib/std_fixes.h\"

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/client/cuda/m4/sah_largefile_breaks_cxx.m4 -- sah_cxx_includes='echo "#include \"BOINCDIR/lib/std_fixes.h\"" ; echo $sah_css_includes'

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/client/cuda/sah_config.h -- #include "std_fixes.h" at the end

#vi /home/guy/Desktop/sah_v7_opt/Xbranch/client/cuda/sah_config.h.in -- #include "std_fixes.h" at the end (maybe not needed since I believe this is a backup file)

32. #cd /home/guy/Desktop/sah_v7_opt/Xbranch

33. #chmod +x _autosetup

34. #./_autosetup

35. #export CFLAGS=-msse2

36. #./configure BOINCDIR=/home/guy/Desktop/boinc CFLAGS='-DUSE_CUDA'

*NOTE: there are several dozen CFLAGS which I have not delved into which may make any particular compile crunch MBs faster based on your software and hardware. Petri has a long list of them.

37. #vi /home/guy/Desktop/sah_v7_opt/Xbranch/Makefile

Find the line (way down):
NVCCFLAGS = -03 --use_fast_math --ptxas-optioons"-v" --compiler-options "$(AM_CXXFLAGS) $(CXXFLAGS) -fno-strict-aliasing" -m64 -gencode arch=compute_10,code=sm_10 -gencode arch=compute_11,code=sm_11 -gencode arch=compute_12,code=sm_12 -gencode arch=compute_13,code=sm_13 -gencode arch=compute_20,code=sm_20 -gencode arch=compute_20,code=sm_21 -gencode arch=compute_20,code=sm_20

and add:

-gencode arch=compute_30,code=sm_30 -gencode arch=compute_35,code=sm_35 //for latest CUDA architecture.

NOTE: Not sure this actually accomplishes anything based on what I see scroll up during the make.

38. #make clean

39. #make

and this creates the NVIDIA/MB app file: seti_cuda setieathome_x41zc_x86_64-pc-linux-gnu_cuda60

There are some warnings, but no hard errors.

README NOTES: exacutable uses CUDA runtime (cudart32_xx_xx.dll), and CUFFT libraries (cufft32_xx_xx.dll), [Cuda 3.2, 4.2 and 5.0] or cudart.dll and cufft.dll [Cuda 2.2 and 2.3] -- these files need to be copied over along with the newly compiled app into your setiathome.berkeley.edu directory. Along with those, create an app_info.xml and it should work.

Now test your newly compiled NVIDIA/MB app:

40. Download:

KWSN Linux MB Bench V2.01.08 (http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=360)
V7 MB Test WUs 1 (http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=296)
V7 MB test WU 2 (http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=310)
WisGen WUs.7z (http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=384)

41. #cd /home/guy/Desktop

42. #mkdir /home/guy/Desktop/MBBenchmark

43. #chmod 777 /home/guy/Desktop/MBBenchmark // make it readable/writable

44. #cd /home/guy/Desktop/MBBenchmark

45. #mv /home/guy/Downloads/* /home/guy/Desktop/MBBenchmark/ // and move those files into it.

46. #chmod 777 * // make them all readable/writeable/executable

47. # cd /home/guy/Desktop

48. #chown guy:guy MBBenchmark // change owner so you can do the next step:

49. Using graphical file explorer, change file name "WisGen WUs.7z" to "WisGenWUs.7z" (command line does not like spaces in file names)

50. #yum install p7zip.x86_64

51. #cd /home/guy/Desktop/MBBenchmark // and unzip them in the following steps

52. #7za e KWSN-Bench-Linux-MBv7_v2.01.08.7z

Extracting KWSN-Bench-Linux-MBv7_v2.01.08/testWUs/old/readme
file readme
already exists. Overwrite with
KWSN-Bench-Linux-MBv7_v2.01.08/testData/readme?
(Y)es / (N)o / (A)lways / (S)kip all / A(u)to rename all / (Q)uit?

I answered "u"

It unzips five "readme" files. All of them are text files with "Compression Undeleter !" in them. So I deleted them to clean up the directory.

