NO CUDA-capable device is detected

Accelerated Computing Aerial Forum
1

I am trying to run the stand alone LDPC decoder example under cuPHY e.g. here
build/cuPHY/examples/error_correction$
cuda_ex_ldpc
I get the following error is it expected?

AERIAL_LOG_PATH set to /home/sbaker/AERIAL_LOG
Log file set to /home/sbaker/AERIAL_LOG/ldpc_decoder.log

15:42:39.626634 WRN 269165 0 [NVLOG.CPP] Using /home/sbaker/zodiacArtemis2/zodiac/trunk/Prototypes/Artemis2/cuda_test_3/cuPHY/nvlog/config/nvlog_config.yaml for nvlog configuration
15:42:39.786198 ERR 269165 0 [AERIAL_CUPHY_EVENT] [CUPHY.PUSCH_RX] EXCEPTION: no CUDA-capable device is detected

I am running on x86 server with RTX3090 GPU

When I compiled cuPHY is set the CUDA architecture 86 in CMakeLists.txt is this the issue e.g.

set(CMAKE_CUDA_ARCHITECTURES 86)

sbaker@ubuntusim12:~/zodiacArtemis2/zodiac/trunk/Prototypes/Artemis2/cuda_test_3$ lspci | grep VGA
1a:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)
68:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)

sbaker@ubuntusim12:~/zodiacArtemis2/zodiac/trunk/Prototypes/Artemis2/cuda_test_3$ lspci | grep -i nvidia
1a:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)
1a:00.1 Audio device: NVIDIA Corporation GA102 High Definition Audio Controller (rev a1)
68:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)
68:00.1 Audio device: NVIDIA Corporation GA102 High Definition Audio Controller (rev a1)

nvidia-smi
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Our IT are investigating why the NVIDIA driver not running

2

This issue was resolved after our IT fixed the NVIDIA driver
I could run the example e.g.
cuPHY/examples/error_correction$ ./cuphy_ex_ldpc
AERIAL_LOG_PATH set to /home/sbaker/AERIAL_LOG
Log file set to /home/sbaker/AERIAL_LOG/ldpc_decoder.log

NVIDIA GeForce RTX 3090: 82 SMs @ 1695 MHz, 23.7 GiB @ 9751 MHz, Compute Capability 8.6, PCI 0000:1A:00

LDPC Configuration:

Source = (generated at runtime)
BG (base graph) = 1
Kb (info nodes) = 22
Z (lifting size) = 384
B (code block size in bits) = 8448
F (filler bits) = 0
K (info bits, B + F) = 8448
mb (parity nodes) = 8
P (punctured parity bits) = 0
N (modulated bits, B+Z(mb-2)-P) = 10752
Number of codewords = 1
R (code rate = B / N) = 0.786
Modulation = QPSK
Punctured info nodes = DISABLED
14:38:54.398632 WRN 9462 0 [NVLOG.CPP] Using /home/sbaker/zodiacArtemis2/zodiac/trunk/Prototypes/Artemis2/cuda_test_3/cuPHY/nvlog/config/nvlog_config.yaml for nvlog configuration
Normalization = 0.750000
Number of iterations = 1

Average (1 runs) elapsed time in usec = 15.4, throughput = 0.55 Gbps
bit error count = 0, bit error rate (BER) = (0 / 8448) = 0.00000e+00, block error rate (BLER) = (0 / 1) = 0.00000e+00
Exiting bg_fmtlog_collector - log queue ever was full: 0

3

Hi @stuart.baker,

Thank you for letting us know. We are happy to hear that the issue has been resolved.

Thank you.

4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.