New improved versions of SSE2 and SSE3 released


Message boards : News : New improved versions of SSE2 and SSE3 released

Message board moderation

To post messages, you must log in.
1 · 2 · Next
AuthorMessage
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1704 - Posted: 7 Sep 2013, 21:06:51 UTC
New improved versions of SSE2 and SSE3 has been released for 32 and 64bit linux and windows.

Radim Vančo (Kyong)
ID: 1704 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
nanoprobe

Send message
Joined: 15 Jan 13
Posts: 12
Credit: 904,320
RAC: 0
Message 1706 - Posted: 7 Sep 2013, 22:10:26 UTC

Last modified: 7 Sep 2013, 22:10:33 UTC

ID: 1706 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
nanoprobe

Send message
Joined: 15 Jan 13
Posts: 12
Credit: 904,320
RAC: 0
Message 1707 - Posted: 7 Sep 2013, 22:17:13 UTC

Last modified: 7 Sep 2013, 22:20:45 UTC
Tried to download some new tasks. Found this error in message log.

9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_17210_132: permanent HTTP error
9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_17210_207: permanent HTTP error
9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_18130_225: permanent HTTP error
9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_18131_37: permanent HTTP error
9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_17239_151: permanent HTTP error
9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_17239_153: permanent HTTP error
9/7/2013 6:14:02 PM | Asteroids@home | Giving up on download of input_17268_88: permanent HTTP error

Something wrong on my end? Never saw that error before. XP 32 bit.

Tried getting new tasks on second machine. Same error. Win7 64 bit.
ID: 1707 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
boroda3

Send message
Joined: 4 Jan 13
Posts: 6
Credit: 4,594,440
RAC: 0
Message 1711 - Posted: 8 Sep 2013, 2:12:43 UTC
The same HTTP permanent error.
No WU for application v.101.00 was downloaded - all generate errors.
But the application itself was downloaded ok.

Asteroids@home 08-09-2013 06:57 Started download of period_search_10100_windows_intelx86__sse3.exe
Asteroids@home 08-09-2013 06:57 Finished download of period_search_10100_windows_intelx86__sse3.exe
Asteroids@home 08-09-2013 07:02 Started download of period_search_10100_windows_intelx86__sse2.exe
Asteroids@home 08-09-2013 07:03 Finished download of period_search_10100_windows_intelx86__sse2.exe
Asteroids@home 08-09-2013 07:02 Giving up on download of input_18180_68: permanent HTTP error
Asteroids@home 08-09-2013 07:03 Giving up on download of input_18113_54: permanent HTTP error
Asteroids@home 08-09-2013 07:04 Giving up on download of input_18123_64: permanent HTTP error
Asteroids@home 08-09-2013 07:05 Giving up on download of input_18182_125: permanent HTTP error
Asteroids@home 08-09-2013 08:46 Giving up on download of input_18191_184: permanent HTTP error
Asteroids@home 08-09-2013 08:47 Giving up on download of input_18190_27: permanent HTTP error
Asteroids@home 08-09-2013 08:48 Giving up on download of input_18156_193: permanent HTTP error
ID: 1711 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Avatar

Send message
Joined: 27 Jun 12
Posts: 129
Credit: 62,725,780
RAC: 0
Message 1712 - Posted: 8 Sep 2013, 3:18:44 UTC - in response to Message 1711.  

Last modified: 8 Sep 2013, 3:23:58 UTC
The same HTTP permanent error.
No WU for application v.101.00 was downloaded - all generate errors.
But the application itself was downloaded ok.

Asteroids@home 08-09-2013 06:57 Started download of period_search_10100_windows_intelx86__sse3.exe
Asteroids@home 08-09-2013 06:57 Finished download of period_search_10100_windows_intelx86__sse3.exe
Asteroids@home 08-09-2013 07:02 Started download of period_search_10100_windows_intelx86__sse2.exe
Asteroids@home 08-09-2013 07:03 Finished download of period_search_10100_windows_intelx86__sse2.exe
Asteroids@home 08-09-2013 07:02 Giving up on download of input_18180_68: permanent HTTP error
Asteroids@home 08-09-2013 07:03 Giving up on download of input_18113_54: permanent HTTP error
Asteroids@home 08-09-2013 07:04 Giving up on download of input_18123_64: permanent HTTP error
Asteroids@home 08-09-2013 07:05 Giving up on download of input_18182_125: permanent HTTP error
Asteroids@home 08-09-2013 08:46 Giving up on download of input_18191_184: permanent HTTP error
Asteroids@home 08-09-2013 08:47 Giving up on download of input_18190_27: permanent HTTP error
Asteroids@home 08-09-2013 08:48 Giving up on download of input_18156_193: permanent HTTP error

That's the issue we've been having ever since the server crash. The server has a heap of work on the database but the actual files aren't there because of the server crash.

Of the few work units that I have managed to download the 101 app is about one third faster than the 100 app (both sse3 versions).
BOINC blog
ID: 1712 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r
Avatar

Send message
Joined: 19 Jun 12
Posts: 21
Credit: 107,293,560
RAC: 0
Message 1716 - Posted: 8 Sep 2013, 9:00:30 UTC - in response to Message 1704.  
New improved versions of SSE2 and SSE3 has been released for 32 and 64bit linux and windows.

Radim Vančo (Kyong)


Is the apps source code available for download somewhere ?


Join BOINC United now!
ID: 1716 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ananas

Send message
Joined: 18 Mar 13
Posts: 32
Credit: 2,506,320
RAC: 0
Message 1717 - Posted: 8 Sep 2013, 10:21:05 UTC - in response to Message 1712.  

Last modified: 8 Sep 2013, 10:21:47 UTC
...

Of the few work units that I have managed to download the 101 app is about one third faster than the 100 app (both sse3 versions).

Less (if any) for the Windows x86 version on C2Q, my Xeon (i7 generation, Windows x64) seems to have gained ~1/4 speed
ID: 1717 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Avatar

Send message
Joined: 27 Jun 12
Posts: 129
Credit: 62,725,780
RAC: 0
Message 1718 - Posted: 8 Sep 2013, 10:39:11 UTC - in response to Message 1712.  
Of the few work units that I have managed to download the 101 app is about one third faster than the 100 app (both sse3 versions).


Now I have a larger sample I would have to say they are taking a little longer. Not by much though. It was about 30-32 mins before, now about 33-35 mins.
BOINC blog
ID: 1718 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile [AF>Libristes] Dilandau

Send message
Joined: 7 Dec 12
Posts: 1
Credit: 5,871,120
RAC: 0
Message 1720 - Posted: 8 Sep 2013, 12:14:35 UTC

Last modified: 8 Sep 2013, 12:17:31 UTC
A little faster.

Xeon E3-1225 v2 (4 Core / 4 Threads):
- v100.00 SSE3 => 1200-1300sec
- v101.00 SSE3 => 1100-1200sec

Xeon E3-1245 v2 (4 Core / 8 Threads):
- v100.00 SSE3 => 2200-2350sec
- v101.00 SSE3 => 2050-2200sec
ID: 1720 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Overtonesinger
Avatar

Send message
Joined: 9 Sep 13
Posts: 23
Credit: 32,670,898
RAC: 216
Message 1725 - Posted: 9 Sep 2013, 7:51:29 UTC - in response to Message 1704.  
SUPER! (mobile CPU i7 720QM, 8 threads at 1.6 GHz)


--- ENG ---
Searching a BOINC-app for my tablet - such that: It has maximum integer computations and minimal FPU computations from all those apps available for BOINC on Android. :-)

Any advice, please?
Thanx.

---------------- Česky: ----------------
Mám dotaz:
Ve výpočtech Asteroids at Home aplikace pro Android: Převažují tam

integer nebo FPU výpočty?

Mám tablet, jehož ARM 1.0 GHz CPU má výrazně rychlejší int. a o 20 procent pomalejší hardwarové FPU (než můj ARM-CPU *TAKÉ* 1.0 GHz, ale v mobilu).

Hledám pro něj tedy vhodnou aplikaci, která by měla co největší podíl integer výpočtů - ze všech dostupných BOINC projektů pro Android... :-)
Děkuji!
ID: 1725 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jan Vaclavik

Send message
Joined: 26 Jan 13
Posts: 31
Credit: 1,550,384
RAC: 213
Message 1726 - Posted: 9 Sep 2013, 12:05:15 UTC
The new apps seems faster on Pentium E2200 and about as fast as the "old" app on Athlon II 240.
ID: 1726 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile HA-SOFT, s.r.o.
Project developer
Project tester

Send message
Joined: 21 Dec 12
Posts: 176
Credit: 136,462,135
RAC: 6
Message 1728 - Posted: 9 Sep 2013, 14:11:39 UTC - in response to Message 1725.  
SUPER! (mobile CPU i7 720QM, 8 threads at 1.6 GHz)


--- ENG ---
Searching a BOINC-app for my tablet - such that: It has maximum integer computations and minimal FPU computations from all those apps available for BOINC on Android. :-)

Any advice, please?
Thanx.

---------------- Česky: ----------------
Mám dotaz:
Ve výpočtech Asteroids at Home aplikace pro Android: Převažují tam

integer nebo FPU výpočty?

Mám tablet, jehož ARM 1.0 GHz CPU má výrazně rychlejší int. a o 20 procent pomalejší hardwarové FPU (než můj ARM-CPU *TAKÉ* 1.0 GHz, ale v mobilu).

Hledám pro něj tedy vhodnou aplikaci, která by měla co největší podíl integer výpočtů - ze všech dostupných BOINC projektů pro Android... :-)
Děkuji!


99% calculations in Period search app uses double precision floating point operations.
ID: 1728 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Overtonesinger
Avatar

Send message
Joined: 9 Sep 13
Posts: 23
Credit: 32,670,898
RAC: 216
Message 1735 - Posted: 10 Sep 2013, 16:02:00 UTC - in response to Message 1728.  
Thank you.

