![]() |
Stephen Brooks 2002-10-28 04:59:19 | I've heard some of you have recompiled the Linux Muon1 with optimisation for various processors. If there's a version out there that a few people have tried and appears to stable, please tell me so I can put a link to it on the main page (also tell me what restrictions on processor etc. there are for it). "As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline" |
DrHanser 2002-10-28 05:37:09 | Hrm. I should do this for the Windows client as well. The mark of an educated man is one who knows a lot about something, and a little about everything. |
[Ars]SmilingJim 2002-10-28 06:00:26 | I did re-port the linux version back to win32. ![]() |
Me@Home 2002-10-28 07:34:42 | Is there a windoze optimized client? |
[DPC] Jive 2002-10-28 07:49:35 | Maybee we should make a page with all the download versions ... linux/windows/architecture I would think that if people could choose, they would use an optimised client. (i could look into hosting some of the files for u since they don't change often). |
DrHanser 2002-10-28 08:00:19 | quote: ugh, a Windows-using, Lunix-wannabe-user. As for your question, there is no optimized Windows client at the moment, except for the one that SmilingJim has, which has not been released yet. -- The mark of an educated man is one who knows a lot about something, and a little about everything. |
[Ars]SmilingJim 2002-10-28 08:42:05 | Not sure my should be that "official" all i did was took the code from the linuxmuon and used djgpp (gcc for win32) and tweaked/commented out what didn't work. Like i took out all the pthread stuff, so it doesn't handle the thread stuff at all now. I just set it to idle with win2k task manager. And I'm still not sure that it runs any faster, still trying to find a good way to test it. SmilingJim |
Stephen Brooks 2002-10-28 08:54:03 | Here's a nice list of many Linux clients optimised for different machine-architectures:quote: If a few other people can vouch for the stability of these clients (i.e. that one is the same thing as the "very fast AthlonXP" client I've heard of), then I'll post this link and a link to the readme on the page under 7F4's port. "As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline" |
Stephen Brooks 2002-10-28 08:58:21 | quote: In the current development source I've inlined a few frequently-called functions and also have decided to compile "with MMX instructions" which I think means for a 686 target processor. Can't promise when 4.22 will arrive due to workload... I'll try my best. "As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline" |
Me@Home 2002-10-28 09:26:39 | OK. If you optimize for Athlon SSE2 instructions, that would be really cool. Haha (I'm watching Orange County - great show) Me /This movie is great... |
[DPC] Jive 2002-10-28 09:59:22 | MMX optimisation (if applicable to the computing) will certainly enhance the efficiency of the cruncher ... And on a side note ... this crunching isn't really for anything less than 686-mmx anyway ... coz it would take far too long. As for the peeps running the old beasties .... maybee make some versions .. AMD version AMD Athlon version Intel version Intel MMX version Intel SSE version Intel SSE2 version I'll see if i can get u some space to put up a download-mirror ... so u won't get such a big hit on ur bills. |
[ARS]odessit 2002-10-28 10:03:25 | Stephen, the original Linux Athlon client was built by TinyHelp and I posted a news burp on TAM's website (link below) I can vouch for the stability & speed of this client. The newer clients are built by Andy (TheEternalVortex) and I just started running the -Athlon client on my Duron. SO far it works, how fast, I donno. Will have to wait couple of days. BTW, DO read the README file. Also, Athlon does not support the SSE2 ![]() ARS Team Atomic Milkshake Unofficial Muon1 FAQ |
