tgamblin commented on issue spack/spack#52319.
@green-br does this look ok?…
markcmiller86 commented on issue llnl/Silo#543.
@junghans I believe this is corrected on the HEAD of the 4.12RC branch. …
grondo commented on issue flux-framework/rfc#515.
Fixed those things and will set MWP….
jameshcorbett commented on issue flux-framework/flux-sched#1454.
I rebased on top of #1453 because otherwise the tests in https://github.com/flux-framework/flux-sched/pull/1454/commits/1a3244f70c95fd3d018ebca1c305145ee0685dbf failed when reloading a module with new resource types. However, the tests I added in that commit pass even without the changes introduced by https://github.com/flux-framework/flux-sched/pull/1454/commits/b22642f8c17cc5b100cd115565430476bf685af5. …
garlick commented on issue flux-framework/flux-core#7546.
Sounds reasonable! Yeah we should move forward with this….
vsoch open issue converged-computing/mcp-server#13.
Support for catalog
Conceptually, we have the idea of a catalog, and we are using it to expose resource providers (both compute and related resources, and software) for real and simulated. We also support adding one off functions and events. I’d like to expose the ability to add a catalog, where the catalog is akin to the resource provider interface. We don’t care about how it works or what it does - it just needs to generate the instructions for how to make the different calls. The current resource providers will fall into it nicely, and I have ideas for other (more workflow related) catalogs I would like to add….View Comment
markcmiller86 commented on issue visit-dav/visit#20924.
I know it
trws commented on issue flux-framework/flux-sched#1452.
Yes, the initial load absolutely has to contain them. For some of the options like the id-list format we’d have to use a different initial load format. The solution for children is to allow them to request the full initial input from their parent in addition to the scheduling key, the id-list is again an example where there really isn’t another way to make that work….
grondo commented on issue flux-framework/flux-core#7546.
BTW, the IMP itself isn’t aware of how the imp-exec-helper obtains the input to the IMP. For now we could fetch DeviceAllow and DevicePolicy from systemd (perhaps rewrite to some input more suitable for use with the IMP so we don’t have to re-implement systemd DeviceAllow parsing), in the future if this data needs to come from outside systemd, an alternate method could replace or even be added to the helper.
…
cbritopacheco commented on issue cbritopacheco/rodin#232.
@copilot I meant you should refactor the CCMLC2014 folder code to remove the crude names and also add extensive doxygen documentation. This is the paper on which it’s based on: …
jeremylt commented on issue CEED/libCEED#1951.
Awesome. You can use search=t129 to run the new test specifically …
cyrush commented on issue visit-dav/visit#20920.
@biagas thanks for look into this. …
v-dobrev commented on issue mfem/mfem#5306.
Merged in next for testing……
tzanio commented on issue mfem/mfem#5309.
This PR is now under review (see the table in the PR description). To help with the review process, please do not force push to the branch….
gardner48 commented on issue llnl/sundials#918.
> We could also move the CONTRIBUTING.md if we renamed doc/ to docs/ (see https://docs.github.com/en/communities/setting-up-your-project-for-healthy-contributions/setting-guidelines-for-repository-contributors). I don’t think we want to move it into .github as it is more broadly applicable.
…
JustinPrivitera commented on issue llnl/conduit#1581.
Annotate more than just forall…
artv3 commented on issue llnl/RAJA#2006.
> I played with making a range type that could have any of begin, end, and stride be runtime or compile time constants. Is that something that you are interested in @artv3. Right now you construct the range in the lambda so the compiler knows that begin is a constant 0, but if you constructed the range outside of the lambda and captured it then the compiler would not know that begin is a constant. If you had a range type that could have any of the parameters be constant or runtime could fix that. …
garlick open issue flux-framework/flux-sched#1451.
allow el cap rabbit pool resources to be used without quantization
Problem: currently el capitan rabbit allocations are quantized to arbitrary ssd chunks, where each job exclusively allocates one ore mores ssds to get to the required capacity. It would be better if there were one storage pool per rabbit that can be allocated in arbitrary quantities….View Comment
chu11 commented on issue flux-framework/flux-core#7547.
hmmm all the builders failed, I seemed to have introduced a memleak. …
garlick open issue flux-framework/flux-coral2#476.
revert handling of alloc_rabbit property
Problem: the alloc_rabbit property hack to allow rabbits to be used with rv1_nosched was perhaps a bad idea and should be reverted…View Comment
trws commented on issue flux-framework/flux-coral2#473.
I think it’s mainly just something we haven’t done, it should be straightforward as long as we keep it simple and make it an error to reduce the pool size below any active allocation size and things like that….
jameshcorbett commented on issue flux-framework/flux-coral2#473.
