Posts by HA-SOFT, s.r.o.

21) (Message 3100)
Posted 23 May 2014 by Profile HA-SOFT, s.r.o.
Post:
AVX is optimized for Haswell. On Ivy and Sandy Bridge SSEx apps are better. Instruction set is not the only difference between apps. AVX has slightly different code than SSE app.

32bit apps are a little out in SSE and AVX world...
22) (Message 3058)
Posted 18 May 2014 by Profile HA-SOFT, s.r.o.
Post:

GTX 750ti 4,736 - 5,660

So newer cards are faster even with dp wu's, use much less power.

Better architecture.

the gpu app should be much faster than a cpu wu.

Definitely not.
23) (Message 3044)
Posted 16 May 2014 by Profile HA-SOFT, s.r.o.
Post:
Asteroids@Home uses single precision floating point calculations or double precision floating point calculations?


double.
24) (Message 3003)
Posted 9 May 2014 by Profile HA-SOFT, s.r.o.
Post:
am I right, that there is no speedup over our apps? I have tested FMA app long time ago on i5-4670 and there was no speedup over avx app.
25) (Message 2977)
Posted 6 May 2014 by Profile HA-SOFT, s.r.o.
Post:
Nice.
26) (Message 2929)
Posted 28 Apr 2014 by Profile HA-SOFT, s.r.o.
Post:
We have now two volunteers who try to make OpenCL app for AMD cards. Next I'm so busy at work for now. If both of them failed to do OpenCL app, I will do one.

Z
27) (Message 2911)
Posted 14 Apr 2014 by Profile HA-SOFT, s.r.o.
Post:
Version itself in xml is not important (version_num tag). Only file names are important. Version in file name is always in form VVVRR so 102.10 is 10210 in file name. 101.12 is 10112 in filename etc.

PS: I have problem viewing some posts with tags in forum.
28) (Message 2905)
Posted 11 Apr 2014 by Profile HA-SOFT, s.r.o.
Post:
I think it's fixed!


Nice.

Thanks
Z
29) (Message 2903)
Posted 6 Apr 2014 by Profile HA-SOFT, s.r.o.
Post:
Can anyone else confirm whether or not SETI@home or Milkyway@home use the Cuda55 to crunch the numbers as well? Because if they do then it's something wrong with Asteroids@home. I'm running those 2 and they are fine.


Please, try latest version of cuda app and let us know if it helps.
30) (Message 2830)
Posted 18 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
Is there any speedup over avx?
31) (Message 2828)
Posted 18 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
Is someone interested in experimenting with a app that uses the AMD fma4 instruction set ?

FWIW, i've been running it for a couple of weeks now at it works (valid results).

If someone is interested i could upload it.

Let me know.


Did you recode the app or change compiler settings only? I have FMA2 app and not difference between AVX and FMA2.
32) (Message 2807)
Posted 17 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
There is a maintenance now. Do not abort or delete tasks, they will be reported as soon as server becomes ready. You can check server status on project homepage.

Z
33) (Message 2803)
Posted 17 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
I have finished work which is never reported.
I've just started working with Astroid's tasks, but I feel like it's a waste of processing time if it's never going to update.
Anyone else having this problem?


There is a maintenance now. Do not abort or delete tasks, they will be reported as soon as server becomes ready. No time wasting.
34) (Message 2759)
Posted 15 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
Meanwhile we have increased wu's lenght. Now it's about 3500s.
35) (Message 2743)
Posted 14 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
Yes. Maintenance.
36) (Message 2722)
Posted 13 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
Last Pentathlon I crunch einstein and have got last credit after 9 months. Nobody loose credits and work is usefull at all events.
37) (Message 2706)
Posted 12 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
I have almost 60 validations waiting to be completed and it's growing everyday. I probably don't understand your question. I'm estimating. I was at about 7500 for avg./work at the end of last month. It's been consistently sliding since the server fix and application additions. I went back to 7.2.39, I was using 7.2.42 and I just wanted to see if there was any difference. It does appear to be working correctly for other clients, however. I'm very close to giving up on A@H...


If your tasks are in "Waiting for validation" state, you simply have to wait for other cruncher to calculate same task and validation will be processed. It's common behaviour of validation.
38) (Message 2701)
Posted 12 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
My RAC has gone down about 4000 points in the last 2 weeks. It's still taking an extraordinary amount of time to complete work units.


Validate or calculate?
39) (Message 2691)
Posted 11 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
WOW, the CUDA code of Asteroids is very well commented and readable!
I can understand it. 99 of it without looking into CUDA API reference! This is COOL.
Re-write this into OpenCL will be a toy :).

Hmm, will need the latest version of it soon.............. :)


I recommend you this document:

http://developer.amd.com/resources/heterogeneous-computing/opencl-zone/programming-in-opencl/porting-cuda-applications-to-opencl/
40) (Message 2690)
Posted 11 Mar 2014 by Profile HA-SOFT, s.r.o.
Post:
I dunno. I'm running BOINC 7.2.33 and I didn't do anything special. After I brought my Haswell online I got a mix of SSE3 and AVX tasks (as expected) but soon the SSE3 tasks stopped coming (as expected) and it's been only AVX tasks ever since. A small percentage, nowhere near 42%, are invalids.


That's because there is a difference between linux and windows in avx detection in boinc core.

Z


Previous 20 · Next 20