Guest xionmark Posted May 10, 2006 Share Posted May 10, 2006 Hi there, I'm getting a strange error when trying to bake out some geo via a geometry ROP. Most frames of the sequence finish fine, memory usage averages at about 1.7G, with 2 or 3 frames peaking at 3.4G but they finish. There are some 4-5 frames that almost always fail with the following error from hscript: > /obj/geo_all/ropnet -> render geometry1 Render failed: Error: No OP found or specified: geometry1 Of course there is a "geometry1" geometry ROP so the error message appears to be unrelated to the actual error that occured. I reduced the amount of geo to be baked out by the geo ROP by half but still get the error. While hscript is running, top reports that it's max mem useage at about 2.3G just before tha failure (not sure how much to trust top). H8.0.474, RH4. Any ideas? TIA! --Mark Quote Link to comment Share on other sites More sharing options...
michael Posted May 10, 2006 Share Posted May 10, 2006 I'm assuming here that you are rendering on a farm: any time that has ever happend to me it was caused by something feeding the rop, not the rop itself...often the network will cook just fine interactively but fail on the farm... I'd try writing out the geometry for one or two of those frames inside a session...you might find the bad node higher up in the network... HTH Quote Link to comment Share on other sites More sharing options...
Guest xionmark Posted May 11, 2006 Share Posted May 11, 2006 I'm assuming here that you are rendering on a farm:any time that has ever happend to me it was caused by something feeding the rop, not the rop itself...often the network will cook just fine interactively but fail on the farm... I'd try writing out the geometry for one or two of those frames inside a session...you might find the bad node higher up in the network... HTH 27456[/snapback] Hi there, There's no errors coming from upstream to the ROP, I can set the frame to anywhere in the sequence and Houdini will (eventually) cook the net and display in the viewport. I can then save the geo right from the node that feeds the ROP. So why the geometry ROP fails is a mystery. You're right, it fails more often on the farm, but will almost always fail on my workstation when executing the ROP from Houdini or hscript. I suppose I'll just have to divide the geo up even more to write out seprate sequnces of files, but I'm wondering what the difference is between doing a "Save Geometry" and writing the geo out with a geo ROP. The hunt continues ... --Mark Quote Link to comment Share on other sites More sharing options...
michael Posted May 11, 2006 Share Posted May 11, 2006 hmmmm...tricky not knowing what your network looks like...one thing that might help would be to write out the network at logical stages through the network...just to give houdini a break once in a while... it's annoying when there is no obvous problem... Quote Link to comment Share on other sites More sharing options...
Jason Posted May 11, 2006 Share Posted May 11, 2006 Your only true friend in these situations is the Performance Monitor. Evaluation errors (like recursion errors) are printed out in here. The SOP caching can make some expressions fail on the first cook but when you MMB on the tile, the recook often goes through and the error disappears. I remember a time when the MMB info was changed to not cook but it caused consternation amongst the masses (for good reason) Quote Link to comment Share on other sites More sharing options...
michael Posted May 11, 2006 Share Posted May 11, 2006 good points jason... I think we still have a RMB 'Force Cook' script around...power to the people!! Quote Link to comment Share on other sites More sharing options...
Guest xionmark Posted May 12, 2006 Share Posted May 12, 2006 Your only true friend in these situations is the Performance Monitor. Evaluation errors (like recursion errors) are printed out in here. The SOP caching can make some expressions fail on the first cook but when you MMB on the tile, the recook often goes through and the error disappears.I remember a time when the MMB info was changed to not cook but it caused consternation amongst the masses (for good reason) 27482[/snapback] Hi there Jason! Yea, the perf monitor is certainly my friend! It did reveal some areas of inefficiency in my networks (mainly caching issues) but unfortunately did not reveal the problem with the ROP failing. I took a very detailed look at it, culled geo down until I could find a "breaking point" where it would all start to work as normal but found that is was very hard to reproduce. Very strange indeed. Anway, still have more to dig into, I'm sure we'll figure it out soon ......... Thanks for the reminder about the perf meter! --Mark Quote Link to comment Share on other sites More sharing options...
Guest xionmark Posted May 16, 2006 Share Posted May 16, 2006 Your only true friend in these situations is the Performance Monitor. Yea baby! I found the problem via the perf monitor. There was a cache SOP that was bypassed and only influenced the SOP chain under rare circumstances, which I only found by digging through the oodles of detail pumped out by the perf mon. Yahooo! Thanks Jason! --Mark Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.