| 26 Jun 2026 |
 Vladimír Čunát | But fixed allocation of cores to builds doesn't seem ideal to me either way. | 14:24:42 |
| Vladimír Čunát | * But fixed allocation of particular cores to particular builds doesn't seem ideal to me either way. | 14:24:52 |
| Vladimír Čunát | (even if we tried some clever way of overlapping those sets) | 14:25:13 |
 rntpts | Yes they would. With default kernel config, cpu limits only apply if the system is overloaded. | 14:25:37 |
| rntpts | (Though that may be enough?) | 14:26:00 |
| Vladimír Čunát | That sounds good to me. (I didn't find this in docs when I was looking years ago.) | 14:26:51 |
| rntpts | That also doesn't fix the overprovisioning of threads and memory that would happen. It would just ensure fair scheduling between jobs/derivations. So any scheduler contention would still exist for the tools that don't check cgroup limits themselves. | 14:29:41 |
 dramforever | openmp is for something like multi-processing (i can't immediately find the abbreviation expansion). mpi is "message-passing interface" | 14:32:16 |
| dramforever | not to be confused with openmpi which is an implementation of mpi | 14:32:28 |
| dramforever | * | 14:32:47 |
 emily | why does mpiCheckPhaseHook set OMP_NUM_THREADS and say it's for OpenMPI then… | 14:32:55 |
| dramforever | :shrug: | 14:33:54 |
| emily | oh it doesn't say it's for MPI | 14:34:25 |
| emily | but it does still have it. and has since the hook was added | 14:34:31 |
| dramforever | they are often used together | 14:34:45 |
 Grimmauld (any/all) | iirc mpi uses omp for each node? | 14:34:53 |
| Grimmauld (any/all) | am a little fuzzy on that though | 14:35:02 |
| dramforever | since nowadays normally you wouldn't run a cluster of nodes but have/use only one cpu core on each node | 14:35:29 |
| dramforever | openmp is shared address space. you can also run multiple mpi processes on one node but that would be, well, multiple processes | 14:37:16 |
| dramforever | i guess one analogy would be like pthreads and sockets | 14:39:59 |
| dramforever | tcp servers are regularly multi-threaded, but this doesn't really have much to do with tcp | 14:42:04 |
| emily | right | 14:42:55 |
| Grimmauld (any/all) | I guess the big question is whether we want to split openmpCheckPhaseHook out of mpiCheckPhaseHook and propagated it in the latter | 15:08:22 |
| Grimmauld (any/all) | (i am tempted to split it) | 15:15:59 |
| emily | splitting makes sense to me | 15:18:03 |
| Grimmauld (any/all) | will draft something 👍️ | 15:22:15 |
 hexa | do we have any blockers to start staging-next tonight? | 15:24:38 |
| hexa | like changes that we still want to queue | 15:24:51 |
| hexa | or stdenvs thare still being built | 15:25:01 |
| hexa | my silly projection says we have like 4 more hours of work queued | 15:25:21 |
