CompuServe Thread

#3ds rendering stall

18 messages in this thread
#169172From: STEVE FENTONMay 9, 1995 1:09 PM
I have 3DS version 4.0 running on a Gateway P5-100. Sometimes during rendering, the system stalls while transforming objects. The system will allow me to escape out of the rendering process, and then restart from the point it stalled. It will then run fine for various numbers of frames before stalling again. I had similar problems due to a power saver in the autoexec.bat, and due to 32 bit disk access, but I fixed both of those problems. HELP!! Its tough to babysit the machine for 30 hours of rendering.
#169180From: Brian Rudolph [Adesk]May 9, 1995 1:34 PM
Steve, You most likely have POWER.EXE in your config files. Remove that line and everything should be fine. -Brian
#169187From: STEVE FENTONMay 9, 1995 2:02 PM
That problem was a couple of months ago. That command was removed from my config.sys file. In my earlier message I reffered to that problem in my autoexec.bat, but I was mistaken, it was the Power.exe command in my config.sys that I fixed earlier. Thanks for the try. Any more ideas?
#169416From: Brian Rudolph [Adesk]May 10, 1995 11:29 AM
<<Any more ideas?>> If the stall is happening durng the tranforming objects stage, the only reported problems are with POWER.EXE or other power saving features. Is the stall still happening at that stage in the rendering? The other most common problems would be memory conflicts. Have you tried to diagnose for those? -Brian
#169457From: STEVE FENTONMay 10, 1995 2:19 PM
The stall only occurs during transforming objects, and it occurs randomly, unlike the problems we previously encountered with POWER.EXE. It will render correctly for hours or days, and then start experiencing problems once every hour or two. However, when we start rendering at the frame it stuck on, it never sticks on that frame a second time. We have had numerous problems with our system not just with 3DS, and although I have checked many things I'm not sure what memory conflict checks I should, or have checked.
#170969From: Brian Rudolph [Adesk]May 19, 1995 2:37 PM
<<The stall only occurs during transforming objects, and it occurs randomly, unlike the problems we previously encountered with POWER.EXE. It will render correctly for hours or days, and then start experiencing problems once every hour or two. However, when we start rendering at the frame it stuck on, it never sticks on that frame a second time. We have had numerous problems with our system not just with 3DS, and although I have checked many things I'm not sure what memory conflict checks I should, or have checked. >> Steve, sorry, I've been out of the office for a few weeks unexpectedly. Did you solve this problem? -Brian
#171491From: Charles DixonMay 23, 1995 1:54 PM
>> <<The stall only occurs during transforming objects, and it occurs randomly, unlike the problems we previously encountered with POWER.EXE. It will render correctly for hours or days, and then start experiencing problems once every hour or two. However, when we start rendering at the frame it stuck on, it never sticks on that frame a second time. We have had numerous problems with our system not just with 3DS, and although I have checked many things I'm not sure what memory conflict checks I should, or have checked. >> Steve, sorry, I've been out of the office for a few weeks unexpectedly. Did you solve this problem? -Brian << Humble request to attach to your thread? I have a similar problem with 3DS v3 hanging during rendering animation. The problem is intermittent, aside from usually happening at the worst possible time, like when it was on the last frame of a 7 hour render, I had just called the client so say all O.K., and it hung 🙁 … I/ve kept the animations pretty small (60 to 120 frames) because of these problems, which besides being costly, is a pain. I'm running a 486DX66, 36mb RAM, QEMM 7.04, DOS 3.2, Stacker 3.0. Video card ('cheap', but O.K.) is a Spea V7 Mirage (S3 chip). Free disk space varies, but try to keep 3x the estimated .flc file size free (on the destination drive for the flic) to allow for temp files. Any hints, help or otherwise most appreciated. Cheers, Chas -Charles Dixon
#171628From: Brian Rudolph [Adesk]May 24, 1995 9:53 AM