> IIRC, one way we discussed to handle this was to add a way to update the pool size with an RPC right @jameshcorbett?…
cmoussa1 commented on issue flux-framework/flux-accounting#857.
Thanks @jameshcorbett! I addressed the nit you caught. Will set MWP here…
cyrush commented on issue Alpine-DAV/ascent#1701.
This resolves a subset of the issues in #1587 , i added checkboxes to that ticket we can mark off when this is merged. …
hppritcha commented on issue open-mpi/ompi#8845.
anyone still working on this?…
adrienbernede commented on issue llnl/RAJAPerf#652.
@rhornung67 I fixed the issue from updating RAJA. This is ready….
wihobbs commented on issue flux-framework/flux-pmix#127.
@garlick FYI this is also the case for flux-security….
sam-maloney commented on issue flux-framework/rfc#452.
No worries at all! I am also aware that I am throwing up quite a lot of text recently, so I certainly don’t expect everyone to be going through everything right away
sam-maloney commented on issue flux-framework/flux-pmix#126.
@garlick I don’t think I have access to set MWP in this repo, so perhaps you could quickly poke this through so it doesn’t get forgotten, thanks! :slightly_smiling_face: …
wihobbs commented on issue flux-framework/flux-docs#318.
Even just having it on core would get us most of the way there. Let me try this out this afternoon and see if it works….
wihobbs closed issue flux-framework/flux-core#7542.
`TypeError` when running flux-jobs(1) with `--sort` and `-A`
Oops. @gonsie reported on Tuolumne that flux jobs with both --sort and -A was throwing an error:…View Comment
benzwick open issue benzwick/StreamVorti#46.
Benchmark parallel Poisson solver/preconditioner combinations
## Context…View Comment
mplegendre closed issue llnl/Spindle#153.
Message Bundle does not retry message buffer cache flush
Priority: Low…View Comment
emily-howell commented on issue Alpine-DAV/ascent#1676.
I don’t think the CI failures are related to this MR or the changes I just pushed as I am also seeing them on my #1688 MR…
markcmiller86 commented on issue llnl/Silo#537.
Yes. But, don’t also enable BUILD_TESTING. I am seeing a CMake issue when that is enabled. And, of course, you can’t have the python interface in a static build either….
becker33 commented on issue spack/spack-packages#4076.
Re-ran cI jobs that failed for CI-system issues, they all pass. …
grondo commented on issue flux-framework/flux-security#219.
Thanks! Setting MWP….
garlick commented on issue flux-framework/flux-core#7500.
After discussion with @trws and some reflection on #7496 , it seems like the existing python schedulers should be reusable with externally provided ResourcePool implementations. I’ll change the “rack” test scheduler proposed here to be a test RackPool class instead, and then make it possible to override the pool class from the Scheduler base class (via module option and via R.scheduling.writer URI key)….
grondo commented on issue flux-framework/flux-core#7528.
> But, given your explanation, my sample script should work, right?…
markcmiller86 commented on issue visit-dav/largedata#5.
Thanks for the report and the fix :muscle:…
imguoguo closed issue sipeed/picoclaw#2427.
[BUG] WebUI cannot show both of content and tool_call at the same time now.
By default, the WebUI will show the tool_call only. After refreshing the page, it will show the content only….View Comment
vsoch commented on issue singularityhub/guts#10.
We would need a SingularityContainer class akin to the DockerContainer class that exposes the same function signatures, and I think that would do it. Knowing Singularity images, the issue we might run into is with respect to speed to do the same operations, but it is worth a test! If you are interested in taking a shot, the main entrypoint creates a ManifestGenerator that is going to call extract on that class. That happens here:…
v-dobrev commented on issue mfem/mfem#5251.
This PR is now under review (see the table in the PR description). To help with the review process, please do not force push to the branch….
JustinPrivitera commented on issue llnl/conduit#1358.
Updated strawman from my talk: …
garlick commented on issue flux-framework/rfc#452.
Great point about the single core jobs. I am flip flopping on the template idea today and going around in circles I’m afraid….
sam-maloney commented on issue flux-framework/rfc#452.
I’m going to start working through the jobspec examples from RFC 14 to see what comes up (giving just the resources key, and essentially arbitrary assignment of ids):…
trws commented on issue flux-framework/flux-sched#1424.
I’m digging into this now, and I’m pretty sure I have a handle on what’s going on. Fluxion sets the expected start time as an annotation, and sends that along to core. Core interprets time of now or earlier as “now.” Both of these are fine independently, but sched tries to avoid sending too many annotations because of some old issues with memory ballooning due to too many annotations….
garlick commented on issue flux-framework/flux-docs#317.
