; We had about 50 visits on the v1.3 download page since rc2 (and more
> than 300 since rc1). I hope people did actually test it.
>
> Brice
>
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/li
hwloc_topology_export_xmlbuffer(t1, , );
It is this export that fails, which implies to me that somehow the import
didn't work right. Note that this code worked fine with libxml2, so this is a
regression.
On Sep 22, 2011, at 9:39 AM, Jeff Squyres wrote:
> Yes, I can get some testing of the o
just to figure out how to use an
> accessor function.
Why not? ;-)
I put the note in a per-section basis; I figured that would be "enough" (vs.
putting something in the description of each function in that section).
--
Jeff Squyres
jsquy...@cisco.com
For corporate l
How's this?
https://svn.open-mpi.org/trac/hwloc/changeset/3853
On Sep 24, 2011, at 10:01 AM, Ralph Castain wrote:
>
> On Sep 24, 2011, at 5:52 AM, Jeff Squyres wrote:
>
>> On Sep 24, 2011, at 7:46 AM, Jeff Squyres wrote:
>>
>>>> The funky thing he
Thanks!
How's this? https://svn.open-mpi.org/trac/hwloc/changeset/3852
On Sep 20, 2011, at 8:57 AM, Brice Goglin wrote:
> Le 20/09/2011 14:44, Jeff Squyres a écrit :
>> 1. Is it permissible to use _PROCESS or _THREAD with
>> get_proc_last_cpu_location() and get_proc_cpubind()
On Sep 24, 2011, at 7:46 AM, Jeff Squyres wrote:
>> The funky thing here is that the parent/child links between the first
>> socket and its core go across level 2 because nothing matches there. In
>> the first socket, you have Socket(depth1)->Core(depth3) while in the
>
all" will
never traverse down there, etc. If you include it in DIST_SUBDIRS, it'll be
included in the tarball (which is necessary if you generate the Makefile[.in]),
but the other usual "make" targets don't have to go in there.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
igure has
> to manage shouldn't be a problem, we already have many .in
> files that are not Makefile.am.
>
> Brice
>
>
>
> Le 23/09/2011 20:59, Jeff Squyres a écrit :
>> Err... is this right?
>>
>> Does "make dist" and friends still work?
>>
>&
| grep -v rename.h: | grep -v .svn >/dev/null \
> + || echo $$name; \
> + done
> ___
> hwloc-svn mailing list
> hwloc-...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-svn
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
ing width will be, etc. This
front-end scheduler needs to know whether the back-end node is capable of
supporting binding, for example.
We manually added this information into the message that we send up to the
scheduler, but it would be much nicer if the XML export/import just handled
that automatica
ewise, I get a "line=64" output, which makes no sense.
The attached patch clarifies the output.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
printattr.diff
Description: Binary data
ow we're slurping OMPI's hwloc into
our tree... nothing wrong in hwloc itself...
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
'opal_hwloc121hwloc_backend_xml_init'
make[1]: *** [topology.lo] Error 1
make: *** [all-recursive] Error 1
We use the hwloc prefix stuff in the OMPI embedded build; did something not get
prefixed properly in the minimal XML stuff?
Back in a few hours...
On Sep 22, 2011, at 9:39 AM, Jeff Squyres wrote:
> Yes
.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
te:
> Le 06/09/2011 17:39, Jeff Squyres a écrit :
>> On Sep 6, 2011, at 11:25 AM, Brice Goglin wrote:
>>
>>>> - I don't know where we ended up in the other thread: do we want JSON or
>>>> no? If we can parse it easily without an external dependency, then I
late in the OMPI v1.5 series... but I've already got some feedback that
>> there might be resistance to what we're proposing to use hwloc for in OMPI,
>> so we'll see :-) ).
>
> Agreed, v1.3 is too young for OMPI. I am not sure many people already
> tried v1.3rc1.
RCs. "RC", to me, has
always meant "feature complete."
> I'm a bit too lazy to release a 1.2.2 except if there's a single RC :)
> And I think I would vote for 1.3rc2 now too.
I can drive the v1.2.x release stuff.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
On Sep 6, 2011, at 8:49 AM, Jeff Squyres wrote:
> Ah, it looks like we got around this in OMPI by explicitly checking to see if
> the compiler vendor is Sun. Terrible! Let me see about back-porting these
> changes from OMPI...