#rm readme* (answer y)

53. #mkdir /home/guy/Desktop/MBBenchmark/Zips

54. #chmod 777 Zips

55. #mv KWSN-Bench-Linux-MBv7_v2.01.08.7z /home/guy/Desktop/MBBenchmark/Zips //move it to Zips folder to clean directory

56. #mkdir /home/guy/Desktop/MBBenchmark/testWUsStorageArea // creates a place to store test work units while not in use

57. #chmod 777 /home/guy/Desktop/MBBenchmark/testWUsStorageArea // loosen up file security to make it easy

58. #cd /home/guy/Desktop/MBBenchmark/testWUsStorageArea

59. # mv /home/guy/Desktop/MBBenchmark/V7_MBtestWUs1.7z /home/guy/Desktop/MBBenchmark/testWUsStorageArea/V7_MBtestWUs1.7z
// move the test file zips into the storage area

60. #mv /home/guy/Desktop/MBBenchmark/V7_WU_FGset.7z /home/guy/Desktop/MBBenchmark/testWUsStorageArea/V7_WU_FGset.7z

61. #mv /home/guy/Desktop/MBBenchmark/WisGenWUs.7z /home/guy/Desktop/MBBenchmark/testWUsStorageArea/WisGenWUs.7z

62. #7za e V7_MBtestWUs1.7z // then unzip them and move the zips to the Zips folder to keep this directory clean

63. #mv /home/guy/Desktop/MBBenchmark/testWUsStorageArea/V7_MBtestWUs1.7z /home/guy/Desktop/MBBenchmark/Zips/V7_MBtestWUs1.7z

64. #rmdir V7_MBtestWUs1 // don't need it

65. #7za e V7_WU_FGset.7z

66. #mv /home/guy/Desktop/MBBenchmark/testWUsStorageArea/V7_WU_FGset.7z /home/guy/Desktop/MBBenchmark/Zips/V7_WU_FGset.7z

67. #7za e WisGenWUs.7z

68. mv /home/guy/Desktop/MBBenchmark/testWUsStorageArea/WisGenWUs.7z /home/guy/Desktop/MBBenchmark/Zips/WisGenWUs.7z

Make them all readable/writable, move them where they need to be:

69. #chmod 777 * // loosen up file security on all the test work units

70. #cp /home/guy/Desktop/sah_v7_opt/Xbranch/setiathome_x41zc_x86_64-pc-linux-gnu_cuda60 /home/guy/Desktop/MBBenchmark/APPS/setiathome_x41zc_x86_64-pc-linux-gnu_cuda60 // copy compiled app to /APPS directory

// If you want an app to compare yours to, you can download 'setiathome_x41g_x86_64-pc-linux-gnu_cuda32 from
// http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=311
// and put it in /MBBenchmark/REF_APPS
// after download (using Firefox):
// #cd /home/guy/Desktop/MBBenchmark
// #mkdir /home/guy/Desktop/MBBenchmark/LunaticsCuda32App
// #mv /home/guy/Downloads/Lunatics_x41g_linux64_cuda32.7z /home/guy/Desktop/MBBenchmark/LunaticsCuda32App/Lunatics_x41g_linux64_cuda32.7z
// #cd /home/guy/Desktop/MBBenchmark/LunaticsCuda32App
// #7za e Lunatics_x41g_linux64_cuda32.7z
// #mv /home/guy/Desktop/MBBenchmark/LunaticsCuda32App/Lunatics_x41g_linux64_cuda32.7z /home/guy/Desktop/MBBenchmark/Zips/Lunatics_x41g_linux64_cuda32.7z
// #cp /home/guy/Desktop/MBBenchmark/LunaticsCuda32App/setiathome_x41g_x86_64-pc-linux-gnu_cuda32 /home/guy/Desktop/MBBenchmark/REF_APPS/setiathome_x41g_x86_64-pc-linux-gnu_cuda32
// #cp /home/guy/Desktop/MBBenchmark/LunaticsCuda32App/libcudart.so.3 /home/guy/Desktop/MBBenchmark/libcudart.so.3

71. #cd /home/guy/Desktop/MBBenchmark/testWUs

72. #mv /home/guy/Desktop/MBBenchmark/testWUsStorageArea/PG* /home/guy/Desktop/MBBenchmark/testWUs // move short/fast test work units

73. #cp /home/guy/Desktop/MBBenchmark/testWUsStorageArea/_WisGenA.wu /home/guy/Desktop/MBBenchmark/testWUs

74. #mv /home/guy/Desktop/MBBenchmark/testWUs/_WisGenA.wu /home/guy/Desktop/MBBenchmark/testWUs/A1_WisGenA.wu // rename it, make it alphabetically first

75. #cp /home/guy/Desktop/MBBenchmark/testWUsStorageArea/_WisGenA.wu /home/guy/Desktop/MBBenchmark/testWUs // copy another

