Christian – happy to hear that that took care of the C::Sensor issue. On a different note, this might be too late now, but if you are interested in C::Sensor (and I imagine control/autonomy) applications, we have a meeting here in Madison next month that focuses, among other things, on issues of autonomy and how Chrono simulation can help design autonomy solutions. Here’s the link: https://sbel.wisc.edu/magic-2023/. Moving forward, I’ll add your email to the list of colleagues with whom we advertise these Chrono meetings.
Best of luck with your Chrono project. Dan --------------------------------------------- Bernard A. and Frances M. Weideman Professor NVIDIA CUDA Fellow Department of Mechanical Engineering Department of Computer Science University of Wisconsin - Madison 4150ME, 1513 University Avenue Madison, WI 53706-1572 608 772 0914 http://sbel.wisc.edu/ http://projectchrono.org/ --------------------------------------------- From: [email protected] <[email protected]> On Behalf Of Christian Feller Sent: Saturday, November 18, 2023 6:35 AM To: ProjectChrono <[email protected]> Subject: Re: [chrono] Re: Problems with Chrono OptiX sensors on Ampere GPUs Hi all, I upgraded to develop + required OptiX release and asked the affected colleague to update his GPU Driver (was older than 530). Et voila, everything runs as expected. Interestingly, it seems that the key was the GPU driver. The "old" Chrono 7 code with OptiX 7.2 is also running fine when using most recent NVIDIA driver. Sometimes the simple solution is the best... Note, however, that on Pascal and Turing GPUs everything is running fine with older drivers. On my Pascal even with <= 500 or so. Ampere GPUs seem to be special in that regard (at least the RTX A1000). Thanks, Christian Christian Feller <[email protected]<mailto:[email protected]>> schrieb am Do., 16. Nov. 2023, 09:47: Hi Dan, Hi Jason, thank you both for your answers and support. We are of course aware that our setup is a bit behind current development. But as we have built up our vehicle modeling toolchain with dependencies on Chrono 7.0 base classes, upgrading to develop would probably be some work. Before starting that endeavor, it would be cool to have some feedback such as - a user/developer who can confirm that OptiX sensors work just fine with the above or a similar setup (--> we should look for errors on our end that are independent of Chrono or OptiX release) - a user/developer experiences or has experienced the same and has a potential fix (--> we can stop looking for stuff that we are doing wrong) - a user/developer who points us to a compatibility issue with Ampere GPUs and the above setup that we are not aware of (--> we can stop looking for other errors and go for newer releases) That being said, I will take a stab at upgrading our toolchain to Chrono develop and newer CUDA + OptiX. Thank you, Christian On Wednesday, November 15, 2023 at 9:56:53 PM UTC+1 JASON Z wrote: Please consider upgrading Chrono to either the 8.0 Release or the main branch of the chrono repo. There have been updates since 7.0.0, and we recently upgraded the Optix to 7.7.0 in the main repo. It will be difficult for us to debug on issues happening on previous versions. On Wednesday, November 15, 2023 at 8:58:26 AM UTC-6 Christian Feller wrote: Dear all, we are observing some strange behavior when trying to run lidar sensors from the Chrono::Sensor module on PCs with Ampere GPU (compute capability 8.6). Our setup is as follows - Chrono 7.0.0 - OptiX 7.2.0 - CUDA 11.6.2 - USE_CUDA_NVRTC=OFF - CUDA_ARCH_NAME = Auto Everything is running fine on Pascal or Turing GPUs and we see the lidar point cloud data that we would expect. However, when a colleague of mine is building and running the same code on his PC with Ampere GPU, the model as such is running OK, but we get "flashing" lidar point cloud data. More precisely, the correct point cloud data is returned at discrete time points, but most of the time the sensor seems to be returning blank data (i.e., zero intensity returns everywhere). We get the same behavior when using CUDA NVRTC. Is this a known issue or has anyone observed something similar? Might this problem be solved by upgrading to Chrono 8.0 and/or a more recent OptiX or CUDA release? To my understanding, the above toolset should be compatible with Ampere 8.6 architecture and the observed behavior appears to be really strange :/ Any help is highly appreciated! Thanks, Christian Christian Feller Lead Simulation Engineer Danfoss Power Solutions GmbH & Co. OHG Virtual Solutions & Verification DPS R&D / Software -- You received this message because you are subscribed to a topic in the Google Groups "ProjectChrono" group. To unsubscribe from this topic, visit https://groups.google.com/d/topic/projectchrono/ob5Xw3jkS-4/unsubscribe. To unsubscribe from this group and all its topics, send an email to [email protected]<mailto:[email protected]>. To view this discussion on the web visit https://groups.google.com/d/msgid/projectchrono/f891e87b-3633-485d-90fe-3115cc9a219bn%40googlegroups.com<https://groups.google.com/d/msgid/projectchrono/f891e87b-3633-485d-90fe-3115cc9a219bn%40googlegroups.com?utm_medium=email&utm_source=footer>. -- You received this message because you are subscribed to the Google Groups "ProjectChrono" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]<mailto:[email protected]>. To view this discussion on the web visit https://groups.google.com/d/msgid/projectchrono/CAOtC80Ma6VZpKG1ukpwatvJ0_sAm9y-UmQcnv6kvoMC6wRpRkA%40mail.gmail.com<https://groups.google.com/d/msgid/projectchrono/CAOtC80Ma6VZpKG1ukpwatvJ0_sAm9y-UmQcnv6kvoMC6wRpRkA%40mail.gmail.com?utm_medium=email&utm_source=footer>. -- You received this message because you are subscribed to the Google Groups "ProjectChrono" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/projectchrono/DM8PR06MB77038D5E3845930F801CEF10B1B6A%40DM8PR06MB7703.namprd06.prod.outlook.com.
