Feiyi Wang

hmm ... the use case has MPI in mind (cluster settting), so multiple processes are a given. In that case, it seems the only way is to have each process...

this is Redhat 7.4, running kernel 4.11.0-44.5.1.el7a.ppc64le. Note this is Power 9 architecture, with Mellanox dual port EDR ...

Actually, it is little-endian, as shown from ppc64le designation. strace output: ``` execve("/root/p27/bin/ibtool", ["ibtool", "sminfo"], [/* 44 vars */]) = 0 brk(NULL) = 0x21050000 access("/etc/ld.so.preload", R_OK) = -1 ENOENT (No...

It seems trying different path: looks like you open `umad0` before this call. `strace sminfo`, the related portion seems: ``` open("/sys/class/infiniband_mad/umad0//ibdev", O_RDONLY) = 3 read(3, "mlx5_0\n", 20) = 7 close(3)...

I tried that - doesn't seem to fix the problem ... > ioctl(3, _IOC(_IOC_READ, 0x1b, 0x03, 0x00), 0) = -1 ENOTTY (Inappropriate ioctl for device) This strace output is a...

still, strace: `ioctl(3, _IOC(_IOC_READ|_IOC_NONE, 0x1b, 0x01, 0x1c), 0x3fffd2667fe0) = -1 ENOTTY (Inappropriate ioctl for device)`

@adammoody my understanding is either "star" or "watching" the repo doesn't give you the push of commit message, but "watching" did provide update on all issues and conversations. bringhurst's fix...

we have a parallel compression implementation in the works, which will reply on dtar for phase 1 processing, when do you expect this move to happen?

Ahana (from Florida state) is doing the actual work. She came up with a parallel compression algorithm and now doing a proof-of-concept implementation. I think she just works off the...

@adammoody, I think what you described make sense ... a custom header would have made many things easier for parallel processing, the only downside is compatibility with conventional tool/format, but...