Optimized BOINC 5.2.5 core client

Author	Message
michael37 Send message Joined: 23 Jul 99 Posts: 311 Credit: 6,955,447 RAC: 0	Message 184564 - Posted: 31 Oct 2005, 22:49:25 UTC Last modified: 31 Oct 2005, 22:50:50 UTC Greetings, I am starting a new thread here. For now, there is no optimized modern Boinc core client for Linux. Trux and TMR put together a nice work into an optimized Windows client. Links for optimized version 4.72 core clients: TMR built optimized Windows and Linux clients. The Linux SSE and SSE2 client didn't work because they were static, not dynamic. TMR built MMX dynamic client which works, but never updated the SSE/SSE2 ones. Harold built optimized Linux client. This seems to be the fastest boinc for now. ID: 184564 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 184574 - Posted: 31 Oct 2005, 23:08:55 UTC - in response to Message 184564. Greetings, I am starting a new thread here. For now, there is no optimized modern Boinc core client for Linux. Trux and TMR put together a nice work into an optimized Windows client. Links for optimized version 4.72 core clients: TMR built optimized Windows and Linux clients. The Linux SSE and SSE2 client didn't work because they were static, not dynamic. TMR built MMX dynamic client which works, but never updated the SSE/SSE2 ones. Harold built optimized Linux client. This seems to be the fastest boinc for now. I can offer you an Boinc 4.72 for Athlon XP. Here's a benchmark: TMR 4.72 SSE: 2029 double precision MIPS (Whetstone) per CPU 4258 integer MIPS (Dhrystone) per CPU My Build: 2341 double precision MIPS (Whetstone) per CPU 4100 integer MIPS (Dhrystone) per CPU You can download it here Join BOINC United now! ID: 184574 ·

Don Erway Volunteer tester Send message Joined: 18 May 99 Posts: 305 Credit: 471,946 RAC: 0	Message 184578 - Posted: 31 Oct 2005, 23:29:30 UTC The sse2 version for windows does not work for me. I downloaded it, and unzipped it, and copied all the files to the boinc directory, while it was not running. BOINC will not run. I did an uninstall and re-install, and all is goo. What am I missing? The trux version works fine... Don ID: 184578 ·

JohnB175 Volunteer tester Send message Joined: 15 Oct 03 Posts: 124 Credit: 321,769 RAC: 0	Message 184589 - Posted: 31 Oct 2005, 23:54:18 UTC Last modified: 31 Oct 2005, 23:55:02 UTC Yea I would really like a 5.x.x CC for a P3 with MMX, SSE for linux. ID: 184589 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 184613 - Posted: 1 Nov 2005, 1:04:26 UTC - in response to Message 184589. Yea I would really like a 5.x.x CC for a P3 with MMX, SSE for linux. Ok i've got something for you to play with :-) I don't know what they did to the source but benchmarks are lower compared to boinc 4.72. I don't like it :-( But anyway here are the links Linux Boinc 5.3.1 P3 SSE Linux Boinc 5.3.1 Athlon XP SSE Join BOINC United now! ID: 184613 ·

michael37 Send message Joined: 23 Jul 99 Posts: 311 Credit: 6,955,447 RAC: 0	Message 184667 - Posted: 1 Nov 2005, 3:20:50 UTC - in response to Message 184574. I can offer you an Boinc 4.72 for Athlon XP. Here's a benchmark: TMR 4.72 SSE: 2029 double precision MIPS (Whetstone) per CPU 4258 integer MIPS (Dhrystone) per CPU My Build: 2341 double precision MIPS (Whetstone) per CPU 4100 integer MIPS (Dhrystone) per CPU You can download it here Two questions. 1. I am looking for P4/SSE2 version. My guess is that ICC will probably produce the fastest build. 2. How did you get TMR 4.72 SSE to work? It instantly segfaults for me. The aforementioned thread has an extensive discussion by yourself on why his static binaries don't run. ID: 184667 ·