I did some internal cleanup on that configury test and
nforced for user-given
> strings (object info attributes).
>
> Brice
>
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
thanks.
> Another thing: Even with visibility disabled, I'm getting test failures
> on, I believe only i386 (so that should mean sparc and sparcv9 are passing)
> Can't confirm for amd64 yet.
Ok. Send along all the data and we'll have a look to see what's going wrong.
--
Jeff Squyres
I think this has now been resolved, right?
On Sep 5, 2011, at 6:45 PM, Samuel Thibault wrote:
> Hello,
>
> Jeff Squyres, le Sat 27 Aug 2011 12:40:20 +0200, a écrit :
>> Here's some output from the nightly build machine about the "make check"
>> errors that are
I do not believe your diagnosis is correct (but I am not at keyboard to check).
The configure visibility stuff turns on 2 different things:
- visibility compiler attributes
- compiler cli flags (fvisibility)
With your sed, u only disabled the cli options, whichight be redundant anyway.
But
t; - I am a bit afraid that we would go from a well-working XML support to
> something much less reliable (do we need to be fully XML compliant so
> that external programs can load our XML files and play with them?)
A fair point.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal info
loc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
I don't know enough about either format to say.
Sent from my phone. No type good.
On Sep 2, 2011, at 6:03 AM, "Samuel Thibault" <samuel.thiba...@inria.fr> wrote:
> Jeff Squyres, le Fri 02 Sep 2011 11:58:05 +0200, a écrit :
>> JSON: sure, it's an easy format, but we're
for these
formats, I'm ok with a totally trivial and
not-necessarily-compatibilte-with-anyone-else's-format format.
On Sep 1, 2011, at 9:38 PM, Christopher Samuel wrote:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
> On 02/09/11 01:30, Jeff Squyres wrote:
>
>> Is there any chance
the back-end nodes have libxml? For us to do what we want,
it would need to be available on all nodes because the OMPI orted processes
would be querying hwloc for the local topology and then sending it to the
"head" node process (usually mpirun) for further analysis and process mapp
Or "htx" for "hwloc tiny xml".
Hah! htx might work. Or:
hst: hwloc simple text
hnx: hwloc NoXML
htt: hwloc trivial text
simple: obvious
serialized: obvious
string: obvious
newline: because the fields are newline-delimited
...?
--
Jeff Squyres
jsquy...@cisco.com
F
' brains. :-) So -- "sometime soon"
would be nice.
I could *probably* write this, but I'm guessing you guys could write it much
faster than I could...
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
Ok, we finally fixed all the nightly tarball build issues with respect to the
build machine (recent versions of doxygen and tex-live now installed).
Nightly tarballs should now resume as normal.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web
ide the system latex install, and set an
environment variable (or something) to point to it?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
Here's some output from the nightly build machine about the "make check" errors
that are failing.
Does this help?
(I also just updated doxygen)
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
eddie-
ve two Misc
> objects (representing switches) above four identical machines.
>
> hwloc_topology_set_custom() initializes my new "custom" backend. It
> basically means that I am going to insert Misc objects and/or topologies
> that hwloc_topology_load() will just have to connect together later.
Done (i.e., I setup that for the env for the build machine).
On Aug 22, 2011, at 3:55 PM, Brice Goglin wrote:
> Le 22/08/2011 21:52, Jeff Squyres a écrit :
>> For trunk and v1.3?
>
> We have the latest ones for v1.2 (2.68/1.11.1/2.4) so I am using the
> same fo
For trunk and v1.3?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
scheduler.sourceforge.net/projects/hwloc/GridEnginehwloc.html
>
> Rayson
>
> ==
> Open Grid Scheduler - The Official Open Source Grid Engine
> http://gridscheduler.sourceforge.net/
>
>
> On Thu, Aug 11, 2011 at 8:33 AM, Jeff Squyres &
Univa notified me that they're now using hwloc, so I added them to our "those
who are using hwloc" list.
http://www.open-mpi.org/projects/hwloc/
w00t.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
t; free(indexes);
> _______
> hwloc-svn mailing list
> hwloc-...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-svn
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
C_CHECK_FUNC for pthread_create in the utils setup m4
- might want to fail configure if it fails (because the utils clearly can't be
built)
- add -lpthread in the relevant Makefile.am's -- it shouldn't be added to the
overall LIBS
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information
oc.
