Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

CUDAPolisher core dumps #21

Open
noncodo opened this issue Aug 21, 2020 · 1 comment
Open

CUDAPolisher core dumps #21

noncodo opened this issue Aug 21, 2020 · 1 comment

Comments

@noncodo
Copy link

noncodo commented Aug 21, 2020

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
Copy link

Experiencing same thing (I think): NVIDIA-Genomics-Research/GenomeWorks#662
No response yet.

# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants