we built a end-end test platform using aerial-accelerated-ran as a gNB and many real UEs. when gNB and many ues ran, cuphy in aerial-accelerated-ran print the error“WRN UlPhyDriver04 0 phychannel.cpp:368 [SB-L1] [DRV.PHYCH] [/opt/nvidia/cuBB/cuPHY-CP/cuphydriver/src/common/phychannel.cpp:368] cuda failed with result1 an illegal memory access was encountered” after the 161st ue accessed to gNB. Then L1 in aerial-accelerated-ran exited. What is happed on the error printed? how should we do to fix the error? Thanks a lot
Hi @harine_wind ,
Can you please provide the following:
- Build flags you used,
- Full phy.log
- Yaml configuration files
Do you see print of the backtrace when the cuphycontroller stops? This should also been captured in the phy.log or on the console logs.
Thank you.
Hi @harine_wind
additional info needed… is it on aerial-testbed with OAI L2+ stack or a different non-opensource L2+ stack?
Thanks.
