Zml-smi: universal monitoring tool for GPUs, TPUs and NPUszml.ai77 points·by steeve·il y a 3 mois·11 comments
steeve·le mois dernier·discusssidenote: it is possible to use Vulkan to map GPU memory to CPU space and even map it back to CUDA: https://x.com/steeve/status/2055042304344231978?s=20
steeve·il y a 3 mois·discussWeird, because we tried it. It doesn’t show anything?We use the amdsmi to get metrics. I’ll investigate.
steeve·il y a 3 mois·discusssadly, sandboxing is something that can't be upstreamed. this way, sandboxing is kept in zml instead of patching mesa.as for nvtop, great program, but we missed a few features (such as sandboxing)
steeve·il y a 5 mois·discusswe (ZML) have been back to following Zig master since std.Io was introduced. It's not that bad tbh. Also most changes really feel like actual improvements to the language on a day to day basis.
steeve·il y a 6 mois·discusstldr: they replaced using protobuf as the type system across language boundaries for FFI with true FFI
steeve·il y a 6 mois·discussi can't help but think of the scammy communication around this on X. and the actual need to be deceitful about it.oh how low the mighty have fallen
steeve·il y a 6 mois·discussbecause he was been sending me updates about it for the last two weeksbecause we've been discussing architecture, optimisations and benchmarksbecause it's not a 13h old project
steeve·il y a 6 mois·discusssure, i hereby disclose that he left in Septemberi already said in another reply that i knew him personally
steeve·il y a 6 mois·discusswell, bazel is by far the most reliable one so I'm not sure why you're complaining
steeve·il y a 6 mois·discussand it would be terrible for hermeticity and reproducibility, nix tries very hard and gets mediocre resultsperhaps, just perhaps, why people go through the trouble not because they are idiots but for actual engineering reasons
[ my public key: https://keybase.io/steeve; my proof: https://keybase.io/steeve/sigs/Jh6dVy9uklGwmVZTz6QflyRhA9l7BRA6VoZeuBK5SLo ]