JohnB175 Volunteer tester Send message Joined: 15 Oct 03 Posts: 124 Credit: 321,769 RAC: 0	Message 184671 - Posted: 1 Nov 2005, 3:44:34 UTC - in response to Message 184613. Last modified: 1 Nov 2005, 3:47:45 UTC Ok i've got something for you to play with :-) I don't know what they did to the source but benchmarks are lower compared to boinc 4.72. I don't like it :-( But anyway here are the links Linux Boinc 5.3.1 P3 SSE Linux Boinc 5.3.1 Athlon XP SSE Yikes... well I just tried it... it hosed all my workunits... why is I never backup before stuff like that happens... ugh... oh well...I had to go back to the stock binary... any ideas on why this happened? I got various errors which included: [SETI@home Beta Test] ACTIVE_TASKS::restart_tasks(); missing files [SETI@home Beta Test] Unrecoverable error for result 03oc03aa.9124.881.736090.161_2 (One or more missing files) I got the same messages for the various projects I am attached to... ID: 184671 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 184747 - Posted: 1 Nov 2005, 12:04:56 UTC - in response to Message 184667. Two questions. 1. I am looking for P4/SSE2 version. My guess is that ICC will probably produce the fastest build. Maybe Harold can build an ICC version for you. 2. How did you get TMR 4.72 SSE to work? It instantly segfaults for me. The aforementioned thread has an extensive discussion by yourself on why his static binaries don't run. I don't know what tmr did that the binary segfaults on your system.On my Athlon XP I could at least run the benchmarks but the client is unable to connect that's due to the static linking. Join BOINC United now! ID: 184747 ·

JohnB175 Volunteer tester Send message Joined: 15 Oct 03 Posts: 124 Credit: 321,769 RAC: 0	Message 184771 - Posted: 1 Nov 2005, 13:14:06 UTC By the way what are the (2) files in your release? boinc_client and boinc_cmd? ID: 184771 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 184822 - Posted: 1 Nov 2005, 16:07:23 UTC - in response to Message 184771. By the way what are the (2) files in your release? boinc_client and boinc_cmd? Take a look here http://boinc.berkeley.edu/client_unix.php This page got all the answers :-) Join BOINC United now! ID: 184822 ·

Charles Dennett Send message Joined: 29 Apr 00 Posts: 27 Credit: 18,785 RAC: 0	Message 184833 - Posted: 1 Nov 2005, 16:38:54 UTC - in response to Message 184613. Last modified: 1 Nov 2005, 16:39:55 UTC Yea I would really like a 5.x.x CC for a P3 with MMX, SSE for linux. Ok i've got something for you to play with :-) I don't know what they did to the source but benchmarks are lower compared to boinc 4.72. I don't like it :-( I compiled 5.2.5 the other day for my Linux box (AMD XP2600+) and noticed the drop in the benchmark numbers from 4.72 which I had been running previosly. I found this message: http://setiathome.berkeley.edu/forum_thread.php?id=21112#177597 which mentions removing a PICFLAGS setting from the client's Makefile.in. Did that, reran configure and then the make and the benchmark numbers jumped back up to near where they were for 4.72. Hope this helps. ID: 184833 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 184835 - Posted: 1 Nov 2005, 16:41:59 UTC - in response to Message 184833. Last modified: 1 Nov 2005, 16:45:51 UTC Yea I would really like a 5.x.x CC for a P3 with MMX, SSE for linux. Ok i've got something for you to play with :-) I don't know what they did to the source but benchmarks are lower compared to boinc 4.72. I don't like it :-( I compiled 5.2.5 the other day for my Linux box (AMD XP2600+) and noticed the drop in the benchmark numbers from 4.72 which I had been running previosly. I found this message: http://setiathome.berkeley.edu/forum_thread.php?id=21112#177597 which mentions removing a PICFLAGS setting from the client's Makefile.in. Did that, reran configure and then the make and the benchmark numbers jumped back up to near where they were for 4.72. Hope this helps. I'll take a look at it. Thanks !!! Join BOINC United now! ID: 184835 ·

