Posts by jason_gee

1) Message boards : Number crunching : CES 2017 -- AMD RYZEN CPU (Message 1850716)
Posted 1 day ago by Profile jason_gee
Post:
Hmmm, certainly is starting to look compelling. Might be time to dust off the compilers and mutithread the CPU apps.

Of all the leaked CPU performance tests, Ryzen beat out the equivalent Intel CPU in 6 of 8 tests. What I found interesting is the R7 was considerably slower in the Prime Number factoring test. I wonder how can that be when it was considerably faster in the Integer and FFT tests. Does prime factoring NOT involve FFT? I thought it did based on my observations of the AIDA64 and OCCT Prime stress tests. Or is it just that the Prime test is optimized to Intel compiler or something?


Unsure on that. A cursory look at algorithms for prime factoring says they vary widely. While i don't immediately spot explicit FFTs involved, they could be being used as a black box in some of the tests.

For problems of any reasonably large size like that, we'd probably have to find out what sortof memory configuration was used.In the few where I have seen it specified timings were loose and speed lackluster. If libraries with run time optimisation (such as fftw) are used, then it'll pick fastest code paths (i.e. 'wisdom'), however given that we're supposedly looking at a clean slate design, it's unlikely that optimal codelets exist yet in bench software.
2) Message boards : Number crunching : CES 2017 -- AMD RYZEN CPU (Message 1850707)
Posted 1 day ago by Profile jason_gee
Post:
Hmmm, certainly is starting to look compelling. Might be time to dust off the compilers and mutithread the CPU apps.
3) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850654)
Posted 1 day ago by Profile jason_gee
Post:
What is sad is that we are currently working 8-9 year old arbitrary data when we should be working BLC data on this system, https://www.nasa.gov/press-release/nasa-telescope-reveals-largest-batch-of-earth-size-habitable-zone-planets-around
The last I heard there was a problem with the delivery of any BLC data, that was a week or so ago. Seems odd with today's high speed internet to have delivery problems.


Yeah, that's the rub alright
4) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850645)
Posted 1 day ago by Profile jason_gee
Post:
Disappointing to see politics erupt on this board :-(

Sad truth indeed, when politics affects the science :-(
5) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850643)
Posted 1 day ago by Profile jason_gee
Post:
Maybe this project has evolved from a science project, to a "food on the table" project?
When the benefactor determines no bacteria have been produced in the Petri Dish, and asks for a final report, how long do you keep the Petri Dish?


That's easy. You put the petri dish aside, and start on the next experiment. You may then either have a spare petri dish, or stumble on a 'happy accident' (allusion to the discovery of penicillin)
6) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850636)
Posted 1 day ago by Profile jason_gee
Post:
Maybe this project has evolved from a science project, to a "food on the table" project?

I agree, though wouldn't necessarily use the word 'evolved'. If the scientists involved would stand up and state their positions, then maybe we could find them a better home, than the apparent wasteland they're in.

Hmm, but the scientists might be in an impossible situation. By standing up and stating their positions, the "food on the table" may all of a sudden turn into dust.
Heck, the whole table and the house might disappear totally.


Yep, and now you know why Nero fiddled while Rome burned.

We're on the same page I believe :-)


World's a changin' . The sleeper has awakened.
7) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850632)
Posted 1 day ago by Profile jason_gee
Post:
Maybe this project has evolved from a science project, to a "food on the table" project?

I agree, though wouldn't necessarily use the word 'evolved'. If the scientists involved would stand up and state their positions, then maybe we could find them a better home, than the apparent wasteland they're in.

Hmm, but the scientists might be in an impossible situation. By standing up and stating their positions, the "food on the table" may all of a sudden turn into dust.
Heck, the whole table and the house might disappear totally.


Yep, and now you know why Nero fiddled while Rome burned.
8) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850629)
Posted 1 day ago by Profile jason_gee
Post:
Maybe this project has evolved from a science project, to a "food on the table" project?

I agree, though wouldn't necessarily use the word 'evolved'. If the scientists involved would stand up and state their positions, then maybe we could find them a better home, than the apparent wasteland they're in.
9) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850622)
Posted 1 day ago by Profile jason_gee
Post:
The fact that Berkeley allows criminals onto campus to beat people, and orders security to stand aside, says everything to me about how much care is being taken with the data.
10) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850575)
Posted 2 days ago by Profile jason_gee
Post:
Anybody remembers CP/M edit?
Tullio


Oddly rings a bell though I can't place it.


. . CP/M was an old OS earlier than MS-Dos and the Commodore 128 ran it as well as its native mode. It had an edit app which was CP/M edit.

