@Crunch3r

Hi!

I just thought this morning it ain't a bad idea to run some test of your optimized seti clients against the reference work unit on my Athlon XP 3200+.
So, I tested your clients V2.7s, V2.08s and V2.09s on this host. The tests finished some minutes ago, but the result is somewhat confusing to me.

Here are the results:
AthlonXP_SSE_FFTW3_Caching_V2.07s: 5808.850943
AthlonXP_SSE_FFTW3_Caching_V2.08s: 5508.440612
AthlonXP_SSE_FFTW3_Caching_V2.09s: 8814.406029

Do you know the reasons for the 2.09 slower than 2.08 or 2.07 (on my host)?

The Hardware is an Athlon XP 3200+ , no overclocking, 512 MB RAM
The OS is Debian 3.1 kernel 2.6.8-2-386


Would be nice if you could give me some enlightment ;-)

Regards
rattelschneck

Have you copied the "wisdom" file into the directory of the setiathome app. ?

If not it should have been created by now and you should rerun the benchmark.
I bet you'll see a difference :-)

---

Join BOINC United now!

]
Have you copied the "wisdom" file into the directory of the setiathome app. ?

If not it should have been created by now and you should rerun the benchmark.
I bet you'll see a difference :-)

The wisdom file is included in you tarball, sure it is there.

See the file date: 2005-11-11 22:51 wisdom

]
Have you copied the "wisdom" file into the directory of the setiathome app. ?

If not it should have been created by now and you should rerun the benchmark.
I bet you'll see a difference :-)

The wisdom file is included in you tarball, sure it is there.

See the file date: 2005-11-11 22:51 wisdom

Could you please check the file date of the app. is it Nov 13 22:20 ?

If that's then the output of client would be helpfull.

I never had any problem with this client and it's the fastes i've ever build for Athlon XP.

---

Join BOINC United now!

]
Have you copied the "wisdom" file into the directory of the setiathome app. ?

If not it should have been created by now and you should rerun the benchmark.
I bet you'll see a difference :-)

The wisdom file is included in you tarball, sure it is there.

See the file date: 2005-11-11 22:51 wisdom

Could you please check the file date of the app. is it Nov 13 22:20 ?

If that's then the output of client would be helpfull.

I never had any problem with this client and it's the fastes i've ever build for Athlon XP.

The file date is:
2895828 2005-11-13 22:20 setiathome-4.7.i686-pc-linux-gnu

The ouptut stored in the stderr.txt is:

rattelschneck@athlonxp:~$ cat reference_AthlonXP_SSE_FFTW3_Caching_V2.09s/stderr.txt
Can't open init data file - running in standalone mode
Windows optimized S@H application by truXoft
Improvements by Tetsuji Maverick Rai, Hans Dorn, Harold Naparst, Ned Slider, Crunch3r, trux,...
$URL: svn://hnaparst.homelinux.com/seti_boinc/tags/naparst-r3.5/client/analyzeFuncs.cpp $
$Build: GCC 4.0.2 STATIC Athlon XP SSE/3DNow V2.09s by Crunch3r $
$Rev: 166 Athlon XP GCC ACML/FFTW V2.09s $
$Date: 2005-11-11 19:23:59 +0100 (Fri, 11 Nov 2005) $
Reading wisdom from file.
cache calculated 53 times

]
Have you copied the "wisdom" file into the directory of the setiathome app. ?

If not it should have been created by now and you should rerun the benchmark.
I bet you'll see a difference :-)

The wisdom file is included in you tarball, sure it is there.

See the file date: 2005-11-11 22:51 wisdom

IIRC the wisdom file is the result of some benchmarks that are run by fftw.
It might be a good idea to work with a newly generated one that fits to your host.

Regards Hans

---

IIRC the wisdom file is the result of some benchmarks that are run by fftw.
It might be a good idea to work with a newly generated one that fits to your host.

Regards Hans

Shell I touch it, or is it created by the client itself?

IIRC the wisdom file is the result of some benchmarks that are run by fftw.
It might be a good idea to work with a newly generated one that fits to your host.

Regards Hans

Shell I touch it, or is it created by the client itself?

Try hiding it. The client should create a new one on the first run, and use it later.

Regards Hans

---

OK, I rerun the test after deleting all relevant files, including the wisdom file.

The result for the initial run without wisdom is now:
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.308632


A second run using the new created wisdom file:
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.294634



So the corrected list of results looks like this now:
AthlonXP_SSE_FFTW3_Caching_V2.07s: 5808.850943
AthlonXP_SSE_FFTW3_Caching_V2.08s: 5508.440612
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.294634


Does this mean as a consequence, if I use V2.09s better not using the delivered wisdom file? If so, why is it included in first place?

Regards
rattelschneck

OK, I rerun the test after deleting all relevant files, including the wisdom file.

The result for the initial run without wisdom is now:
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.308632


A second run using the new created wisdom file:
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.294634



So the corrected list of results looks like this now:
AthlonXP_SSE_FFTW3_Caching_V2.07s: 5808.850943
AthlonXP_SSE_FFTW3_Caching_V2.08s: 5508.440612
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.294634


Does this mean as a consequence, if I use V2.09s better not using the delivered wisdom file? If so, why is it included in first place?

Regards
rattelschneck

It's included because it was recommended by harold to avoid the creation in the first run because it could be confusing if the processing time would be higher than previous version (only for the first run while creating the wisdom file).

---

Join BOINC United now!

