very fast tree?

[jianshu93]

Hi Mike,

I was planning to build very larger trees via fasttree2 option but I found that for most recent AMD and Intel CPUs, SIMD version of fasttree2, which is called veryfasttree can be 2-8 times faster. Would it be possible to add it as an additional support? The bad thing is users have to compile it according to their machine type (e.g., whether AVX2 or 512 is supported), conda version may not support all instructions, maybe only avx2.

Thanks,

Jianshu

[AstrobioMike]

Hi there, Jianshu :)

Thanks for the note, I didn't know about veryfasttree yet. Seems invaluable for large trees!

I think you're right that the conda install only supports avx2, based on the recipe here. That does make me hesitant to add it in...

You may have seen this already, but there is a way to run GToTree doing everything except the treeing part (by adding the -N flag). So you could consider doing that and then plugging the alignment into veryfasttree without GToTree wasting the time trying to build the tree.

[AstrobioMike]

Note for myself, it looks like as of a few weeks ago this might be more flexible to OS based on the recipe update here 👍 I need to investigate

[AstrobioMike]

since the conda recipe has been improved to deal with mac/darwin too under-the-hood (and i finally got around to testing it), this is included as a treeing option as of GToTree v1.8.11 (set with the -T parameter) 👍

Thanks again, @jianshu93!