DrHanser 2002-10-28 10:10:38 | Stephen, I'd be willing to host any and all versions of the client at no charge, as a primary mirror, if you like. No bandwidth limits. Just let me know. -- The mark of an educated man is one who knows a lot about something, and a little about everything. |
AySz88 2002-10-28 14:30:00 | Which ( MMX/SSE/SSE2 ) is which (Intel processor)? |
[DPC] Jive 2002-10-28 15:02:13 | mmx = pentium 1 sse = pentium3 sse2 = pentium4 this assume's my memor is correct & intel is consistent (which they aren't for the lower mhz's processors anyway) from memory & in a drunk stage (just got back from teh bar ![]() |
[ARS]odessit 2002-10-28 15:03:54 | Pentium Classic- Nothing Pentium MMX - MMX Pentium II - MMX Pentium III - MMX+, SSE Pentium IV - MMX+, SSE, SSE2 Duron < 1 GHz - MMX+, 3DNow+ Duron => 1 GHz - MMX+, 3DNow+, SSE Athlon - MMX+, 3DNow+ Athlon XP - MMX+, 3DNow+, SSE ARS Team Atomic Milkshake Unofficial Muon1 FAQ |
[Ars]SmilingJim 2002-10-28 15:33:43 | is there a gcc flag for mmx+ over mmx? or 3dnow+ over 3dnow? and if so what are the differences. |
prokaryote 2002-10-29 12:18:49 | Hi SmilingJim, Any luck with the Windoze optimized clients? www.ninjamicros.com mathematical projects |
px3 2002-10-31 10:47:56 | Hi Stephen, i'd recompiled the clients on Solaris 2.8 and on a special Redhat version used on RAQ Cobalt. If you're interested, i can create a pkg/rpm and post the link. Regards Rolf PX3 |
[Ars]SmilingJim 2002-10-31 11:13:06 | Linuxmuon for win32 Here is a link to the source and generic i686 + mmx exe. There are experimental and I'm not sure how must faster they are. If you can improve them at all, let me know. SmilingJim |
Stephen Brooks 2002-11-02 09:46:24 | quote: I'm currently evading that issue nicely ![]() quote: If you could put up a simple HTML page with your e-mail address for support, some basic readme, and links to the downloads, I'll link to it from my site. Best thing is to e-mail me the URL and a brief description of the platform(s) your version is for (I need the latter to get the wording for the link on my site appropriate). "As every 11-year-old kid knows, if you concentrate enough Van-der-Graff generators and expensive special effects in one place, you create a spiral space-time whirly thing, AND an interesting plotline" |
Goner 2002-11-02 12:07:37 | quote: i'm using the athlon-xp client from that site on Mandrake 8.2 with an Athlon XP 1700+ and it runs OK ... i wouldn't know how to benchmark it, but it is fast & stable. Goner |
Goner 2002-11-02 12:43:06 | quote: this one runs OK on Win98SE & XP 1700+ ... your http://web.ics.purdue.edu/~bauerlej/muonathlon.exe did _not_ run ... Goner |
Bill[Romulus2] 2002-11-02 13:00:54 | I got a big no go on the muonathlon.exe as well. Copy/paste of message below. System is an Athlon 1ghz running Win98se. Placed the exe in the muon folder and fired it up, nothing else was done. Loading bending magnet fields... y=0cm... y=5cm... y=10cm... y=13cm... Done Interpreting lattice file...Exiting due to signal SIGILL Invalid Opcode at eip=0000b89d eax=00000000 ebx=00000001 ecx=000adf09 edx=00000002 esi=000adf08 edi=00000004 ebp=000ad178 esp=000acec0 program=C:\MUON\MUONAT~1.EXE cs: sel=00a7 base=8414d000 limit=fff69fff ds: sel=00b7 base=8414d000 limit=fff69fff es: sel=00b7 base=8414d000 limit=fff69fff fs: sel=0087 base=00010250 limit=0000ffff gs: sel=00c7 base=00000000 limit=0010ffff ss: sel=00b7 base=8414d000 limit=fff69fff App stack: [000adc4c..0002dc4c] Exceptn stack: [0002dba8..0002bc68] Call frame traceback EIPs: 0x0000b89d 0x0000cd1d 0x0000d2bb 0x00012625 0x00018d48 |
Goner 2002-11-02 13:16:43 | quote: i got this : Searching for auto-saved file... Restored simulation at 64.5ns Exiting due to signal SIGSEGV General Protection Fault at eip=0000752e eax=00ffffff ebx=00000000 ecx=00000000 edx=000ad490 esi=5fffffa0 edi=000ad490 ebp=000ad088 esp=000acee0 program=D:\UTILIT~1\MUON_A~1\MUONAT~1.EXE cs: sel=00b7 base=85918000 limit=00b0ffff ds: sel=00bf base=85918000 limit=00b0ffff es: sel=00bf base=85918000 limit=00b0ffff fs: sel=008f base=0000f080 limit=0000ffff gs: sel=00cf base=00000000 limit=0010ffff ss: sel=00bf base=85918000 limit=00b0ffff App stack: [000adc4c..0002dc4c] Exceptn stack: [0002dba8..0002bc68] Call frame traceback EIPs: 0x0000752e 0x000078b6 0x00008476 0x00012a98 0x00018d48 btw : i just tried "linuxmuon.exe" and after it finished it's 1st result, there was no results.txt, but results.dat had a new result added ... ?! Goner |
[Ars]SmilingJim 2002-11-02 15:23:28 | Okay, ill see what i can do. Its been working for me for awhile and getting nice results. I'll replace the athlonmuon.exe with the one i am currectly using for my duron. It been up for going on 47 hour on my box. If anyone can figure out what happened. Let me know, I'm not that good at debugging core dumps yet. SmilingJim |
prokaryote 2002-11-02 15:56:26 | Hi SmilingJim, Got your linuxmuon win 32 client working on a W2K system. Originally, I too had problems similar to those described above. Here's how I installed the client to get it to work. To do this, you must have been running the old windows client first, because you're going to need some of the created files. 1. Extract the linuxmuon.zip file to the "Program Files" folder. The download will create a new folder called "linuxmuon". 2. From the old Muon1 directory, copy over to the linuxmuon folder the following files:{"user.txt", "config.txt", "results.dat", "l2y0.txt", "l2y10.txt", "l2y13.txt", "l2y5.txt", "Lattices" folder, "pi_plus_2G2eV.csv"}. Do NOT copy over the autocli.sav file. This is what causes the General Protection Fault error for some reason. You'll have to sacrifice the current simulation that the old Muon1 executable was working on. 3. Now fire up a command line window and run "linuxmuon" from this new folder. Okay, now I don't know if the linuxmuon.exe will create a "results.txt" file or not. Otherwise, you'll have to create one manually from the "results.dat" file. I place a comment in the "results.dat" file to demark newly crunched simulations from the older, already downloaded ones. Thanks, prok www.ninjamicros.com mathematical projects |
John Campbell 2002-11-02 17:22:07 | quote: Was muonathlon.exe built with SSE optimizations? That would cause it to throw this sort of invalid opcode error on the older Athlon Classics and Thunderbirds (a GHz Athlon might be either of those, but is certainly one or the other). SSE support wasn't added until the Palomino (Athlon XP/MP). |
[Ars]SmilingJim 2002-11-02 17:24:10 | Ya mine is on a Win2k system. And I did have some trouble to begin with with the old autocli.sav file. After you delete that I think you should be fine. Source and i686 client Athlon/Duron (Non XP) client And you do need to download the old Command Line and use all the files. I think this is the same for the Linux Client [Edit] I removed the AthlonXP client cause I though it was causing problems and replaced it with the Old Athlon Client. If anyone wants the XP client back, Let me know and I will repost it |
prokaryote 2002-11-02 17:50:42 | Got it to run on an athlon T-bird and on an XP (both with W2K). Also, multi-threading throws errors as well. Just have to run 2 clients on a dually (Tyan 2460 XP1900+ W2K). What about the results.txt file? Thanks www.ninjamicros.com mathematical projects |
[Ars]SmilingJim 2002-11-02 17:59:55 | No Idea on the result.txt. Ive never had the problem before. Ill try to look through the source and see if I can find anything. I never remember commenting out anything about it when i was reporting back to win32. The source is there if you want to play around with it yourself. Let me know if you find anything I can fix. SmilingJim |
prokaryote 2002-11-02 21:56:28 | quote: I'll give it a shot. I've got the linuxmuon win32 going which is the Athlon +mmx and not the athlon XP with SSE, right? Oh, and thanks for all of the work that you've done as well. prok www.ninjamicros.com mathematical projects |