Return-Path: X-Flags: 0001 Delivered-To: GMX delivery to joof@gmx.de Received: (qmail 23525 invoked by uid 0); 29 Jun 1999 20:20:46 -0000 Received: from front6.grolier.fr (194.158.96.56) by mx7.gmx.net with SMTP; 29 Jun 1999 20:20:46 -0000 Received: from club-internet.fr (ppp-163-185.villette.club-internet.fr [195.36.163.185]) by front6.grolier.fr (8.9.3/No_Relay+No_Spam_MGC990224) with ESMTP id WAA03342 for ; Tue, 29 Jun 1999 22:20:29 +0200 (MET DST) Sender: francois@front6.grolier.fr Message-ID: <37792A8E.F9150FF1@club-internet.fr> Date: Tue, 29 Jun 1999 22:20:30 +0200 From: Francois Dispot Organization: Landfelder research X-Mailer: Mozilla 4.07 [en] (X11; I; Linux 2.0.36 i586) MIME-Version: 1.0 To: Flierl Jakob Subject: Re: pvmpov References: <26534.930673942@www4.gmx.net> Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Notify-Sent-To: 9435576@Scall.de X-Mozilla-Status2: 00000000 Flierl Jakob wrote: > > Some days ago you wrote: > > > I compare it with the previous release (for 3.0x): > > - When SIGINTed (ctrl-C), it exits without displaying anything, and the > > slave tasks remain alive > I´ve taken a look into the source, but I cannot find that bug. Can you > help? Well, let's try :-) I have 2 machines; for Berg's sake, let's call them wozzeck (p200) and doktor (dual PII/450). When I start pvm form doktor, I automatically add wozzeck, so I get this: 1 successful HOST DTID wozzeck.localdomain 80000 3.3.11 t40001 2 hosts, 1 data format HOST DTID ARCH SPEED doktor.localdomain 40000 LINUX 1000 wozzeck.localdomain 80000 LINUX 1000 pvm> Then I exit, pvm still runs, and I am ready to continue. Now I start pvmpov with an ini file that contains the following lines: pvm=on pvm_hosts=wozzeck.localdomain,doktor.localdomain,doktor.localdomain,doktor.local domain,doktor.localdomain pvm_slave=x-pvmpov pvm_width=20 pvm_height=20 pvm_tasks=5 pvm_nice=0 The duplicate hosts enable me to run 4 clients on the fast machine, which is the fastest way to go. Persistence of Vision(tm) Ray Tracer Version 3.1e.Linux.gcc This is an unofficial version compiled by: Jakob Flierl - PVM-POV Version 3.1e.1 The POV-Ray Team(tm) is not responsible for supporting this version. Copyright 1999 POV-Ray Team(tm) Initializing PVMPOV Spawning x-pvmpov with 5 PVM tasks on 2 hosts... ...5 PVM tasks successfully spawned. Waiting up to 120s for first slave to start... Slave 0 successfully started. Parsing Options Input file: verre.pov (compatible to version 3.1) Remove bounds.......Off Split unions........Off Library paths: /opt/povray31/include Output Options Image resolution 400 by 300 (rows 1 to 300, columns 1 to 400). Output file: verre.tga, 24 bpp Targa Graphic display......On (type: 0, palette: 3, gamma: 2.2) Mosaic preview......Off CPU usage histogram.Off Continued trace.....Off Allow interruption..Off Pause when done......On Verbose messages.....On Tracing Options Quality: 9 Bounding boxes.......On Bounding threshold: 25 Light Buffer.........On Vista Buffer.........On Draw Vista Buffer...Off Antialiasing.........On (Method 2, Threshold 0.400, Depth 2, Jitter 1.00) Radiosity...........Off Animation Options Clock value.... 0.000 (Animation off) PVM Options Block Width.... 20 Block Height... 20 PVM Tasks...... 5 PVM Nice....... 0 PVM Arch....... PVM Slave...... x-pvmpov PVM WorkingDir. /home/francois/POV/VERRE Pvm Hosts...... wozzeck.localdomain doktor.localdomain doktor.localdomain dokt or.localdomain doktor.localdomain Redirecting Options All Streams to console..........On Debug Stream to console.........On Fatal Stream to console.........On Render Stream to console........On Statistics Stream to console....On Warning Stream to console.......On Starting frame 0... Displaying frame 0... Using 24 bit TrueColor visual... Slave 1 at doktor.localdomain successfully started. Slave 2 at doktor.localdomain successfully started. Slave 3 at doktor.localdomain successfully started. Slave 4 at wozzeck.localdomain successfully started. 3.00 of blocks complete. The calculation is quite long and if I interrupt it, I get no special output, pvmpov exits, but ps gives: 7710 p0 S 0:00 -bash 7728 p0 S 0:00 /usr/pvm3/lib/LINUX/pvmd3 7737 p0 S 0:05 /home/francois/pvm3/bin/LINUX/x-pvmpov -nI/home/francois/POV 7738 p0 S 0:05 /home/francois/pvm3/bin/LINUX/x-pvmpov -nI/home/francois/POV 7739 p0 S 0:05 /home/francois/pvm3/bin/LINUX/x-pvmpov -nI/home/francois/POV 7740 p0 S 0:05 /home/francois/pvm3/bin/LINUX/x-pvmpov -nI/home/francois/POV 7741 p0 R 0:00 ps And on wozzeck, I still have: 974 ? S 0:19 /home/francois/pvm3/bin/LINUX/x-pvmpov -nI/home/francois/POV Note that in the present case, the tasks are stopped, but sometimes they continue to run and to eat time. Now just tell me how I can help. There are so many things I would also like pvmpov to do... but maybe one of the first things I really wonder is why pvmpov is 15-25% slower than "raw" pov 3.1. I mean, it is normal to get some overhead, but why so much? And it is the same on a very complex scene which takes several hours to run, so I hope this should not be network overhead. I do not know how it works, but I find it surprising. Do not hesitate to ask more questions about my ctrl-C problem. Sometimes I also get segfaults during the render, but it is more difficult to repeat. About pvm-pov, I recently contacted the author of povfront, a gtk front-end to povray (a .ini gui in fact), and asked him if he could add a pvm-pov part to his interface, and he seemed to agree. Bye and happy Poving! -- François DISPOT http://www.geocities.com/vienna/7709 _ _ _ _ ______ _____ _____ _____ _____ _ || ||// // // // /___ / /___ / / ___/ / ___/ // || |// // // // // // / /__ / / //__ || // // // // // // / ___/ / / //|__| ||//|// // // //___ //___ / /__ / /__ // || |//|// //___// /____/ /____/ /____/ /____/ // ||