Segmentation fault when creating PuschRx instance in PyAerial

Accelerated Computing Aerial Forum
1

Hello,

I’m testing PUSCH channel decoding on Aerial Data Lake data. (with this documentation: Using pyAerial for PUSCH decoding on Aerial Data Lake data — Aerial CUDA-Accelerated RAN)

When attempting to create a PuschRx instance in PyAerial, the Python process immediately terminates with a segmentation fault.

setup details: GH200, with driver version 575.57.08, cuda version 12.9

minimum reproducible code

from aerial.phy5g.pusch import PuschRx
from aerial.phy5g.config import PuschLdpcKernelLaunch

pusch_rx = PuschRx(
    cell_id=0,
    num_rx_ant=4,
    num_tx_ant=4,
    enable_pusch_tdi=False,
    eq_coeff_algo=0,
    ldpc_kernel_launch=PuschLdpcKernelLaunch.PUSCH_RX_LDPC_STREAM_SEQUENTIAL,
)

Segmentation fault (core dumped)

Any guidance would be very helpful.
Thank you in advance!

2

Hello,
Thanks for finding this bug in the current version, we will resolve this in the next release.
In the meantime, can you please use eq_coeff_algo >= 1