Armin Töpfer

Avoid buffer overflow in mm_sync_regs

9

@lh3 This is not meant for merging, but rather a question if this stopgap solution makes sense and would do any harm? I need a bandage this week to fix...

kalloc panic

2

Hi Heng, our IsoSeq clustering algo relies on mm2 and so far no problems in the last 1.5 years. We just hit a reproducible kalloc panic: [kfree] The end of...

>50% cache miss rate

6

I've been benchmarking software that is using mm2 internally. It turns out that there is massive cache miss rate caused by mm2. I ran vanilla mm2.15 on some HG002 CCS...

Assertion failed

12

Hi! I've been using SPSCQueue for some time and just updated to the latest version and now hit this: ``` SPSCQueue.hpp:188: void Rigtorp::SPSCQueue::pop() [with T = PacBio::CCS::Chunk; Allocator = std::allocator]:...

Use original docopt

1

Use https://github.com/docopt/docopt.cpp as a submodule instead of copying into our source.

Only provide BAM output

3

Drop all output files except for BAM. Annotate metrics to each record using custom BAM tags, such as super-read coverage or the read names of that were used.

Try different enumeration techniques

1

Try out heuristic MCE algos by Eppstein or enumerate quasi-cliques. Potential drop in replacements: - https://github.com/darrenstrash/quick-cliques - http://c0de-x.com/how-to-find-maximal-cliques/

Add naive regression test

1

Add a very simple test case of perfect reads from two haplotypes and assemble them correctly.

Use include/ for headers and src/ for sources + private headers