OK, I rerun the test after deleting all relevant files, including the wisdom file.

The result for the initial run without wisdom is now:
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.308632


A second run using the new created wisdom file:
AthlonXP_SSE_FFTW3_Caching_V2.09s: 5508.294634

Does this mean as a consequence, if I use V2.09s better not using the delivered wisdom file? If so, why is it included in first place?

Regards
rattelschneck

I'm a bit confused now :o)
Generating the wisdom should take some time. I wonder why the first and second run took the same time.

Did you use the "time" command to get the cpu times?
The cpu times reported by the client in standalone mode are plain wrong!

Regards Hans

---

Did you use the "time" command to get the cpu times?
The cpu times reported by the client in standalone mode are plain wrong!

Regards Hans

No, I didn't use the time command. The reported times are taken from the init_data.xml file.
I could rerun the tests using the time commands, if you like.

No, I didn't use the time command. The reported times are taken from the init_data.xml file.
I could rerun the tests using the time commands, if you like.

If you can spare the cycles :o)

It might help us to decide how to get the best results with fftw.

Regards Hans

---

Hi

why at the results Rev. 3.4 stands
Windows optimized ....

Thanks

---

Hi

why at the results Rev. 3.4 stands
Windows optimized ....

Trux made several attempts to get rev 3.x working under Windows. So far, he's been unsuccessful. As far as I remember, he said that the application crashes randomly, and the crashes are not reproducible. If you have experience in creating and debugging Windows applications, please post here and I'm sure you can get this effort jump-started. Please note that you need an expensive commercial license for ICC and IPP for Windows (Linux non-commercial version is free).

---

Here is an update for the results on Crunch3r optimized clients V2.07 - V2.09 running on my AthlonXP 3200+.
This time I used the time command to run the tests.

---------------------------------------
AthlonXP_SSE_FFTW3_Caching_V2.07s
real 81m11.129s
user 81m7.585s
sys 0m0.726s

<wu_cpu_time>5808.771954</wu_cpu_time>
---------------------------------------
AthlonXP_SSE_FFTW3_Caching_V2.08s
real 79m16.031s
user 79m11.879s
sys 0m0.802s

<wu_cpu_time>5508.358624</wu_cpu_time>
---------------------------------------
AthlonXP_SSE_FFTW3_Caching_V2.09s (without wisdom file)
real 129m17.573s
user 129m8.977s
sys 0m5.463s

<wu_cpu_time>8513.813726</wu_cpu_time>
---------------------------------------

The wisdom file has been created on the end of the run above. The computer is just busy with another run using the newly created wisdom file.
I will keep you informed...

Regards
rattelschneck

Here is the missing result from the last run with wisdom file.

---------------------------------------
AthlonXP_SSE_FFTW3_Caching_V2.09s with wisdom file
real 127m42.493s
user 127m38.403s
sys 0m0.409s

<wu_cpu_time>8513.650751</wu_cpu_time>
---------------------------------------

To be honest, I'm still a bit confused when looking at the results.
The time command revealed a little better results for the run with the wisdom file, but how come that V2.08 and even V2.07 seems to be faster?

Here is the missing result from the last run with wisdom file.

---------------------------------------
AthlonXP_SSE_FFTW3_Caching_V2.09s with wisdom file
real 127m42.493s
user 127m38.403s
sys 0m0.409s

<wu_cpu_time>8513.650751</wu_cpu_time>
---------------------------------------

To be honest, I'm still a bit confused when looking at the results.
The time command revealed a little better results for the run with the wisdom file, but how come that V2.08 and even V2.07 seems to be faster?

Actually i don't know what's causing this. I'm using the V2.09 version myself and had no problems at all.

You should give it a try on "live" WUs and see what happens.

---

Join BOINC United now!

First of all the ia64 was build with gcc and sure it is a bit outdated. If you want anew client i can build one for you.

That would be nice. Is there a reason why you haven't tried icc and ipp? Both are available for Itaniums.

Would be interesting if you could post processing times on you itanium ( your computers are hidden so i can't look myself)

Here it is. The average time is 9,757s.

I've build the IA64 setiathome app. and it's available for download @ my page.
The fftw wisdom file will be created while processing the fist wu so the second Wu processed should give a nice speedup.

It's build with gcc because icc gave to much errors while compiling.

---

Join BOINC United now!

I've build the IA64 setiathome app. and it's available for download @ my page.
The fftw wisdom file will be created while processing the fist wu so the second Wu processed should give a nice speedup.

I'll take it for a spin.

It's build with gcc because icc gave to much errors while compiling.

I am absolutely sure someone who knows both boinc and seti application from inside out needs to do quite a few code changes to get boinc/seti working with icc. For example, -DW7 and -xW options make no sense for ia64. I tried to get the build working on EM64T and failed miserably. Itanium got to be harder than EM64T.

---

I've build the IA64 setiathome app. and it's available for download @ my page.
The fftw wisdom file will be created while processing the fist wu so the second Wu processed should give a nice speedup.

I'll take it for a spin.

It's build with gcc because icc gave to much errors while compiling.

I am absolutely sure someone who knows both boinc and seti application from inside out needs to do quite a few code changes to get boinc/seti working with icc. For example, -DW7 and -xW options make no sense for ia64. I tried to get the build working on EM64T and failed miserably. Itanium got to be harder than EM64T.

Not as hard as you might think. IPP/ICC have full support for Itanium.
It is only a matter of linking against the correct libs and including the
correct header files in config.h.in

---

Harold Naparst