How to setup app_info.xml?

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

10 Nov 2014 17:04:22 UTC

Topic 197792

(moderation:

)

Hi,

i managed to compile the BRP4G and BRP5 cuda application (same app) on my linux with cuda 6.0.

1) Now I need help setting up a working app_info.xml.

I'd like to run CPU APPS:
- einstein_S5R6_1.01_graphics_i686-pc-linux-gnu
- hsgamma_FGRP4_1.04_x86_64-pc-linux-gnu__FGRP4-SSE2
- einstein_S5R6_1.01_graphics_i686-pc-linux-gnu
and MY GPU APP
- einsteinbinary_x86_64-pc-linux-gnu-cuda
.

2) I also need help about what to do with db.dev and dbhs.dev files? rename?

3) Should I put the NVIDIA v6.0 cudart and cufft files to einstein project directory too? Is it ok if they are in LIBPATH/PATH?

4) Before I try anything with production database I'd like to test my compilation against the standard app to validate results. How to do that off-line (not beta)? Is there a scrip for developers/optimizers?

--
petri33

Sid

Joined: 17 Oct 10

Posts: 164

Credit: 966214047

RAC: 372944

How to setup app_info.xml?

10 Nov 2014 18:49:29 UTC

Message 124618

(moderation:

)

Could you please guide how to compile BRP with cuda 6.0?
I was trying to compile the application too but stuck on some steps.
What exact linux did you use? What applications have to be installed before script can be run? What need to be fixed in provided script, etc.
Step by step guide would be ideal.

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

I use linux mint Petra and

10 Nov 2014 19:13:59 UTC

Message 124619 in response to message 124618

(moderation:

)

I use linux mint Petra and gcc (Ubuntu/Linaro 4.8.1-10ubuntu9) 4.8.1
.

I got the source code from here: Latest release

I had to apt-get a ton of packages.
"./build.sh --linux-cuda profile" followed by "cat build.log" told me what to do next (missing this and that).

Edit: libcurl and libjpg were missing or wrong version on my machine. libcurl needed some other packages and so on.

I'm sorry I did not keep a log of all the required steps. But mainly the task was to get all the required stuff.

The ./build.sh needed

...
# NVIDIA CUDA compiler wrapper options for GTX780
export NVCCFLAGS="-Xptxas -v -arch=compute_35 -code=compute_35  -maxrregcount=64 -g --verbose"

# component versions
BINUTILS_VERSION=2.22
GSL_VERSION=1.12
FFTW_VERSION=3.3.4
LIBXML_VERSION=2.6.32
ZLIB_VERSION=1.2.8
...

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

Where should I run

10 Nov 2014 19:26:24 UTC

Message 124620 in response to message 124619

(moderation:

)

Where should I run this?

./einsteinbinary_x86_64-pc-linux-gnu-cuda -t /home/petri/einstein/src/../test/templates_400Hz_2_short.bank -l /home/petri/einstein/src/data/zaplist_232.txt -A 0.04 -P 3.0 -W -z -i /home/petri/einstein/src/../test/J1907+0740_dm_482.binary -c status_profile.cpt -o results_profile.cand

I do not have a test directory under einstein. Nor do I have src/data.

Sid

Joined: 17 Oct 10

Posts: 164

Credit: 966214047

RAC: 372944

RE: I use linux mint Petra

10 Nov 2014 20:56:18 UTC

Message 124621 in response to message 124619

(moderation:

)

Quote:

I use linux mint Petra and gcc (Ubuntu/Linaro 4.8.1-10ubuntu9) 4.8.1
.

I got the source code from here: Latest release

I had to apt-get a ton of packages.
"./build.sh --linux-cuda profile" followed by "cat build.log" told me what to do next (missing this and that).

Edit: libcurl and libjpg were missing or wrong version on my machine. libcurl needed some other packages and so on.

I'm sorry I did not keep a log of all the required steps. But mainly the task was to get all the required stuff.

The ./build.sh needed
...
# NVIDIA CUDA compiler wrapper options for GTX780
export NVCCFLAGS="-Xptxas -v -arch=compute_35 -code=compute_35  -maxrregcount=64 -g --verbose"
# component versions
BINUTILS_VERSION=2.22
GSL_VERSION=1.12
FFTW_VERSION=3.3.4
LIBXML_VERSION=2.6.32
ZLIB_VERSION=1.2.8
...

Thank you. I tried on ubuntu, fedora, suse and some more Linuxes. Doing so I had to change compilation script, for example to set ZLIB version to 1.2.9, then I got compilation error for one source file.
Now I have a hint.

