open-mpi/hwloc

misc TODO

Closed this issue · 41 comments

Tools

  • bind process on 2 cores "near" physical proc id 3 ?
    • hwloc-calc: add an option to request a cpuset containing of n close entries among the generated cpuset
  • internationalize the output of lstopo? object types and memory size units
  • hwloc-top, like lstopo, but keeps printing every 3s or so, and show bound threads as well as the used CPU%

Doc

  • automatically generate the pngs?
    • see doc/images/HACKING

Support

  • add info about supported instructions (sse, avx, ...)
  • add info about available execution units (fpu)
    • and say if they are shared between threads/cores
      • this could help improving the current ambiguity between two real cores, one hyperthreaded core, and AMD dual-fake-core compute units
  • reduce distance matrices so that parent objects get distances between them as well (just like we do when computing group distances after inserting groups)
  • parallelize the discovery ? :)

I/O

  • CCI interoperability to get cci_device and/or cci_device->name locality
    • use cci_device->pci.{domain,bus,dev,func} to retrieve the PCI device
    • wait for the CCI API to be stable
  • Add a ofed plugin to gather ofed device info without relying on Linux sysfs
    • Not sure whether ofed works the same on other OS anyway

Backends and Ports

  • Try to make the distance grouping code a separate component ?
  • QNX
    • _syspage_ptr() SYSPAGE_ENTRY(entry)
    • ThreadCtl/Thread_ctl_r(_NTO_TCTL_RUNMASK)
  • BSD
    • sys/sched.h: sched_bind/sched_unbind, but that's in-kernel only for now.
  • AIX
  • Cray Catamount?

Imported from trac issue 15. Created by bgoglin on 2009-10-09T03:22:34, last modified: 2014-09-12T04:53:34

Trac comment by sthibaul on 2010-01-24 13:12:24:

(In [1660]) Add x86-cpuid backend. Refs #15.

Trac comment by sthibaul on 2010-02-14 15:52:24:

AMD64 topology done in r1699, r1700, r1701.

Trac comment by bgoglin on 2010-02-28 07:54:26:

fix title

Trac comment by bgoglin on 2010-02-28 14:22:11:

distance stuff moved to #32

Trac comment by bgoglin on 2010-02-28 14:25:56:

updates:

  • AMD magny-cours support is OK
  • numa distance in XML is in #32
  • envvar are properly documented (except the ugly debugging one, on purpose)

Trac comment by bgoglin on 2010-03-31 02:01:49:

Use parts of /proc/cpuinfo in the sysfs backend

Trac comment by bgoglin on 2010-05-01 17:20:24:

add hwloc-calc TODOs

Trac comment by bgoglin on 2010-05-19 01:03:14:

link to taskset support discussion

Trac comment by bgoglin on 2010-05-22 05:38:00:

  • taskset is done
  • rename hwloc_cpuset_from_string into hwloc_cpuset_sscanf ?

Trac comment by bgoglin on 2010-05-25 11:35:08:

Author: bgoglin
Date: 2010-05-25 11:28:20 EDT (Tue, 25 May 2010)
New Revision: 2142
URL: https://svn.open-mpi.org/trac/hwloc/changeset/2142

Log:
When nothing (except options) is given on the hwloc-calc command-line,
read from stdin and process line-by-line. Multiple objects in the same
line should be separated by spaces.

Trac comment by bgoglin on 2010-07-10 03:24:37:

  • misc ideas from Jirka Hladky (additions to tools)
  • remove some stuff that was implemented (insert misc obj in topo/lstopo)

Trac comment by bgoglin on 2010-07-13 04:13:08:

hwloc-calc --intersect added in r2307

Trac comment by bgoglin on 2010-09-19 13:20:57:

Need to use /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages

Trac comment by bgoglin on 2010-09-26 16:58:30:

cpuset_from_string renamed into bitmap_sscanf in the bitmap branch

Trac comment by bgoglin on 2010-10-30 10:56:19:

update stuff, remove obsolete/done stuff, ...

Trac comment by sthibaul on 2010-11-16 18:44:59:

Added idea of topology configuration to restrict detection to a portion of the machine.

Trac comment by bgoglin on 2011-01-04 10:44:45:

bitmap<->stringlist functions added in trunk@3013

Trac comment by bgoglin on 2011-01-04 10:47:26:

misc ideas

Trac comment by bgoglin on 2011-02-16 11:30:00:

random ideas probably good for 1.2

Trac comment by bgoglin on 2011-02-17 12:09:59:

hwloc-calc --number-of added in 3177

Trac comment by bgoglin on 2011-02-18 05:23:17:

HT doc in r3180

Trac comment by bgoglin on 2011-03-28 15:27:24:

Relaxed distance-based grouping

Trac comment by bgoglin on 2011-04-05 17:18:07:

random remaining stuff for the main I/O ticket (#5 now closed)

Trac comment by bgoglin on 2011-04-27 04:51:56:

remove misc stuff that are done or have their own ticket

Trac comment by sthibaul on 2011-08-12 08:04:26:

Use cpuid's apicid to implement get_last_cpu_location(THREAD)

Trac comment by bgoglin on 2011-09-23 15:52:29:

Jeff/Ralph requests binding info in XML and a "symmetric children" flag

Trac comment by bgoglin on 2011-10-03 01:01:51:

(In [3894]) Add a symmetric_subtree object attribute

Refs #15

Trac comment by bgoglin on 2011-10-03 01:28:08:

symmetric attribute done

Trac comment by bgoglin on 2011-10-04 08:13:13:

trunk@3907 can now combine cpuinfo (for cpu model) and sysfs info.

Trac comment by bgoglin on 2011-10-23 02:01:55:

misc additions from my todo list

Trac comment by bgoglin on 2012-04-23 08:12:44:

cci interoperability

Trac comment by bgoglin on 2012-07-22 02:06:24:

cleanup/update stuff

Trac comment by bgoglin on 2012-11-17 03:20:35:

misc updates

Trac comment by bgoglin on 2012-12-11 16:15:47:

misc updates

Trac comment by bgoglin on 2013-01-18 09:16:39:

(In [5249]) hwloc-info: Add --ancestors and --ancestor

They dump one or all ancestors of the given objects.

Refs #15

Trac comment by bgoglin on 2013-01-18 15:24:52:

two requests from the hwloc tutorial at ComPAS last monday

Trac comment by bgoglin on 2013-02-15 16:38:14:

BlueGene is done for real, it's not a random joke anymore.

Trac comment by bgoglin on 2013-09-11 06:10:57:

misc updates

Trac comment by bgoglin on 2014-03-31 16:37:24:

these lines have their own tickets.

Moved all slightly-relevant remaining items to their own ticket.