<< I'm running a 486DX66, 36mb RAM, QEMM 7.04, DOS 3.2, Stacker 3.0. Video card ('cheap', but O.K.) is a Spea V7 Mirage (S3 chip). Free disk space varies, but try to keep 3x the estimated .flc file size free (on the destination drive for the flic) to allow for temp files.>> Depending on various paramaters in the animation, and how much RAM you have. 3 times the final size of the FLC might not be enough. If you are using HIGH or medium palette modes, you will probably need more hard disk space than that. Try rendering to LOW or CUSTOM palette. When you freeze like this, what does your status say about paging to disk? Can you render the exact same scenes at a lower resolution? Or if you turn off mapping? These would indicate you are running out of disk space. Also, we don't recommend the use of Stacker or other disk compression programs. While many people use them with no problems, there have been reports that these can sometimes get in the way of Phar Lap writting the temporary files. -Brian
#171873From: Charles DixonMay 25, 1995 2:24 PM
Brian, Thanks for the quick response re: animation hangs >> << I'm running a 486DX66, 36mb RAM, QEMM 7.04, DOS 3.2, Stacker 3.0…. >> >>Depending on various parameters in the animation, and how much RAM you have. 3 times the final size of the FLC might not be enough. If you are using HIGH or medium palette modes, you will probably need more hard disk space than that. Try rendering to LOW or CUSTOM palette. << I usually use low palette. >>When you freeze like this, what does your status say about paging to disk? Can you render the exact same scenes at a lower resolution? Or if you turn off mapping? These would indicate you are running out of disk space.<< I use lots of maps, especially reflection. Running 32MB RAM, the machine still often pages 20MB to disk (I know, get more RAM… <g>) >>Also, we don't recommend the use of Stacker or other disk compression programs. While many people use them with no problems, there have been reports that these can sometimes get in the way of Phar Lap writing the temporary files. << I was not aware Stacker could interfere with the temp files. I have had problems with Stacker in the past when using Stacker's 'de-fragment' utility to de-frag a Stacked drive — like corrupting *.prj files I really needed. Give Stacker its due though, it hasn't harmed old *.prj files I don't care about — they are safe as houses…. I recently installed a SyQuest 270mb removable drive to my system — _not_ Stacked. Perhaps using this as the target disk for rendering — for the temp dir too? (although I am wary of that option since the "drive" is not always there/installed, ie. removed). Thanks again for the advise, Cheers, Chas -Charles Dixon
#171973From: Brian Rudolph [Adesk]May 26, 1995 9:58 AM
<< I recently installed a SyQuest 270mb removable drive to my system — _not_ Stacked. Perhaps using this as the target disk for rendering — for the temp dir too? (although I am wary of that option since the "drive" is not always there/installed, ie. removed).>> That would certaintly remove some of the strain on your hard disk. Especially if you use a new cartridge. remember, though, that the temp files the FLC creates are not the only temp files written. Pharlap writes files to the drive specified in the CFIG386 settings. -Brian
#172048From: Charles DixonMay 26, 1995 3:41 PM
Brian, >> That would certainly remove some of the strain on your hard disk. Especially if you use a new cartridge. remember, though, that the temp files the FLC creates are not the only temp files written. Pharlap writes files to the drive specified in the CFIG386 settings. << Thanks — that is a useful piece of information. I'm aware of the temp file location setting in CFIG386. When you referred to the temp file conflict with Stacker, I thought these temp files were the ones you were referring to. Your info means I can maintain the temp dir location set in CFIG386, i.e. on the permanent drive, and safely render to the Sysquest drive, to which I assume, Pharlap will automatically send the flc related temp files, irrespective of the CFIG386 setting. By the way, this problem has been a serious headache for a long time. Your help is really appreciated — I wish I had asked sooner! Cheers, Chas ps: If you are ever in Newcastle upon Tyne I'll buy you pint! ;-} -Charles Dixon
#173142From: PETER CASANOJun 1, 1995 5:30 PM
I had a similiar problem using Personal Netware. The Map paths embedded in the project to be rendered pointed to drives that didn't exist in the context of the slave. When I removed these, it worked.