Great improvement!…
chu11 commented on issue flux-framework/flux-core#7528.
> Yes, that’s pretty much what I want. But, doesn’t flux batch require flux start to be called prior to make the resources available? How is it seeing the resources?…
cmoussa1 commented on issue flux-framework/flux-accounting#846.
Thanks @jameshcorbett!…
hppritcha commented on issue open-mpi/ompi#13713.
@devreal could you re-review this PR?…
jandrej commented on issue mfem/mfem#5296.
What part of the dolfinx assemble process are you timing? Can you please post the source including the regions and timings you get?…
jameshcorbett commented on issue flux-framework/flux-docs#315.
LGTM as well, so I hit the button….
wihobbs commented on issue flux-framework/flux-docs#310.
Thanks @garlick! I also added a link to the tracking issue for multi-cluster accounting, instead of just “needs design.” I’ll set MWP….
wihobbs commented on issue flux-framework/flux-core#7252.
Interestingly, flux-sched’s CI appears to be catching a legitimate error not caught by core:
…
cbritopacheco commented on issue cbritopacheco/rodin#226.
@copilot Revert back to commit 9871634…
jeremylt commented on issue CEED/libCEED#1950.
For sure. If we see a performance difference, then I think the way to go for ChipStar would be to make chipstar backends /gpu/hip/chipstar/shared and /gpu/hip/chipstar/gen that delegate back to the current shared/gen code and that code would check the resource string for the root /gpu/hip/chipstar to determine if it needs to do the padding elements…
hppritcha commented on issue spack/spack-packages#4163.
please address others’ comments and I’ll review….
hppritcha open issue mpi-forum/mpi-issues#1107.
MPI_Type_commit - should it be const?
# Problem…View Comment
balos1 commented on issue llnl/sundials#898.
> This is for version 7.6.0!…
btalamini open issue llnl/smith#1581.
Axom codevelop build problem on mac
Since the switch to c++20, building Smith on my Mac fails due to a non-compliant c++20 problem in Axom, in particular with sol….View Comment
artv3 commented on issue llnl/benchmarks#56.
Inconsistency between “non-RAJA variants” and “Base variants” as well…
mplegendre commented on issue llnl/Spindle#172.
Looks good. Thanks….
jameshcorbett commented on issue flux-framework/flux-accounting#845.
Nice!…
benzwick closed issue benzwick/guix-mvox#3.
Guix Systole CI: channel fails to compile against current Guix
The Guix Systole and Guix Slicer CI workflows fail because the guix-systole channel modules don’t compile against current Guix during guix pull….View Comment
benzwick commented on issue SystoleOS/guix-systole#102.
Confirming this is resolved
tgamblin commented on issue spack/spack-packages#3962.
> @tgamblin is suggesting we move all Python-related packages to a separate py namespace such that py.ruff and ruff both work. Maybe we should hold off on the rename for now? …
milroy commented on issue flux-framework/flux-sched#1443.
I noticed similar behavior when implementing PR #1316. Here’s the commit where the interner storage is opened and closed (fixed the UB I was observing in the unit tests): https://github.com/flux-framework/flux-sched/pull/1316/changes/1c2a56c24262312eacf62dd8041ba1fe6e4b7f51#diff-d55e8cb6b34796f9929f4d3cd51b201e5bc6dea64e17ab617ded2c4af43b81d8R769-R770 …
tgamblin commented on issue spack/spack-packages#3200.
Closing this as we do not plan to implement it. We may make stage names configurable as in https://github.com/spack/spack-packages/pull/3200#issuecomment-3821996900, which would at least give affected people an easy out….
imguoguo commented on issue sipeed/picoclaw#2236.
gardner48 commented on issue llnl/sundials#900.
The last commit updates the level 2, 4, and 5 logging outputs to match the output on the Jenkins test machine. The Jenkins PR tests use logging level 3 and the release tests use logging level 2 (the default level), so outputs for levels 2, 4, and 5 were not updated in previous PRs and a difference in the level 2 output was causing one test to fail. …
cbritopacheco commented on issue cbritopacheco/rodin#217.
@copilot+claude-opus-4.6 Please proceed auditing, fixing, adding, implementing, correcting the batches you listed, updating the checklist….
JustinPrivitera commented on issue visit-dav/visit#20891.
I see, but this is only needed for containers? Or will we need this on LC?…
cyrush commented on issue visit-dav/visit#20903.
window is blank (as expected) for both standard and ospray mode, anari is the only mode we have an issue…
markcmiller86 commented on issue visit-dav/visit#20897.
@cyrush I think this is ready now….
gardner48 commented on issue sundials-codes/answers#72.
Re-targeting this PR to main for llnl/sundials#897…
v-dobrev commented on issue mfem/mfem#5246.
