Open ply2022 opened 11 months ago
That sounds like too much. Could you disable the GPU kernel call and use valgrind/mem-check to find leaks? Did you check allocations performed and their size?
Hi,
I don't know how to do that on my end. I will get help from the Research Computing staffs. I will provide feedback once I hear from them. Thank you.
Hi,
I am doing all-vs-all pairwise alignment of filtered ONT reads (phred score > 9). Due to extensive memory usage, I have to split reads into 10,000 reads per file, which is about 140 MB. Yet, it took 230 GB to complete one pairwise alignment. Does that sound normal?
I used 8/32/64 (Number of CPU threads / number of blocks / number of threads per block) setting for GPU accelerated Darwin on NVIDIA A100. How can I lower the CPU memory usage? Thanks.