JohnB175 Volunteer tester Send message Joined: 15 Oct 03 Posts: 124 Credit: 321,769 RAC: 0	Message 184838 - Posted: 1 Nov 2005, 17:01:16 UTC Any idea as to why running ./boinc_client from my boinc directory gave those error messages and lost workunits for all the projects I was attached too? I hope to get this resolved so I can run the optimized client. ID: 184838 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 184858 - Posted: 1 Nov 2005, 18:14:21 UTC - in response to Message 184838. Last modified: 1 Nov 2005, 18:17:19 UTC Any idea as to why running ./boinc_client from my boinc directory gave those error messages and lost workunits for all the projects I was attached too? I hope to get this resolved so I can run the optimized client. I'm looking at it atm. Join BOINC United now! ID: 184858 ·

JohnB175 Volunteer tester Send message Joined: 15 Oct 03 Posts: 124 Credit: 321,769 RAC: 0	Message 185411 - Posted: 3 Nov 2005, 16:27:22 UTC - in response to Message 184858. I'm looking at it atm. Any luck? ID: 185411 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 185531 - Posted: 3 Nov 2005, 22:31:47 UTC - in response to Message 185411. I'm looking at it atm. Any luck? Haven't had enough time to look into that any further. Besides i think there's no need of using a boinc 5.x.x version. I'll stick with 4.19 and 4.72 Join BOINC United now! ID: 185531 ·

Mark Day Send message Joined: 19 Aug 02 Posts: 81 Credit: 502,830 RAC: 0	Message 185636 - Posted: 4 Nov 2005, 5:46:12 UTC As was mentioned in another thread, I was not able to use -malign-double with any of the 5.x boinc sources (JohnB175 just got bit by this lol) Crunch3r's 4.19 static client was the fastest on Athlon XP, giving some really inflated numbers (over 6000 in Dhrystone). Using the PICFLAGS mod mentioned, I was able to milk this out of 5.3.1... 2005-11-04 00:37:16 [---] 2288 double precision MIPS (Whetstone) per CPU 2005-11-04 00:37:16 [---] 4123 integer MIPS (Dhrystone) per CPU Which is in line with what Crunch3r was getting from his 4.72 clients. M ID: 185636 ·

JohnB175 Volunteer tester Send message Joined: 15 Oct 03 Posts: 124 Credit: 321,769 RAC: 0	Message 185681 - Posted: 4 Nov 2005, 12:53:15 UTC I just wish I could get my hands on a linux 5.x.x CC for a P3 SSE :) Someone has to have one out there. ID: 185681 ·

Crunch3r Volunteer tester Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0	Message 185714 - Posted: 4 Nov 2005, 15:41:56 UTC - in response to Message 185681. I just wish I could get my hands on a linux 5.x.x CC for a P3 SSE :) Someone has to have one out there. Why ? There's no special thing about 5.x.x. Join BOINC United now! ID: 185714 ·

Ingleside Volunteer developer Send message Joined: 4 Feb 03 Posts: 1546 Credit: 15,832,022 RAC: 13	Message 185724 - Posted: 4 Nov 2005, 16:22:41 UTC - in response to Message 185714. Last modified: 4 Nov 2005, 16:23:31 UTC Why ? There's no special thing about 5.x.x. Well, for one thing it's easier to join projects, since don't need to use a more or less random key to "attach". Also, another new feature in v5 is the Result duration correction factor, so actually gets much closer to that you asks for. This didn't work correctly in v4.7x Not to forget, Seti_Enhanced paired with v5.2.6 counts flops, meaning claimed credit isn't based on the imperfect benchmark at all... There's also numerous bug-fixes, probably the most important is the download-bug in v4.45 and v4.7x is fixed. ID: 185724 ·

©2026 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.