BTW: Do mobile processors ARMv7 (v7 rev.2) have *DOUBLE* precision floating point instructions in their hardware??? I am surprised! :O

- they must have, otherwise it would not be efficient and no sense in making Asteroids app for them.
But it clearly runs on some hardware FPU, because 1 WU on 1 GHz ARMv7.2 CPU (Snapdragon type) is completed in less than 18 hours.
:-)
ID: 1735 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
terencewee*

Send message
Joined: 28 Jul 12
Posts: 17
Credit: 700,427
RAC: 0
Message 1739 - Posted: 11 Sep 2013, 10:36:04 UTC

Last modified: 11 Sep 2013, 10:39:26 UTC
Will there be new compile for Intel MacOSX with SSE2/3 instruction sets?

Thanks!
terencewee*
Sicituradastra.
ID: 1739 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar

Send message
Joined: 9 Jun 12
Posts: 584
Credit: 52,667,664
RAC: 0
Message 1740 - Posted: 11 Sep 2013, 12:51:35 UTC
Yes, when there is no new other versions of SSE, because I have have to send it to someone else for compiling.
ID: 1740 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
terencewee*

Send message
Joined: 28 Jul 12
Posts: 17
Credit: 700,427
RAC: 0
Message 1756 - Posted: 12 Sep 2013, 5:14:08 UTC
Wonderful!

Thanks Radim!

terencewee*
Sicituradastra.
ID: 1756 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Overtonesinger
Avatar

Send message
Joined: 9 Sep 13
Posts: 23
Credit: 32,670,898
RAC: 216
Message 1763 - Posted: 14 Sep 2013, 7:59:02 UTC

Last modified: 14 Sep 2013, 8:06:39 UTC
SSE3 ver. rocks! It completes avg.in:
53 minutes on 3GHz AMD A8-3870K 8GB DDR3 @ 2262 MHz,SSE2 in 60 minutes.
98 min. on 1.6 GHz Intel i7 720QM, SSE2 in 101 min., "Period Search Application v101.10" (no SSE) in 206.48 min.

778.5 min on 1.6 Ghz Intel "Atom N270", "Period Search Application v101.10" (no SSE) in AMAZING 2075 minutes.
(Does anyone from Asteroids-at-home have "INTEL C++ compiler / ICC ver.12" to make an Atom-SSE3 optimized version, please? I cannot find any up-to-date and optimized computations for my 2 Atom devices :) )

Android-apps:
842 min. on HTC One-V, 1 Ghz Qualcomm(Snapdragon) "ARMv7 Processor rev 2 (v7l)"
2072+ min. (computation ERROR at 85 percent!!! WU: http://asteroidsathome.net/boinc/workunit.php?wuid=6094290) on "POINT OF VIEW Mobii ProTab2 IPS 9.7", CPU "Cortex A8 1.2 GHz".

Please, can someone try to make a "Cortex A8 ARMv7 Processor rev 2 (v7l)"-compatible version of the app ? I am afraid it might have no hardware-DOUBLE-PRECISION floating point instruction and it seems to emulate it really badly. But maybe there is some other way around it... ..... like a proper emulation written directly in the code? I dont know if it is possible... :-)[/url]
ID: 1763 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Overtonesinger
Avatar

Send message
Joined: 9 Sep 13
Posts: 23
Credit: 32,670,898
RAC: 216
Message 1766 - Posted: 16 Sep 2013, 5:08:34 UTC

Last modified: 16 Sep 2013, 5:09:34 UTC
OK now. First WU completed on ARMv7 Cortex A8 , in 37.75 hours:
http://asteroidsathome.net/boinc/workunit.php?wuid=6309689
ID: 1766 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jan Vaclavik

Send message
Joined: 26 Jan 13
Posts: 31
Credit: 1,550,384
RAC: 213
Message 1767 - Posted: 16 Sep 2013, 7:40:53 UTC - in response to Message 1763.  
778.5 min on 1.6 Ghz Intel "Atom N270", "Period Search Application v101.10" (no SSE) in AMAZING 2075 minutes.
(Does anyone from Asteroids-at-home have "INTEL C++ compiler / ICC ver.12" to make an Atom-SSE3 optimized version, please? I cannot find any up-to-date and optimized computations for my 2 Atom devices :) )
I don't know why, but my SSE3 PCs downloaded few vanilla units after the new app versions were released. I cancelled those units and the next batch was SSE2 or SSE3.
ID: 1767 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Avatar

Send message
Joined: 27 Jun 12
Posts: 129
Credit: 62,725,780
RAC: 0
Message 1768 - Posted: 16 Sep 2013, 11:50:26 UTC - in response to Message 1767.  

Last modified: 16 Sep 2013, 11:51:46 UTC
I don't know why, but my SSE3 PCs downloaded few vanilla units after the new app versions were released. I cancelled those units and the next batch was SSE2 or SSE3.

The BOINC server tries to work out which app is faster. It does that by trying a number of work units using each app. Once its got an idea how long they take it can then give you work for the fastest app.

I got some vanilla 101 as well as the SSE2 and SSE3 versions when the 101 app was first released. While they can be annoyingly slow (well the vanilla app was) if you just let them run the server will quickly work out what is best for your machine.
BOINC blog
ID: 1768 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · Next

Message boards : News : New improved versions of SSE2 and SSE3 released