Armin Töpfer

Hey! For our CCS binary we are ricing like hell, as every percentage performance is actual money saved and reduces time to result for future clinical samples. Who wouldn't like...

>Are you including snmalloc in the PGO / LTO bundle? yes I'll test `nm` and let you know.

Statically linked w/PGO&LTO&GLIBC2.32 on 2x7H12: | Run | Wall time | CPU Time | Peak RSS | | - | - | - | - | | MIMALLOC Tuned* |...

``` $ cat /sys/kernel/mm/transparent_hugepage/enabled [always] madvise never ``` Following all static, PGO/LTO builds on 2x7H12 systems. | Metric | SN 16 QEMU=On | SN 16 | MI Tuned | MI...

Sure thing. But even without `MIMALLOC_LARGE_OS_PAGES`, `mimalloc` is much faster than `snmalloc` per default. So there is something beyond pages that is not working out in my use case. Ping...

Updated.

>are you using the same number of OS threads as hardware threads in your application? Correct. I'm using the same number of compute heavy threads as I have (logical) cores...

Since we are using mm2 as a library, reproducing outside is tricky because this step uses ~300k sequences as references, I'm still working on reproducing it with the minimap2 binary....

I have the read and the 300k references. Do you only need the read?

My core dump tells me I'm using those parameters: ``` {mapOpts_ = {seed = 11, sdust_thres = 0, flag = 524293, max_qlen = 0, bw = 50, max_gap = 500,...