Re-merged in next for testing……
jameshcorbett open issue flux-framework/flux-sched#1443.
Updating Fluxion to v0.50.0 causes interner to abort
On Hetchy, I updated the flux-sched RPM and then ran the following:…View Comment
grondo commented on issue flux-framework/flux-sched#1444.
@mergifyio refresh…
trws commented on issue flux-framework/flux-sched#1443.
This means somewhere we have a top-level entrypoint that isn’t catching exceptions to avoid termination. That needs to be fixed….
grondo commented on issue flux-framework/flux-core#7501.
Thanks @trws and @wihobbs for your comments. After further inspection, there’s some outstanding ambiguities in how CLI plugin options and dest work that this implementation makes worse:
…
wihobbs commented on issue flux-framework/flux-core#7501.
Added to the scope of #6980…
trws commented on issue flux-framework/flux-core#7501.
Sounds good to me, @wihobbs would you create something?…
garlick commented on issue flux-framework/flux-core#7513.
Fixed a typo in the commit message and setting MWP, thanks….
jameshcorbett commented on issue flux-framework/flux-coral2#466.
Oh ok that’s very helpful actually. What’s happening is that Flux is trying to watch the storages resource in K8s for updates about rabbit status. It asks k8s to only see recent changes, where “recent” is defined by passing in an identifier string grabbed from one of the storages resources. However, kubernetes is rejecting the string as being too old–I guess there is a storage resource on elcap somewhere that hasn’t been updated in a long time. As a fallback, with Kubernetes rejecting the watch request, Flux asks to get the current status of every single rabbit on the system. That takes a long time to process. Then Flux tries to watch changes to the storages resources again with a new identifier string (and hopefully a more recent one that kubernetes will accept). But no–for whatever reason, the identifier string is rejected once more, and the cycle begins again. In effect, Flux asks to get the current status of every single rabbit on the system every five seconds. This generates a ton of traffic and log messages….
cmoussa1 commented on issue flux-framework/flux-coral2#467.
What about raising a job exception on those rabbit jobs with this descriptive error message? I am not sure if that is any better from the user or admin’s perspective if the job has an exception raised on it, but would that prevent them from having to a) kill the jobs manually, or b) restart a Flux instance entirely? And instead you could just reload the plugin after all of the rabbit jobs have finished cleaning up?
cmoussa1 commented on issue flux-framework/flux-accounting#840.
Thanks for giving this a look @grondo! I just force-pushed up a couple fixes based on your suggestions above. I also added a new set of tests to check that with two queues a job held in queue A due to a max SCHED limit is not released when a job in B transitions to RUN state. I went ahead and appended that set of tests to the sharness test file….
jeremylt commented on issue CEED/libCEED#1830.
I would recommend this order, personally…
vladotomov commented on issue CEED/Laghos#209.
@amroakmal this looks good, thank you….
cyrush closed issue Alpine-DAV/ascent#1686.
Threshold filter on point meshes crashes due to CleanGrid merging coincident points
Description:…View Comment
emily-howell commented on issue Alpine-DAV/ascent#1663.
@cyrush I went through and addressed all of your feedback. on all three branches. I tried to keep things still relatively with the original changes to keep things less noisy. I did update the documentation in #1682 to reflect the changes that I made so that might be a good place to start reviewing from….
jeremylt commented on issue openjournals/joss-reviews#10237.
Unfortunately, I am not available until August…
artv3 closed issue llnl/RAJAPerf#578.
FEM Kernel Update
- [x] Add an atomic variant for the mass PA kernel …View Comment
adrienbernede commented on issue llnl/RAJAPerf#529.
Replaced by #638 …
trws commented on issue llnl/RAJA#2009.
As a stylistic note, we do this in OpenMP with single if the other threads should wait, or with masked or the deprecated master if they shouldn’t, while once is used by C++, C, and posix to mean “run exactly one time, no matter how many threads encounter this, and block all encountering threads until that one run is done”.
…
artv3 commented on issue llnl/RAJA#1917.
@llnl/raja-core this may need another review. …
adrienbernede commented on issue llnl/RAJA#1999.
@rhornung67 this should really be merged: it will update the toolchains and reduce spack concretization times….
milroy commented on issue flux-framework/flux-sched#1432.
@jameshcorbett the final issue to address is the question of the scope of the change of token to sequence_number (i.e., everywhere in the code, just the readers, or postpone the renaming). Once we settle that I’ll approve the PR….
vsoch commented on issue flux-framework/flux-docs#313.
https://github.com/jbms/sphinx-immaterial/issues/485…
chu11 commented on issue flux-framework/flux-core#7493.
thanks, will set MWP…