CUDAPolisher core dumps

[noncodo]

I've been trying to get racon-gpu to run on my data (2x ~1kb contigs) but keep hitting a wall

$ racon -m 8 -x -6 -g -8 -c 1 reads.fastq aligned.paf dirtyCotigs.fa

Using 4 GPU(s) to perform polishing
Initialize device 0
Initialize device 1
Initialize device 2
Initialize device 3
[CUDAPolisher] Constructed.
[racon::Polisher::initialize] loaded target sequences 0.000031 s
[racon::Polisher::initialize] loaded sequences 22.377420 s
[racon::Polisher::initialize] loaded overlaps 3.298172 s
[racon::Polisher::initialize] aligning overlaps [====================] 12.168808 s
[racon::Polisher::initialize] transformed data into windows 2.449713 s
[racon::CUDAPolisher::polish] allocated memory on GPUs for polishing 6.232884 s
Floating point exception (core dumped)

Fiddling around with the parameters a bit, I get some more verbose output and seem to make it a bit further when specifying: $ racon -c 1 --cudaaligner-batches 1 reads.fastq aligned.paf dirtyCotigs.fa

Using 4 GPU(s) to perform polishing
Initialize device 0
Initialize device 1
Initialize device 2
Initialize device 3
[CUDAPolisher] Constructed.
[racon::Polisher::initialize] loaded target sequences 0.000029 s
[racon::Polisher::initialize] loaded sequences 21.847521 s
[racon::Polisher::initialize] loaded overlaps 3.320273 s
GPU 0: Aligning with band width 46
GPU 1: Aligning with band width 46
GPU 2: Aligning with band width 46
GPU 3: Aligning with band width 46
[racon::CUDAPolisher::initialize] allocated memory on GPUs for alignment 6.027738 s
Alignment skipped by GPU: 1243722 / 1295352overlaps [=====>              ] 8.770285 s
[racon::Polisher::initialize] aligning overlaps [====================] 93.255284 s
[racon::Polisher::initialize] transformed data into windows 2.451490 s 121.277171 s
[racon::CUDAPolisher::polish] allocated memory on GPUs for polishing 5.430252 s
GPU Error:: invalid configuration argument /home/apps/racon-gpu/vendor/GenomeWorks/cudapoa/src/cudapoa_kernels.cuh 546

Aborted (core dumped)

Compiled with: C/CXX compiler = GNU 7.5.0 CUDA v10.2

Running with: 4x RTX2080Ti

[schorlton]

Experiencing same thing (I think): https://github.com/clara-parabricks/GenomeWorks/issues/662 No response yet.