Sid

Joined: 17 Oct 10

Posts: 164

Credit: 966214047

RAC: 372944

RE: Where should I run

10 Nov 2014 20:58:31 UTC

Message 124622 in response to message 124620

(moderation:

)

Quote:

Where should I run this?

./einsteinbinary_x86_64-pc-linux-gnu-cuda -t /home/petri/einstein/src/../test/templates_400Hz_2_short.bank -l /home/petri/einstein/src/data/zaplist_232.txt -A 0.04 -P 3.0 -W -z -i /home/petri/einstein/src/../test/J1907+0740_dm_482.binary -c status_profile.cpt -o results_profile.cand

I do not have a test directory under einstein. Nor do I have src/data.

Why just not to replace existing application and give it a try?
Worse scenario would be a validation error.

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

I think there is a protection

10 Nov 2014 21:09:23 UTC

Message 124623 in response to message 124622

(moderation:

)

I think there is a protection mechanism to prevent maliscious software impersonating the original. A checksum or something. My guess is that you can not even edit the cuda files in the project directory.

The last routine does not need shared memory caching, but try to remove that just by editing. The file gets replaced and the tasks that were 99% complete are lost into the darkest bit heaven.

Holmis

Joined: 4 Jan 05

Posts: 1118

Credit: 1055935564

RAC: 0

RE: I think there is a

10 Nov 2014 21:39:44 UTC

Message 124624 in response to message 124623

(moderation:

)

Quote:

I think there is a protection mechanism to prevent maliscious software impersonating the original. A checksum or something. My guess is that you can not even edit the cuda files in the project directory.

When you're assigned a task from a new search the scheduler tells Boinc where to get the executables needed and what file signature they should have, Boinc then downloads the files and calculate the signature, if it's a match all is well but if not you get a error on the task with something like "Could not get input file".
If you know what type of file signature Boinc uses and edit the client_state.xml you should be able to trick Boinc into using your own app. The anonymous platform mechanism would be better as that's why it's there.

I responded to your thread in the "Problems and Bugs forum" before I noticed this one...

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

RE: I responded to your

11 Nov 2014 0:14:25 UTC

Message 124625 in response to message 124624

(moderation:

)

Quote:

I responded to your thread in the "Problems and Bugs forum" before I noticed this one...

Thanks, I'll take a look at that.

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

RE: RE: I use linux mint

22 Nov 2014 8:42:17 UTC

Message 124626 in response to message 124621

(moderation:

)

Quote:

Quote:
I use linux mint Petra and gcc (Ubuntu/Linaro 4.8.1-10ubuntu9) 4.8.1
.

I got the source code from here: Latest release

I had to apt-get a ton of packages.
"./build.sh --linux-cuda profile" followed by "cat build.log" told me what to do next (missing this and that).

Edit: libcurl and libjpg were missing or wrong version on my machine. libcurl needed some other packages and so on.

I'm sorry I did not keep a log of all the required steps. But mainly the task was to get all the required stuff.

The ./build.sh needed
...
# NVIDIA CUDA compiler wrapper options for GTX780
export NVCCFLAGS="-Xptxas -v -arch=compute_35 -code=compute_35  -maxrregcount=64 -g --verbose"
# component versions
BINUTILS_VERSION=2.22
GSL_VERSION=1.12
FFTW_VERSION=3.3.4
LIBXML_VERSION=2.6.32
ZLIB_VERSION=1.2.8
...
Thank you. I tried on ubuntu, fedora, suse and some more Linuxes. Doing so I had to change compilation script, for example to set ZLIB version to 1.2.9, then I got compilation error for one source file.

Now I have a hint.

and another change in src/Makefile.linux.cuda

LIBS += /usr/local/cuda/lib64/libcufft.so  # CUDA FFT
LIBS += /usr/local/cuda/lib64/libcudart.so # CUDA runtime library
#LIBS += $(EINSTEIN_RADIO_SRC)/cuda/redist/lin$(ARCH)/3.2/lib/libcufft.so.3  # CUDA FFT
#LIBS += $(EINSTEIN_RADIO_SRC)/cuda/redist/lin$(ARCH)/3.2/lib/libcudart.so.3 # CUDA runtime library

petri33

Joined: 3 Nov 14

Posts: 12

Credit: 3892550

RAC: 0

I compiled a CUDA 6.0 version

23 Nov 2014 6:10:34 UTC

Message 124627

(moderation:

)

I compiled a CUDA 6.0 version and I have run it succesfully. It takes about 72 minutes to process one BRP5.