76. #mv /home/guy/Desktop/MBBenchmark/testWUs/_WisGenA.wu /home/guy/Desktop/MBBenchmark/testWUs/A2_WisGenA.wu // make it second

77. suspend or shutdown boinc (if running)

78. #cd /home/guy/Desktop/MBBenchmark

79. #chown guy:guy * // loosen up security so you can browse the output files using the graphical file explorer

80. #cp /home/guy/Desktop/MBBenchmark/LunaticsCuda32App/libcudart.so.3 /home/guy/Desktop/MBBenchmark // need libcudart.so.3 for cuda32

81. #cp /home/guy/Desktop/MBBenchmark/LunaticsCuda32App/libcufft.so.3 /home/guy/Desktop/MBBenchmark // need libcufft.so.3 for cuda32

82. #./benchmark


And that'll produce some output files in /MBBenchmark/testData with the results. In my case, my compiled app was just a few seconds faster, and also showed Q's of 99% or greater, which indicates it's probably a good build.

If you want to test on some full length test files, move the FG* files from the /testWUs_NOT IN USE STORAGE AREA into the /testWUs directory and move the PG* files into the storage area and run another ./benchmark.

As soon as my active linux cruncher runs out of work, I'll see if my newly compiled app works.

My app_info.xml file is going to look like this:

<app_info> <app> <name>setiathome_v7</name> </app> <file_info> <name>setiathome_x41zc_x86_64-pc-linux-gnu_cuda60</name> <executable/> </file_info> <file_info> <name>libcudart.so.3</name> <executable/> </file_info> <file_info> <name>libcufft.so.3</name> <executable/> </file_info> <app_version> <app_name>setiathome_v7</app_name> <version_num>700</version_num> <platform>x86_64-pc-linux-gnu</platform> <plan_class>cuda_fermi</plan_class> <avg_ncpus>0.05</avg_ncpus> <max_ncpus>1.0</max_ncpus> <coproc> <type>CUDA</type> <count>1.0</count> </coproc> <file_ref> <file_name>setiathome_x41zc_x86_64-pc-linux-gnu_cuda60</file_name> <main_program/> </file_ref> <file_ref> <file_name>libcudart.so.3</file_name> </file_ref> <file_ref> <file_name>libcufft.so.3</file_name> </file_ref> </app_version> </app_info>



**********Thanks once again Jason Claggy and Petri!
10) Message boards : Politics : PitifulPeacePrizin' POtus aka PrisonerExchanging You Can Believe In, 'is' Tee'd Up & Ready tO Act. HOle in One fO da Vrold. (Message 1524042)
Posted 3 Jun 2014 by Bubba Winkerbean
--
11) Message boards : Politics : So now Obama has negated a 200 year policy of not negotiating with terrorists (Message 1524040)
Posted 3 Jun 2014 by Bubba Winkerbean
--
12) Message boards : Politics : U.S. Presidential/V.P. candidates (Message 1524035)
Posted 3 Jun 2014 by Bubba Winkerbean
--
13) Message boards : Politics : So now Obama has negated a 200 year policy of not negotiating with terrorists (Message 1524033)
Posted 3 Jun 2014 by Bubba Winkerbean
--
14) Message boards : Politics : PitifulPeacePrizin' POtus aka PrisonerExchanging You Can Believe In, 'is' Tee'd Up & Ready tO Act. HOle in One fO da Vrold. (Message 1524029)
Posted 3 Jun 2014 by Bubba Winkerbean
--
15) Message boards : Politics : U.S. Presidential/V.P. candidates (Message 1523900)
Posted 3 Jun 2014 by Bubba Winkerbean
--
16) Message boards : Politics : U.S. Presidential/V.P. candidates (Message 1523380)
Posted 1 Jun 2014 by Bubba Winkerbean
--
17) Message boards : Politics : U.S. Presidential/V.P. candidates (Message 1523375)
Posted 1 Jun 2014 by Bubba Winkerbean
--
18) Message boards : Politics : U.S. Presidential/V.P. candidates (Message 1523373)
Posted 1 Jun 2014 by Bubba Winkerbean
--
19) Message boards : Politics : U.S. Presidential/V.P. candidates (Message 1523367)
Posted 1 Jun 2014 by Bubba Winkerbean
--
20) Message boards : Politics : Double standard on violence (Message 1523363)
Posted 1 Jun 2014 by Bubba Winkerbean
--


Next 20

Copyright © 2014 University of California