I'll tinker and see if a newer autoconf fixes the problem (i.e., no caching),
because it definitely didn't happen to me last night.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
wrote:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
> On 13/06/11 23:11, Jeff Squyres wrote:
>
>> Chris -- could you send your config.log?
>
> Posted but currently held for moderation due to size..
>
> - --
>Christopher Samuel - Senior Systems Ad
:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
> On 13/06/11 22:22, Jeff Squyres wrote:
>
>> Chris -- does nm on your libpci not show this?
>
> Nope, there is no libpci.so* on RHEL5.6, just a libpci.a.
>
> (and no, I've no idea why either!)
>
> cheers!
0\231:
>> topology-x86.c:222:61: warning: ?\200\230apic_id?\200\231 may be used
>> uninitialized in this function
>
> It's odd that I haven't seen those in my tests. I guess we don't
> automatically add -O2?
OMPI automatically adds annoying flags like -pedantic to developer
Ok, I get it now -- cache is probably about the only object where this could
occur, right?
On Apr 14, 2011, at 9:08 PM, Jeff Squyres wrote:
> On the subject of asymmetric trees, I see the description of
> hwloc_type_get_depth() in hwloc.h:
>
> /** \brief Returns the depth of obj
Is that worth a new release? Possibly even a 1.1.3?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
:1321
1321 while (fscanf(file, "%lx,", ) == 1) /* read one kernel cpu mask
and the ending comma */
(gdb) print file
$1 = (FILE *) 0x0
(gdb)
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
>
> I am tired of getting these undebuggable reports from this crappy RHEL4.
>
> Brice
>
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyr
ou should probably check for libz (if/when needed)
before adding -lz.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
io ports.
K.
Are we *requiring* libpci to build? Or do we allow building without it (and
therefore you just don't get the PCI functionality)?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
uct hwloc_obj * hwloc_get_next_pcidev(struct
> hwloc_topology *topology, struct hwloc_obj *prev);
> +
> +/** \brief Get the next OS device in the system.
> + *
> + * \return the first OS device if \p prev is \c NULL.
> + */
> +HWLOC_DECLSPEC struct hwloc_obj * hwloc_get_next_osdev(struct hwloc_topology
> *topology, struct hwloc_obj *prev);
> +
> +/** @} */
> +
> +
> +
> #ifdef __cplusplus
> } /* extern "C" */
> #endif
>
>
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
Coverity or whether we just enroll hwloc as a "normal" open source project, but
I could find out, if we're interested.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
e -fpic
> -DPIC on the compile line).
>
> I think earlier problems with PGI were caused by versions <= 10.
>
> Anybody else tried PGI 11? Any idea what's going on above?
>
> Brice
>
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
FWIW OMPI is moving to a model like that.
Sent from my phone. No type good.
On Mar 1, 2011, at 4:47 AM, "Bernd Kallies" <kall...@zib.de> wrote:
> On Mon, 2011-02-28 at 22:17 +0100, Brice Goglin wrote:
>> Le 28/02/2011 22:04, Jeff Squyres a écrit :
>>&
ctions on vma properties. They would normally happen before any
> physical pages
> are allocated.
Ah -- good to know!
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
and bus
bandwidth, etc.).
I'm thinking outside the realm of HPC environments, where swapping isn't
uncommon.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
location
close to the receiver.
So: binding + pinning = binding (as long as you can ensure that the binding +
pinning was atomic!).
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
if all the data is going to be read-only, discover once and
share the data among threads. Depending on your layers of abstraction, this
may or may not be easy to do.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
; N0=256407 N1=1594468
> 7ffd861da000 default anon=2484496 dirty=2314066 swapcache=170430
> active=2396456 N0=1083215 N1=1401281
I'm sorry; I'm not too familiar with the output of /proc/*/numa_maps -- what is
this showing? I see some entries switching from active=X to swapcache=X,
assumedly meaning that they have been swapped out...?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
information when pages are swapped out?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
Done!
But we still need that info from Brad. :-)
(just in case he's only lightly reading this thread...)
