Program Closes Computing DepthMap

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • dtmcnamara
    3Dfollower
    • Aug 2015
    • 11

    Program Closes Computing DepthMap

    Im having a problem with Zephyr Pro v2.0 closing when it starts to compute depthmaps on High settings. My RAM usage at the time is only at 25% so I am not sure why this is happening. When looking at the log files there does not seem to be any clues to why it shuts down. Below are the last few lines of the LOG file, any help would be awesome.

    [09:36:51] Starting 3DF Stasia
    [09:36:51] Selected CUDA Device(s):
    [09:36:51] Tesla M2090
    [09:36:51] ClockRate: 1301000 Khz - Memory: 6143 Mb - Cuda Capabilities: 2.0
    [09:36:51] GeForce GTX 780
    [09:36:51] ClockRate: 993000 Khz - Memory: 3072 Mb - Cuda Capabilities: 3.5
    [09:36:51] GeForce GTX 780
    [09:36:51] ClockRate: 993000 Khz - Memory: 3072 Mb - Cuda Capabilities: 3.5
    [09:36:51] Prepare PPMs...
    [09:36:51] Prepare point data...
    [09:36:51] Prepare bounding box...
    [09:36:51] [GPU 0] Computing DepthMap for IMG_0977.JPG
    [09:36:51] [GPU 2] Computing DepthMap for IMG_0975.JPG
    [09:36:51] [GPU 1] Computing DepthMap for IMG_0976.JPG
    [09:36:54] [GPU 1] MrfInitData [(C) 2501 MB] [12420 MB]
    [09:36:54] [GPU 1] MrfInitData [(C) 2501 MB] [12428 MB] -> 1s
    [09:36:54] [GPU 1] MrfInitImages [(C) 2501 MB] [12428 MB]
    [09:36:54] [GPU 1] MrfInitImages [(C) 2496 MB] [12428 MB] -> 1s
    [09:36:54] [GPU 1] MrfDepthRange [(C) 2496 MB] [12428 MB]
    [09:36:55] [GPU 2] MrfInitData [(C) 2382 MB] [12462 MB]
    [09:36:55] [GPU 2] MrfInitData [(C) 2382 MB] [12479 MB] -> 1s
    [09:36:55] [GPU 2] MrfInitImages [(C) 2382 MB] [12479 MB]
    [09:36:55] [GPU 2] MrfInitImages [(C) 2378 MB] [12479 MB] -> 1s
    [09:36:55] [GPU 2] MrfDepthRange [(C) 2378 MB] [12479 MB]
    [09:36:55] [GPU 0] MrfInitData [(C) 6076 MB] [12513 MB]
    [09:36:55] [GPU 0] MrfInitData [(C) 6076 MB] [12505 MB] -> 1s
    [09:36:55] [GPU 0] MrfInitImages [(C) 6076 MB] [12505 MB]
    [09:36:55] [GPU 0] MrfInitImages [(C) 6072 MB] [12505 MB] -> 1s
    [09:36:55] [GPU 0] MrfDepthRange [(C) 6072 MB] [12505 MB]
    [09:36:58] [GPU 0] MrfDepthRange [(C) 6072 MB] [12525 MB] -> 4s
    [09:36:58] [GPU 0] AllRegions (2, 2) [300x300] [(C) 6072 MB] [12521 MB]
    [09:36:58] [GPU 0] AllRegions (2, 2) [300x300] [(C) 6072 MB] [12521 MB] -> 0s
    [09:36:58] [GPU 0] MrfAllRegions [(C) 6072 MB] [12521 MB]
    [09:36:58] [GPU 0] MrfAllRegions [(C) 6027 MB] [12401 MB] -> 1s
    [09:36:58] [GPU 0] Refining DepthMap.
  • Andrea Alessi
    3Dflow Staff
    • Oct 2013
    • 1335

    #2
    Hello dtmcnamara,

    I think Roberto already replied you privately, sorry if this is redundant.

    With Zephyr 2.0 we upgraded to Cuda7, so my guess is that your drivers might not be the latest available. We added a driver version check but i'm guessing it's some kind of driver version check issue with your multi gpu rig (i'm very jelly about it btw)

    Could you please tell me your specific driver version, and if it's not the latest available, try to update it?

    If it's the latest avilable, maybe try installing the cuda toolkit (it includes developer drivers) as a test

    Comment

    • dtmcnamara
      3Dfollower
      • Aug 2015
      • 11

      #3
      Yes, it looks like the problem is with the Tesla card that I run. When I disable the card and run just the GTX cards everything works just fine. I will play around with the CUDA toolkit later this weekend and see if I can fix the problem, Im really interested in seeing how well the Zephyr software works with the co-processors.

      Thanks for the reply

      Comment

      • Till
        Blossoming 3Dflower
        • Sep 2015
        • 1

        #4
        Hi,

        I'm working with a GeForce GTX 570 and recently upgraded to Zephyr Pro 2.0. I have the same Cuda problem. After much installation trouble two weeks ago, my graphics driver was finally recognized as up to date (version 347.62). But Cuda made Zephyr crash anyway. So as suggested by 3DF I upgraded Cuda:

        Selected Cuda driver components:
        - Cuda Toolkit (V. 7.0 -> 7.0) (yes, installed this two week ago already, didn't help)
        - GPU Deployment Kit (n/a -> 347.62)
        ---> Graphics driver (355.82 -> 347.62) FORCED DOWNGRADE!

        Not selected after trying:
        - Cuda Visual Studio Integration (n/a -> 7.0)
        => Cuda Toolkit wants Visual Studio.
        - Installed VS Code (quite small) --> "no supported version found"
        - VS Community -> 7GB that I don't need? NO WAY!

        I really hope that making Zephyr work does not mean that I need to install 7GB of unwanted stuff on my computer!

        So after the Cuda installer downgraded my graphics driver Zephyr again tells me at start-up that a compatible graphics card was found on my system but the drivers are not updated and computation will run in CPU mode and can be considerably slower.

        Good job. I'd rather go back to the previous Cuda version, at least it worked! I'm wasting an ungodly amount of time on this "upgrade".

        Comment

        • Andrea Alessi
          3Dflow Staff
          • Oct 2013
          • 1335

          #5
          Hello Till,

          sorry to hear about your issue.

          I can tell for sure this is not related to your installation problem of two weeks ago (as you might recall, your license was expired and we had to manually renew it).

          You don't need to install VS community, that's for sure and you can at anytime fall back to 1.6 and Cuda 6.5: if you wish to do so, old installers are available at the usual download URL.

          I'll try to find somebody willing to borrow me a GTX570 or similar generation card since it's a pretty old one and there might be some driver issues that we need to investigate. Not saying that's the reason, but it might be worth checking it, i'll followup shortly!

          Andrea

          Comment

          • Andrea Alessi
            3Dflow Staff
            • Oct 2013
            • 1335

            #6
            I can confirm that we isolated the issue.

            The issue happens when running CUDA 7 on older cards (Capabilities 2.0).

            We are working on a fix which will be available in the next update, scheduled for next week!

            Andrea

            Comment

            • Roberto
              3Dflow
              • Jun 2011
              • 559

              #7
              We just released a new version (2.050) with the fix for the crash.
              The old Nvidia cards now works properly, but please note that due to their low Cuda capabilities, some specific parts of the multiview stereo pipeline will run on cpu.

              Comment

              Working...