"CUDA device kernel image is invalid" error with cn/r

[kusayuzayushko]

xmrig-nvidia-v2.14.0 OS: SimpleminingOS Kernel version: 4.17.19-smos10 Driver version: Nvidia 415.13, Cuda 10.0 (Latest SMOS) GPUs: 1080t, 1060 and 4 1050Ti, multiple rigs/cards/users are affected

miner crashes with this error

[2019-03-09 22:42:02] nvrtcCompileProgram failed: NVRTC_ERROR unknown
[2019-03-09 22:42:02] Program compile log: nvrtc: error: failed to load builtins
[CUDA] Error gpu 0: _cryptonight_gpu_hash_:878 _device kernel image is invalid_
terminate called after throwing an instance of 'std::runtime_error'
what(): [CUDA] Error: device kernel image is invalid

Any idea what't the reason?

[xmrig]

Library libnvrtc-builtins.so (part of NVRTC) is missing. Also please check https://github.com/xmrig/xmrig-nvidia/pull/235#issuecomment-463040311 might trick with LD_LIBRARY_PATH help. Thank you.

[marcel1974]

I had the same problem NOW It's working....!!! on graft but not lethean. If I mine directly to pool.graft.hashvault.pro:3333 with xmrig-2.14.0 it report cn/rwz and it mine with no error. If I change the pool to pool.hashvault.pro:3333 or :80 it reports cn/r and boms with: [2019-03-10 07:22:14] new job from 192.168.0.200:3311 diff 18720 algo cn/r height 392058 [2019-03-10 07:22:14] nvrtcCompileProgram failed: NVRTC_ERROR_COMPILATION [2019-03-10 07:22:14] Program compile log: CryptonightR.cu(382): error: identifier "__funnelshift_l" is undefined

CryptonightR.cu(383): error: identifier "__funnelshift_r" is undefined

2 errors detected in the compilation of "CryptonightR.cu".

[CUDA] Error gpu 0: <cryptonight_gpu_hash>:878 "device kernel image is invalid" terminate called after throwing an instance of 'std::runtime_error' what(): [CUDA] Error: device kernel image is invalid Aborted

[xmrig]

@marcel1974 your issue solved in 2.14.1. #246 Thank you.