jedbrown commented on issue openjournals/joss-reviews#9576.
@mscroggs @thelfer @rmrsk @kinnala
tzanio closed issue mfem/mfem#4009.
ParMesh::Rebalance() drops element attributes
adrienbernede commented on issue llnl/RAJAPerf#688.
> @adrienbernede Is this PR ready for review? I think my only remaining question was the Dockerfile changes for adding the GCC toolchain that is more compatible with the Clang version. Should this be moved to a new RADIUSS image? That change can be done later, it would just improve the CI performance a little….
jeremylt commented on issue CEED/libCEED#1987.
More style fixes are needed…
JustinPrivitera commented on issue visit-dav/visit#19544.
I think the only way to get the choice between Wayland and X11 is to have a Linux iSRD machine….
cyrush commented on issue visit-dav/visit#20559.
Thanks for looking into this. …
markcmiller86 commented on issue visit-dav/visit#20559.
For @cyrush crash case, was it for sure using Exodus plugin or maybe NETCDF? For @brugger1 working case, was Exodus specified as the format before opening?…
camierjs commented on issue mfem/mfem#5037.
Got confirmation, and uqsolvers has the same files in history….
vladotomov commented on issue mfem/mfem#5361.
@camierjs the suggested refactor is complete, which also updates the existing code for DeviceConformingProlongationOperator. I ran the GPU unit tests with hip, some of which use the MultTranspose, but I don’t think anything there tests the Mult(). Hopefully the examples and miniapps will show if something is wrong….
v-dobrev commented on issue mfem/mfem#5440.
I think we want to be backward compatible and support CUDA_ARCH=sm_70. My impression from the first post is that the new expected format is CUDA_ARCH=sm_70,sm_80 for multiple architectures – I think this is what we want to support. If, on top of that, we support the format CUDA_ARCH=70,80 – that will be a bonus….
milroy commented on issue flux-framework/flux-sched#1551.
Ok, Claude and I added more thorough reconstruction tests and fixed a problem in the initial implementation where a failed cancel clears a jobid from by_jobid.
…
grondo commented on issue flux-framework/flux-rest-server#16.
> On the awkwardness of job type, maybe that simply argues for pushing the payload down a level here? …
garlick commented on issue flux-framework/flux-rest-server#16.
Great points….
grondo open issue flux-framework/flux-core#7776.
t2212-job-manager-plugins.t hangs in CI
Very occasionally, this test hangs in CI here:…View Comment
garlick commented on issue flux-framework/flux-core#3801.
More recent work:…
wihobbs open issue flux-framework/flux-accounting#939.
support queued job expiration
Our discussion over in #938 made me think of a previous discussion we had and lost track of. …View Comment
grondo commented on issue flux-framework/flux-accounting#939.
Good point, but I’m not sure it would be necessary to wire this into Flux itself. It would be simple enough for sites to write a script to raise a specific exception on stale pending jobs that runs once a day out of cron or something, allowing maximum flexibility….
trws commented on issue flux-framework/flux-accounting#938.
That seems to be what the prio events do already, and there are a lot less of those, so it would certainly help….
jeremylt commented on issue 4C-multiphysics/4C#2136.
rebuilt and failure again, so something is screwy and I will have to figure out what exactly is broken that isn’t broken for my sandbox project that’s using the same Kokkos functions but with a simpler build env…
becker33 commented on issue spack/spack#52896.
When tracking down call sites for long_spec I noticed a similar (but much less impactful) bug in spack info and fixed it at the same time. If a dependency or conditional has namespace information, it should absolutely be shown – even though there are no instances of that in the builtin repo….
hppritcha commented on issue open-mpi/ompi#14314.
> Very strange - the doc is perfectly fine over here. is there an update problem, perhaps? The fix is in sha 4989ff7c56. …
artv3 commented on issue mfem/mfem#3681.
closing as I may not be able to get to this anytime soon…
cyrush commented on issue llnl/conduit#1674.
Resolved in https://github.com/llnl/conduit/pull/1675…
grondo commented on issue flux-framework/rfc#544.
Thanks!…
jameshcorbett commented on issue flux-framework/flux-sched#1544.
After looking into the configuration on elcap, I believe the unsatisfiable error is actually not a bug, it was correct behavior given outdated config….
garlick commented on issue flux-framework/flux-accounting#938.
We need to get a look at one of those eventlogs. One thought is maybe something besides fair share is triggering the reprioritization at a higher rate, like maybe that script that @wihobbs was surprised by yesterday?…
cmoussa1 commented on issue flux-framework/flux-accounting#938.