On Feb 22, 2011, at 7:05 PM, Samuel Thibault wrote:
> Jeff Squyres, le Wed 23 Feb 2011 00:20:42 +0100, a écrit :
>> On Feb 22, 2011, at 6:02 PM, Samuel Thibault wrote:
>
+ hwloc_report_error("object intersection without inclusion!",
__LINE__);
/* We can't handle that. */
return;
case HWLOC_OBJ_CONTAINS:
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
wondering whether it's worth it to try to back-port the
de-C99 stuff to the v1.1 branch. If we're planning on doing 1.2 within a few
weeks, it's probably not worth it.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
On Feb 18, 2011, at 10:55 AM, Jeff Squyres wrote:
> Let's do pthread_once() -- we don't have any of the infrastructure mojo for
> memory barriers.
Better yet, let's avoid all the memory barrier issues and pthread_once issues
by just hard-coding the order of constants and putting a st
type_order. A simpler way is to use
>>> pthread_once().
Let's do pthread_once() -- we don't have any of the infrastructure mojo for
memory barriers.
What will we need on Windows? Are there other OS's without pthread_once()?
>> Is there no portable way to define a library _init function?
>
>
On Feb 17, 2011, at 6:53 PM, Samuel Thibault wrote:
> Jeff Squyres, le Thu 17 Feb 2011 23:15:26 +0100, a écrit :
>> -uint64_t cacheconfig[n];
>> -uint32_t cacheconfig32[n];
>> -uint64_t cachesize[n];
>> +uint64_t *cacheconfig = NULL;
>>
I've pushed this patch up to bitbucket:
https://bitbucket.org/jsquyres/hwloc-de-c99/
I've attached a new diff as compared to the current SVN trunk.
I'll now go reply to the other mails in this thread...
On Feb 18, 2011, at 3:37 AM, Brice Goglin wrote:
> Le 17/02/2011 23:15, Jeff Squy
I don't know what code that turns off in the ports.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
et_instant_cpuexec()
hwloc_get_cpuexec_now()
hwloc_get_my_last_cpuexec()
...?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
g?
Trunk, yes -- after Samuel comments and we talk about how to do the other back
ends.
Once we're done, I'll probably take a whack at back-porting to 1.1 (even if
it's not a straight svn merge) because it certainly is nicer to have stock
versions of hwloc in ompi rather than a hacked ve
On Feb 17, 2011, at 5:15 PM, Jeff Squyres wrote:
> I also attach an SVN diff against the trunk so that you can see the whole
> thing.
I forgot to mention that "make check" passes on both OS X and Linux (which was
an *enormous* help!).
--
Jeff Squyres
jsquy...@cisco.com
For
;now" in there
somewhere? That would certainly connote that as soon as the function returns,
the information may be stale.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
dependent (e.g.,
>> whether it makes a silent/temporary array or pointers, or converts indexing)?
>
> I don't think it can even be implementation dependent: here we just
> define a pointer, and assign to it a casted value.
Got it; thanks.
--
Jeff Squyres
jsquy...@cisco.com
quot;current" in here, somehow. Gives the connotation
that it might change in the future (if it's not bound).
Maybe get_current_cpuset?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
On Feb 14, 2011, at 8:14 AM, Jeff Squyres wrote:
>>> static void
>>> hwloc_setup_groups_from_distances(struct hwloc_topology *topology,
>>> unsigned nbobjs,
>>> struct hwloc_obj **objs,
>>>
On Feb 11, 2011, at 10:51 PM, Samuel Thibault wrote:
> Jeff Squyres, le Sat 12 Feb 2011 04:34:36 +0100, a écrit :
>> static void
>> hwloc_setup_groups_from_distances(struct hwloc_topology *topology,
>>unsigned nbobjs,
>>
Both the Autotools and CMake are fine systems. We have used the Autotools for
the past several years in OMPI projects for the following reasons:
- Autotools bootstrap a tarball (ie, the user does not need to have the
Autotools installed), whereas CMake requires that the user have it installed.
, distances is treated as a 2D array -- i.e., distances[i][j] is
accessed.
I have a bunch of guesses as to what this means, but none of them seem quite
right, so I figured I just ask -- what the heck does that declaration statement
mean?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal
c99 hwloc. I started
an hg branch for this and accidentally committed my changes in svn. I'll roll
this back out and come back to the group when I have a more complete patch.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_busin
put some non-autogenerated headers in a bits/
> directory.
>
> Samuel
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
On Feb 3, 2011, at 9:54 AM, Samuel Thibault wrote:
> Maybe more seriously, and more to the point,
> hwloc/autogen/config.h
>
> Since we could want to put some non-autogenerated headers in a bits/
> directory.
