CUDA error in ORBextractor.cc

[chivas1000]

Hi, I got the code built and run, but after I parse the commands, it pops out the viewer and stuck.

CUDA error at /home/ubuntu/SwarmMap/code/src/cuda/Fast_gpu.cu:400 code=13(cudaErrorInvalidSymbol) "cudaMemcpyToSymbol(c_u_max, u_max, count * sizeof(int))"

using breakpoint I saw this error at ORBextractor.cc line 403

seem to be like a cuda issue

is that because my GPU not compactible with cuda10.2? My GPU is RTX3060 which version of GPU would work?

has anyone ran into this error or is there any suggestions could help? any assistance would be appreciated. thanks.

environment setup: i7 12700H RTX3060 LAPTOP

ubuntu18.04

spdlog (didn't specified so I used 1.7.0) Pangolin (v0.5) Eigen3 (18.04 debian packages) CUDA (10.2) OPENCV (3.4.6) BOOST (1.70.0) BEAST (included in BOOST)

and I manually added the libboost_serialization.so in CMakeLists.txt but not find package for boost_serialzation moudule since it said the build would fail for linking a static library

FIND_PACKAGE(Boost REQUIRED COMPONENTS system filesystem date_time serialization)

FIND_PACKAGE(Boost REQUIRED COMPONENTS system filesystem date_time)

target_link_libraries(${BASE_LIBRARY} ${OpenCV_LIBS} ${EIGEN3_LIBS} ${Boost_LIBRARIES} ${Pangolin_LIBRARIES} ${SRC_DIRECTORY}/Thirdparty/DBoW2/lib/libDBoW2.so ${SRC_DIRECTORY}/Thirdparty/g2o/lib/libg2o.so nvToolsExt pthread spdlog::spdlog /usr/local/lib/libboost_serialization.so.1.70.0 )

commands: /home/ubuntu/SwarmMap/debug/bin/swarm_map -d /home/ubuntu/SwarmMap/config/mh01-03.yaml -v /home/ubuntu/SwarmMap/code/Vocabulary/ORBvoc.bin -c 3 -l debug

outputs: ORB-SLAM2 Copyright (C) 2014-2016 Raul Mur-Artal, University of Zaragoza. This program comes with ABSOLUTELY NO WARRANTY; This is free software, and you are welcome to redistribute it under certain conditions. See LICENSE.txt.

Input sensor was set to: Monocular

Loading ORB Vocabulary. This could take a while... Vocabulary loaded!

Camera Parameters:

• fx: 458.654
• fy: 457.296
• cx: 367.215
• cy: 248.375
• k1: -0.283408
• k2: 0.0739591
• p1: 0.00019359
• p2: 1.76187e-05
• fps: 20
• color order: RGB (ignored if grayscale) CUDA error at /home/ubuntu/SwarmMap/code/src/cuda/Fast_gpu.cu:400 code=13(cudaErrorInvalidSymbol) "cudaMemcpyToSymbol(c_u_max, u_max, count * sizeof(int))"

[Halcao]

You can change cudaMemcpyToSymbol to checkCudaErrors(cudaMemcpyToSymbol()) to get more specific information for this error. I'm sorry that I have no experience on CUDA and cannot offer more help.

[lukeluocn]

I encountered the same issue with RTX3080Ti (driver version: 535.54.03). I then switched to a different platform using GTX1070 (driver version: 440.33.01), which successfully resolved the problem without requiring any additional modifications.