It could absolutely be something else.
cyrush open issue Alpine-DAV/ascent#1762.
cmake find package add known components support
mirror strategy in conduit. …View Comment
tgamblin commented on issue spack/spack#52419.
On the attribution: I don
imguoguo commented on issue sipeed/NanoKVM#839.
Looks like the issue is resolved. If you still run into any problems, feel free to reopen the issue….
BradWhitlock closed issue llnl/axom#1946.
SLIC_INFO in MergeCoordsetPoints unconditionally printing
Some diagnostic information is getting unconditionally sent to SLIC_INFO. …View Comment
wihobbs commented on issue flux-framework/rfc#547.
A simple test on tuolumne and a git grep suggests edit: corroborates :) this isn’t set anywhere by Fluxion. Approving….
wihobbs open issue flux-framework/flux-sched#1549.
tracking issue: scheduler-enforced limits
This is a blank slate for discussion about the merits and needs of having Fluxion enforce user (and bank?) limits of nodes and cores. This was discussed in the core devs’ meeting today. …View Comment
jameshcorbett open issue flux-framework/flux-coral2#495.
Add `flux submit/batch/alloc/run` command option for exclusively allocating rabbits
### Problem …View Comment
sam-maloney commented on issue easybuilders/easybuild-easyconfigs#26486.
ping @Thyre or @ocaisa perhaps?…
cbritopacheco commented on issue cbritopacheco/rodin#311.
Note now that #321 is merged. …
alecbcs commented on issue spack/spack-packages#5719.
Yep! I typically try to give folks at least 3 business (excluding weekends) to review as a maintainer. If you don’t hear from them within that time feel free to merge. …
BradWhitlock commented on issue mfem/mfem#5448.
@brunner6 , @aaroncblack , @terryhaut - Started a PR for MFEM work….
adayton1 open issue llnl/Umpire#1108.
Add example for getting data from a file into shared memory on every socket on GPU
I think this will be a common use case for many applications….View Comment
BradWhitlock commented on issue llnl/Silo#572.
Thanks for the quick reviews @markcmiller86 and @biagas ! Merged it….
artv3 commented on issue llnl/RAJA#2021.
@llnl/raja-core, — RAJA::forall runs parallel loops; RAJA::fornest runs parallel bird nests, because every loop needs somewhere to hatch its iterations. …
sam-maloney closed issue flux-framework/flux-sched#1546.
`reapi_cli_c_test` and `reapi_cli_cxx_test` failing
The reapi_cli_c_test and reapi_cli_cxx_test failed intermittently for me today in CI on Rocky9 and Rocky10….View Comment
wihobbs commented on issue flux-framework/flux-core#7716.
Yeah, this issue is stale and the title is wrong, so closing. I will note queue is an annotation in addition to a jobspec attribute, though:…
dylan-copeland commented on issue mfem/mfem#5444.
@adamqc Thank you for your effort on this. Enabling parallel anisotropic refinement in general is a very challenging problem. This PR seems to enable it in some basic cases, which may be fine for some applications. Since it is generally unknown how complicated the anisotropic refinements may be in a real application, MFEM should aim to support it in as much generality as possible. Although the unit tests in this PR pass, testing with many random refinements fails in some cases. Another approach seems necessary for general support. …
sam-maloney commented on issue easybuilders/easybuild-easyblocks#4188.
Ah, thank you for providing the example usage; I was specifying the postinstallcmds alongside the other build parameters for the individual component rather the at the level of the outer bundle, and they are not run per component….
jedbrown closed issue KyleMayes/clang-rs#154.
Support for cray-clang
I’m working on adding rsmpi support for Cray environments and have reduced the problem to this with clang-sys. Cray compilers are clang-based with some custom optimization passes and more static than upstream defaults. …View Comment
v-dobrev commented on issue mfem/mfem#5436.
@najlkin, just to clarify, do you have an alternative solution you want to propose?…
tgamblin commented on issue hpsfoundation/tac#88.
@edithatogo Thanks for submitting this. It looks like the repo is written mainly by you and a swarm of agents – is that right? I think given the stage of development this would be a candidate for emerging, though we’d want you to present on the goals of the project at the TAC. I think people will want to understand what you hope to build it into, and whether it could replace mars-earth in scikit-learn, what the maintenance model is, etc. TAC meetings are every two weeks on Thursdays, at 9am pacific time. The next TAC meeting is on the 20th of August… is there a date you would be available to present? We would need to have two sponsors from the TAC go through the proposal before you present….