#173460From: Charles DixonJun 3, 1995 11:56 AM
>> I had a similar problem using Personal Netware. The Map paths embedded in the project to be rendered pointed to drives that didn't exist in the context of the slave. When I removed these, it worked. << What a timely reply particularly re: >> pointed to drives that didn't exist <<!! Brian Rudolph's [Adesk] reply/advice about the locations of the different types of temp files/locations lead me write a short COM program earlier this week for determining if a drive is available. It returns an errorlevel code, which a batch file can then use to help select the appropriate *.SET file; ie., if the drive is available, use the SET file which uses that drive, if not then use an alternative SET file. I've started using it with a 3DS SET file to send related TEMP files to the removable drive when it is there, an more importantly, not sending TEMP files there when it is not there ;-). Whether the notion of sending the TEMP files to the removable drive _actually_ works is another matter… The program, ISDISK.COM, has only been tested on my machine. So far it has worked with a CDROM, SYSQUEST removable drive, and the floppy drives. It will only run in DOS due to its use of DOS interrupts, however running it from a DOS window in Windows works, too, of course. If anyone in these environs think they could use it, I will load it along with the C source code onto one of the ASOFT libraries for either free usage or an all-expenses-paid trip to somewhere sunny and warm 😉 Cheers, Chas -Charles Dixon
#172611From: STEVE FENTONMay 30, 1995 12:28 PM
I have tried several fixes since our last discusion, but with no luck. As a matter of fact, I spent last weekend trying to pin down the problem on a file which the problem is happening repeatedly on. I have 32MB of ram, and only 10mb is being used for this animation. Thus, no swapping is occuring. I am rendering to a Micropolis hard drive using a par board. I thought it might be a problem with the micropolis, so I tried sending the files to my normal hard drive. Same problem. I also tried changing the block limit settings on the PAR board, and that seemed to fix it for a while, but then back to the same old thing. The PAR book indicated that this could cause failure, so I am continuing to play with the settings. However, I haven't had any luck so far. Luckily our Memorial Day Weekend weather was terrible, so I've had plenty of time to struggle with this one. Help if you can,
#172631From: Brian Rudolph [Adesk]May 30, 1995 2:11 PM
<< I am rendering to a Micropolis hard drive using a par board. I thought it might be a problem with the micropolis, so I tried sending the files to my normal hard drive. Same problem. I also tried changing the block limit settings on the PAR board, and that seemed to fix it for a while, but then back to the same old thing. >> When you were rendering to the regular hard drive, were the drivers for the PAR board still loaded? I would try booting without any of the drivers, and render to the regular hard drive. See if that solves anything. Also, what about a clean boot (FILES and BUFFERS only) rendering to the regular hard drive. Does that still have the problem? Also, does your computer/motherboard have any powersaving features that are built into the BIOS? Is it possible there is still a "green" feature that has not been turned off? -brian
#172831From: STEVE FENTONMay 31, 1995 12:00 PM
I think you isolated the problem. After your messeage, I exited and the re-entered 3ds normally. However, I did not load the PAR PXP routine as I normally would. I then rendered TGA's to my hard drive. Seven hours later, all 280 frames rendered without a hitch. I then loaded the PAR PXP routine and rendered a known trouble spot to my hard drive. No glithes? I then rendered the same section to the par hard drive. It stalled after 2 frames, confirming that last nights rendering success was not a fluke. Any ideas on where to go from here? At least I feel like I'm gaining now, THANKS. -steve
#172880From: Brian Rudolph [Adesk]May 31, 1995 5:59 PM
<< Any ideas on where to go from here? At least I feel like I'm gaining now, THANKS.>> Not really, I don't use that IPAS routine. I know the IPAS routines that Greg Pyros wrote for the PAR are very stable. I've never had any problems using those. -Brian
#173396From: Lydell AndersonJun 2, 1995 9:20 PM
PMJI– but I have a similar question to the one answered by this post: >> Not quite. Whatever you have set for -swapdir in your CFIG386 settings will be where Phar Lap writes those temporary files. Then there are another set of temporary files that you can direct the location. This is the TEMP= variable in the 3DS.SET file. << I recently tried to go back to rendering a small to medium size (300frame) *.flc animation on my revised system. I have only 40mb free on the 3DS3.exe disk, but I thought I had redirected all my swap files to my g:\ drive which has 2+ gigs free. However, twice I have gotten to approx frame 225 in render and get a "Error writing file—not enough space" message. When I say that I have redirected my swap files I mean that I used cfig386 -swapdir command to make g:\swap my swap directory, and the 3ds.set file has been modified to make g:\3dstemp my pharlap temp file location (even though animation was not complex enough to swap out of RAM). Is there another file that is paged to disk that I have overlooked? I thought I had this all down a year ago, but I think I am forgetting something this time around. In case it matters, I was using Video Post render with Stars.PXP and compositing a *.gif and a *.flc under the keyframer geometry. I keep thinking it was the 40 mb on the 3ds3 disk that is the problem, but I can't remember anything that needs to be redirected—it was too early for the palette optimization process, not sure that could be assigned to the g: drive anyway.. I have "last image" save turned off in 3ds.set… At least the second time I was offered the option of saving the partial flic file. Thanks. Lydell