Stephen

:)


LMAO, that's it. My father had some gizmo about 1981, and went away for work. When he returned I was playing blackjack, and he was annoyed that I got something working that he didn't. Naturally I did so by reading the instructions to loading the tape.... LMAO
11) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850528)
Posted 2 days ago by Profile jason_gee
Post:
Anybody remembers CP/M edit?
Tullio


Oddly rings a bell though I can't place it.
12) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850522)
Posted 2 days ago by Profile jason_gee
Post:
Me Sir, Me SIR , MEEEEE


You sir won a free Rick roll: https://youtu.be/r0IwfHdO36Y
13) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1850517)
Posted 2 days ago by Profile jason_gee
Post:
hands up if you remember copy con
14) Message boards : Number crunching : SETI's 12th birthday challenge (March 20-25) (Message 1850469)
Posted 2 days ago by Profile jason_gee
Post:
We're coming out of some kindof 'funk' that I'd argue no-one completely understands.

I'd say it's borked.

Yeah, well I was trying to be nice.
15) Message boards : Number crunching : SETI's 12th birthday challenge (March 20-25) (Message 1850462)
Posted 2 days ago by Profile jason_gee
Post:
We're coming out of some kindof 'funk' that I'd argue no-one completely understands. Can you postpone for a week ?
16) Message boards : Number crunching : Back to 99% GPU's? (Message 1849725)
Posted 5 days ago by Profile jason_gee
Post:
Normally I would suggest a couple of instances for cuda apps, though with 750/750ti, single instance Cuda50 should be enough.


Ok, I'm ignorant, is that something I can do in the MCuda*.txt file?


Nah, number of instances for a given application is controlled via Boinc's optional app_config.xml file. Someone else will have to chime in with what to put in there for the various applications, just because I use an older Boinc version, and so different methods myself (on the one machine I run multi instance anyway)
17) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1849507)
Posted 6 days ago by Profile jason_gee
Post:
Would recommend checking a vanilla zi Vs vanilla CPU with Gaussians in extreme circumstances, because they have been problematic on and off at different times. They are relatively sensitive to platform variation (though that shouldn't impact best selection logic of course)
18) Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use (Message 1849506)
Posted 6 days ago by Profile jason_gee
Post:
out of curiosity, which programming language are people using to create their own special apps? is it a matter of improving code or optimizing during compiling?


Primarily hand coded optimisations in both the GPU Cuda/OpenCL and Host CPU C/C++ code. In the case of Cuda special apps, that tends to be a combination of higher level structural optimisations, such as adapting to use Cuda Streams, along with low level PTX assembly (&/or intrinsics). The higher level structural enhancements tend to allow better scaling on newer hardware, while the low level assembly tends to improve memory access patterns and instruction level paralellism. Compiler options tended to decrease in importance after about Cuda 4, since nVidia switched to an LLVM based compiler. That relatively modern type of compiler tends to do a reasonable job without much twiddling.
19) Message boards : Number crunching : Back to 99% GPU's? (Message 1849333)
Posted 7 days ago by Profile jason_gee
Post:

So the next question is what can I do to raise the GPU loading (and presumably speed up the Cuda 50 processing?).


[Edit:] Note, in addition to below suggestions, you may also be getting 'Shorty' (Very High Angle Range Arecibo ) tasks, which can present a lower load naturally.

Normally I would suggest a couple of instances for cuda apps, though with 750/750ti, single instance Cuda50 should be enough. So your guess of trying the processpriority key will probably help somewhat. Aside from that, any background applications such as the network streaming fruit and telemetry/diagnostics present with newer drivers [and GeForce Experience] can impact that loading under Cuda. There are guides around on the internet for disabling much bloat there. The reasons OpenCL should be less impacted there would be that OpenCL is wired in at a lower level than cuda, and is written to 'spread' more anyway. Probably future applications will have some more intuitive loading controls (and similarly conservative defaults), but for now getting the best out a particular GPU can be a matter of trial and error.
20) Message boards : Number crunching : Panic Mode On (104) Server Problems? (Message 1848776)
Posted 9 days ago by Profile jason_gee
Post:
20. The End-of-file (EOF) check may not be
enough
...
On top of that, incorrect data will be processed in the loop, as unknown values
will be getting to the tmp variable.
To avoid issues like that, we need to use additional methods to check the stream status:
bad(), fail().
...

Interesting. Wonder if that could also be an explanation for tapes that when failing end up with end channels full of errors?


With codebases so complex, it'd certainly be one possibility of many. The code resides in a utility library set of functions probably used by all the servers, it could certainly account for some amount of fragility under stress. Either way, refactoring to harden this won't hurt.


Next 20


 
©2017 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.