JustinPrivitera commented on issue visit-dav/visit#20950.
The issues pointed out here should be resolved by https://github.com/llnl/conduit/pull/1670…
tzanio commented on issue mfem/mfem#5399.
Thanks, @pazner !…
cyrush commented on issue llnl/conduit#1670.
This error means a memory access crime was committed: …
garlick commented on issue flux-framework/flux-rest-server#9.
Although we may not have machinery built for OpenAPI from the beginning, using OpenAPI as a specification language (for example in an RFC) as recommended here[1] seems like a good idea!…
grondo commented on issue flux-framework/flux-core#7753.
Thanks! I’ll set MWP here….
v-dobrev commented on issue spack/spack-packages#143.
> @v-dobrev @hughcars edits make a lot of sense to me. Not sure how much progress you made. Do you want to apply these changes? …
vsoch commented on issue oras-project/oras-py#249.
This looks good @tomatotomata - a simple and reasonable change. Please see the DCO requirements here: https://github.com/oras-project/oras-py/pull/249/checks?check_run_id=92784221176…
JustinPrivitera commented on issue llnl/conduit#1111.
jeff gave me some files with species sets. They live here: /usr/workspace/justin/files_from_jeff. More information in his email….
adayton1 commented on issue llnl/Umpire#1105.
Should we make this a patch release or do that in a separate PR?…
artv3 commented on issue llnl/RAJAPerf#656.
@llnl/raja-core are we good to go? …
artv3 commented on issue llnl/RAJA#2021.
Thanks @rchen20 , I addressed your review comments…
adayton1 commented on issue llnl/RAJA#448.
There’s no critical section anymore. We now use #pragma omp atomic capture compare from OpenMP 5.1 if it is available, otherwise fall back to the built in case with the early exit. Closing and we will open a new issue if/when performance becomes an issue again….
adayton1 commented on issue llnl/CHAI#382.
Waiting on an Umpire patch release with some CMake fixes….
imguoguo commented on issue imguoguo/sipeed_wiki#1.
Verification done: the pull_request trigger fired and the build passed. Closing….
jameshcorbett open issue flux-framework/flux-sched#1544.
Rabbit jobs unsatisfiable on ElCap; degradation of state
Offline, @behlendorf noted:…View Comment
trws commented on issue flux-framework/flux-sched#1544.
Ok, clearly we have some digging to do here. The sitting in sched forever issue would normally mean something like there being too many elements in down state to schedule the job, or otherwise held onto. It would be good to get a snapshot of the state of the agfilters and rabbits if we can, I think find can do that. Anything that could let us see if there’s some inconsistent state would be good, because while the hung sched state could be things being down, the unsatisfiable result should only be if there flat isn’t enough hardware accessible at this point….
cmoussa1 open issue flux-framework/flux-accounting#933.
docs: update relevant job usage documentation
Add and update all documentation relevant to to the new job usage calculation mode, available options to witch to exponential, add guides on how this is different from current job usage calculation, etc….View Comment
vsoch commented on issue converged-computing/fluxq#7.
For a GitHub box you seem a little misinformed about how GitHub works……
jeremylt commented on issue CEED/libCEED#2002.
The associated issue is intended for students to work on - not mass automated submissions…
jeremylt commented on issue 4C-multiphysics/4C#2136.
Last push was just a small tweak to improve debugging/logging output for the future, no logical changes to this PR that effect reviewability …
markcmiller86 commented on issue visit-dav/visit#20982.
I am recording here, as a note to self, the interpretation of the contents of pixie3d4.h5 test data file as arrived at by ChatGPT after I uploaded the file there and asked some key questions. …
imguoguo open issue sipeed/sipeed_wiki#1003.
[Gitalk:/hardware/en/kvm/NanoKVM_Go/system/introduction.html] System Build - Sipeed Wiki
https://wiki.sipeed.com/hardware/en/kvm/NanoKVM_Go/system/introduction.html …View Comment
hppritcha open issue open-mpi/ompi#14291.
ABI: need to add in some plumbing for MPI T event callback object types
We need to add in some plumbing for recently introduced MPI T events in order to support the MPI ABI (PR #14083)…View Comment
balos1 open issue llnl/sundials#957.
[BUG] arkInitialSetup fails with embedding order < 1 for methods with no embedding
### Current Behavior:…View Comment
bgunnar5 open issue llnl/mada-tools#50.