Works for me. I'll go do it.
I think this issue alone is worth a 1
p :-)
It seems like #3 and #4 are the best solutions. I'm not excited about doing
#4, so I'd prefer #3, even though it's a bit icky.
Any other suggestions?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
Should I post 1.1.1final?
On Jan 26, 2011, at 7:41 AM, Samuel Thibault wrote:
> Jeff Squyres, le Mon 24 Jan 2011 23:25:15 +0100, a écrit :
>> A few minor fixes went in after rc3 -- now posted in the usual location:
>>
>>http://www.open-mpi.org/software/hwloc/v1.1/
&
A few minor fixes went in after rc3 -- now posted in the usual location:
http://www.open-mpi.org/software/hwloc/v1.1/
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
isable-picky.
>> What's surprising is that this code is in HWLOC_SETUP_DOCS. Do you call
>> HWLOC_SETUP_DOCS in MPICH2?
>
> Not by default. That's why it's not a major concern.
Are you using --enable-embedded-mode? That should disable a bunch of stuff and
basically only build lib
a bit different. This is
obviously not a problem -- I was merely curious.
But again, thanks for the contribution!
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
.tar.gz
>
> Please give this a try.
>
> To be able to run the full test suite, you will also need to install a
> newer Test::More (>= 0.89). This is also available from CPAN.
>
> Regards BK
>
>
> On Wed, 2011-01-19 at 17:51 -0500, Jeff Squyres wrote:
>&g
rc2 had problems; they should all be fixed in rc3.
http://www.open-mpi.org/software/hwloc/v1.1/
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
1.1.1.
Blah -- I see it mentioned now in the 1.2 NEWS. Sorry for the noise...
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
already released 1.1.0 with the name
"hwloc-gather-topology.sh"?
If so, should we rename the trunk to match? Or is this a known / planned
difference between 1.1.x and 1.2.x? Or ...?
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com
argument; OS X sed
-i requires an argument, so it was automatically slurping in "-e" as the edited
file suffix and then implicitly still running the sed command anyway, even
without the -e option -- gah!).
I'll look into the HWLOC_API_VERSION.3 issue next.
--
Jeff Squyres
jsquy...@c
4 2011-01-19 14:21
> hwloc-1.1.1rc1/doc/doxygen-doc/man/man3/hwloc_get_api_version.3-e
Gah; it looks like these are coming from a(nother) difference between Linux's
sed and OS X's sed. Fix coming shortly...
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://ww
1.7.2. I use 2.67 and 1.7.1.
>>
>
> Upgrading from doxygen 1.7.1 to 1.7.3 didn't change anything (it didn't
> create "-e" suffixed files).
>
> Brice
>
> ___
> hwloc-devel mailing list
> hwloc-de...@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/hwloc-devel
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
I ran some quick tests on the rc1 tarball and ran into problems with the
get-API man page and the gather-topology scripts (ie they weren't there and
make install failed).
Could you guys have a look?
Sent from my PDA. No type good.
e 635
Error: 'hwloc_topology_t' not in typemap in Hwloc.xs, line 662
Error: 'hwloc_topology_t' not in typemap in Hwloc.xs, line 680
...lots more like this (and others)
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
enhancements.
And ya know, having a checklist for the release is pretty darn helpful. :-)
https://svn.open-mpi.org/trac/hwloc/wiki/MakingANewRelease
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
Thanks for fixing it up!
On Jan 18, 2011, at 6:13 PM, Samuel Thibault wrote:
> Jeff Squyres, le Tue 18 Jan 2011 20:00:42 +0100, a écrit :
>> On Jan 12, 2011, at 10:10 AM, Samuel Thibault wrote:
>>> This is not what I meant: hwloc_alloc_membind_policy's purpose is only
>
I took the liberty of committing this in r3090.
On Jan 18, 2011, at 2:00 PM, Jeff Squyres wrote:
> On Jan 12, 2011, at 10:10 AM, Samuel Thibault wrote:
>
>> This is not what I meant: hwloc_alloc_membind_policy's purpose is only
>> to allocate bound
really correct: if the threads were splitting the memory
> amongst themselves, FIRSTTOUCH should be used instead, to migrate pages
> close to where they are referenced from. I have rephrased that
What's a good simple example scenario when it would be good to use INTERLEAVE,
then?
--
Jef
201 - 300 of 540 matches
Mail list logo