Now I'm trying to get it working with BOINC ... Boinc starts ok, finds config files but does not request any work.

EventLog shows this:

su 23. marraskuuta 2014 09.04.21 | Einstein@Home | [error] State file error: missing application file einsteinbinary_x86_64-pc-linux-gnu-cuda5

I have set up an app_config.xml

einsteinbinary_BRP4G

0.5
0.75

einsteinbinary_BRP5

0.5
0.75

... and an app_info.xml

einsteinbinary_BRP5

einsteinbinary_BRP5
099
7.2.39

einsteinbinary_x86_64-pc-linux-gnu-cuda5

einsteinbinary_x86_64-pc-linux-gnu-cuda5-db.dev
db.dev

einsteinbinary_x86_64-pc-linux-gnu-cuda5-dbhs.dev
dbhs.dev

einsteinbinary_BRP4G

einsteinbinary_BRP4G
099
7.2.39

einsteinbinary_x86_64-pc-linux-gnu-cuda4G

einsteinbinary_x86_64-pc-linux-gnu-cuda4G-db.dev
db.dev

einsteinbinary_x86_64-pc-linux-gnu-cuda4G-dbhs.dev
dbhs.dev

Why does my computer not get any work?

linux1 BOINC # cat sched_request_einstein.phys.uwm.edu.xml

21238049849428023948029384022414
11683437
1353
7
2
39
0.000000
1.000000
0.000000
0.573224
0
0
0.000000
0.000000
0.000000
0.000000
1
anonymous

1024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.

http://einstein.phys.uwm.edu/
1415046160.000000
90.000000
40.000000
0
1
1
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
1
0
0
1
10.000000
10.000000
50.000000
60.000000
60.000000
100.000000
50.000000
0.100000
70.000000
95.000000
95.000000
1.000000
0.000000
0.000000
100.000000
0.000000
0
1
1
18

http://einstein.phys.uwm.edu/
http://einstein5.aei.uni-hannover.de/EinsteinAtHome_cgi/cgi

1415046160
0
1
1
1
0
0
1
60
6
50
100
1
0.03
50
3600
75
50
90
0
0.1
0
0
0
0
0
0
0

0
1
0
1
0
25
0
60
18
100
100
100
0.1
50
60
75
50
90
1
1
0
0
0
0
0
0
0

0
1
0
1
0
25
1
200
0
50
100
1
0.1
50
60
75
50
90
1
1
0
0
0
0
0
0
0

0
1
0
1
0
25
0
60
0
50
100
100
0.001
50
60
75
50
90
1
1
0
0
0
0
0
0
0

6xxxxxxxxxxxxxxxxxxxxxx8
dbxxxxxxxxxxxxxxxxxxxxxxxxxxxxx3

0.982510
-1.000000
0.999922
0.999922
0.999922
1416646195.128399
1245.415428
1416647440.543827

21932.878713
37831064.365775
1416646939.539175
5273146.652706
929294720.215371
1416647437.766123

7200
linux1
127.0.1.1
bxxxxxxxxxxxxxxxxxxxxxxxxxxxxf
12
GenuineIntel
Intel(R) Core(TM) i7-3930K CPU @ 3.20GHz [Family 6 Model 45 Stepping 6]
fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer aes xsave avx lahf_lm ida arat epb xsaveopt pln pts dtherm tpr_shadow vnmi flexpriority ept vpid
2751274813.387191
25810809610.739918
1000000000.000000
1416338171.455112
0
8315920384.000000
12582912.000000
12693319680.000000
108858007552.000000
79361695744.000000
Linux
3.13.0-031300-generic

381907363.000000
69814403.000000
5442900377.600000

4
GeForce GTX 780
3147472896.000000
1
1
0.000000
0.000000
0.000000
4336128000000.000000
6000
0
3221028864.000000
49152.000000
65536
32
2147483647.000000
1024
1024 1024 64
2147483647 65535 65535
941000
65536.000000
3
5
512.000000
1
12

GeForce GTX 780
NVIDIA Corporation
4318
1
0
63
63
1
1
cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64
3221028864
49152
941
4647714815446351884
OpenCL 1.1 CUDA 6.0.1
OpenCL 1.1 CUDA
337.12

JPLEPH.405
9319680.000000
1

rand_BRP5b.bank
11953776.000000
1

stochastic_full.bank
313687.000000
1

How to setup app_info.xml?

Forums › Cruncher's Corner

Comment viewing options

Forums › Cruncher's Corner