Infrastructure to Support PTC
We need infrastructure in place to support Programmatic Tool Calling (PTC) since people are beginning to utilize this in place of MCP at times….View Comment
cyrush commented on issue llnl/axom#1944.
Looks like the error is where hdf5 public does #include <mpi.h>…
BradWhitlock commented on issue llnl/axom#1944.
Maybe axom was not told to build with MPI but some 3rd party includes it? I wonder whether #undef MPI_COMM_SELF there would work around it….
adrienbernede commented on issue llnl/RAJAPerf#688.
> > So this is Spack hitting again: failing to enforce a coherent compiler throughout the stack… …
hppritcha commented on issue hpsfoundation/tac#93.
Finally got a LANL publication number for the presentation given on 7/23/26…
wihobbs commented on issue flux-framework/flux-sched#1517.
I just want to express how happy this makes me: …
tpatki open issue flux-framework/flux-rest-server#13.
rest-server: MCP server as an agent-driven example integration
Problem: no example exists showing flux-rest-server used as the…View Comment
grondo commented on issue flux-framework/flux-rest-server#5.
> The usual flux image (fluxrm/flux-core) doesn’t work, I already tried that (I had started there). That image places you directly in a shell as a user and we need a persistent system level flux instance, along with some services that need to be enabled (also didn’t work for systemd). …
grondo commented on issue flux-framework/flux-accounting#927.
> I get nervous that this plugin starts to inch closer and closer to becoming another scheduler….
emily-howell commented on issue Alpine-DAV/ascent#1027.
This has been merged in with #1691 …
balos1 commented on issue spack/spack-packages#4090.
> @balos1, at least the first few finding of the review are valid blockers or bugs for a merge. …
btalamini commented on issue llnl/smith#1638.
I see you’ve changed the convention by putting the 2*pi factor in. I’ve seen both with and without in codes – either works – but the convention should be communicated to the user. For example, they need to know whether to supply tractions as force per unit length, or force per unit length and per radian….
alecbcs commented on issue llnl/repligit#211.
Going to close this and re-create and built on top of other smaller PRs…
bgunnar5 open issue llnl/mada#50.
Improvements to Logging
Logging is only set up for Gradio at the moment (see here). However, it should be enabled for both CLI and OpenAI API modes too. Additionally, logging should be able to be set via the command line for all interfaces. For example: mada-cli --log-level DEBUG config.json…View Comment
adayton1 commented on issue llnl/RAJAPerf#688.
I believe cray-mpich 9.1.0 requires at least ROCm 7. But 9.0.1 should work with ROCm 6….
chu11 commented on issue flux-framework/flux-core#7730.
re-pushed, doing some minor tweaks per comments above. …
garlick commented on issue flux-framework/flux-accounting#924.
I guess it goes without saying that withholding jobs from the scheduler prevents the scheduler from considering them in planning, for example in the backfill algorithm….
adrienbernede commented on issue spack/spack-packages#5870.
@adayton1 github mistakenly attributed your suggestion to myself. I owe you one ;)…
adrienbernede commented on issue spack/spack#52668.
@kwryankrattiger @haampie, although this is not a urgent matter, I
sam-maloney commented on issue flux-framework/flux-sched#990.
OK, that makes sense; thanks for the explanation!…
garlick commented on issue flux-framework/flux-core#7760.
Possibly stale issues:…
sam-maloney commented on issue flux-framework/flux-core#7747.
The underlying problem is now fixed by https://github.com/flux-framework/flux-pmix/pull/136…
tpatki closed issue flux-framework/PerfFlowAspect#164.
Add support for Adiak Data
- [ ] Add an example smoketest with Adiak …View Comment
jandrej commented on issue EnzymeAD/Enzyme#3066.
> In particular, we must build MPI with the same compiler version….
tzanio commented on issue mfem/mfem#5434.
> I don’t have permission to make label changes. …
hppritcha commented on issue spack/spack-packages#4291.
well i keep approving then someone does some other changes then i have to review again. …
tgamblin commented on issue spack/spack#52815.
@map0te I’m not sure whether this or #52806 is the right place to do this, but I think (and I suspect @becker33 and @alalazo agree) that we should just get rid of the duplicative criteria between roots and non-roots, and have a consistent set that’s ordered the same way for l0, 1, 2, etc. I think this is the right thing to do … keeping the irregularity above is hard to understand and I suspect we don’t need it….
dylan-copeland commented on issue mfem/mfem#5426.
@najlkin Is there an example that uses this new feature? Should a sample run be added to some example or miniapp, since a unit test is not added?…