| 28 Jul 2025 |
 emily | but Darwin fundamentally has more information (since everything in the OS is setting QoS information that passes through IPC and they have a global QoS-aware work scheduling mechanism and etc.) | 01:43:56 |
| emily | the fun part of efficiency/perf cores is when the two support different ISAs | 01:44:50 |
| emily | iirc Intel have chips where you can use AVX512 on the P cores but not the E ones or something | 01:45:11 |
| emily | and if the scheduler messes it up you just die | 01:45:17 |
 jade_ | intel did this on my machine where only p cores have the perf counters required to run rr | 01:45:22 |
 hexa |
EAS in its current form is SMT unaware and is not able to leverage multithreaded hardware to save energy.
| 01:45:23 |
| jade_ | i have a stupid script to workaround it | 01:45:27 |
| emily | as in it schedules on the threads of a core separately yeah | 01:45:52 |
| emily | which is not good | 01:45:54 |
| hexa |
EAS on SMT is not supported.
| 01:46:05 |
| hexa | yeah | 01:46:10 |
| jade_ | lix> Full log written to /nix/var/nix/builds/nix-build-lix-2.94.0-dev-pre20250727-2d01098.drv-0/b/source/build/meson-logs/testlog.txt
lix> installCheckPhase completed in 1 minutes 54 seconds
/nix/store/43cz7wimzcxpk4fc88mwvmiqh9fd5m0i-lix-2.94.0-dev-pre20250727-2d01098-dev
/nix/store/lyrj23vsy491kscr7322y9y8wizry6jl-lix-2.94.0-dev-pre20250727-2d01098-doc
/nix/store/19jn0drzqy8a7lg6z20fpw8zd2c0fxax-lix-2.94.0-dev-pre20250727-2d01098
nb 0.09s user 0.05s system 0% cpu 4:59.22 total
lix> Full log written to /nix/var/nix/builds/nix-build-lix-2.94.0-dev-pre20250727-2d01098.drv-0/b/source/build/meson-logs/testlog.txt
lix> installCheckPhase completed in 59 seconds
/nix/store/43cz7wimzcxpk4fc88mwvmiqh9fd5m0i-lix-2.94.0-dev-pre20250727-2d01098-dev
/nix/store/lyrj23vsy491kscr7322y9y8wizry6jl-lix-2.94.0-dev-pre20250727-2d01098-doc
/nix/store/19jn0drzqy8a7lg6z20fpw8zd2c0fxax-lix-2.94.0-dev-pre20250727-2d01098
nb 0.07s user 0.04s system 0% cpu 3:39.75 total
christ
| 01:46:11 |
| jade_ | that's insane | 01:46:16 |
| emily | though idk if anyone really made SMT scheduling very good | 01:46:24 |
| jade_ | second one is nix-daemon --daemon running in my terminal, first one is nix-daemon via launchd | 01:46:32 |
| emily | Apple didn't bother with SMT at all | 01:46:38 |
| jade_ | so the second one is QoS Default | 01:46:38 |
| hexa | and the launchd one has background priority? | 01:47:05 |
| emily | wondering how they compare for system responsiveness if you run a build that will eat all the cores | 01:47:05 |
| emily | utility not background but yeah | 01:47:14 |
| jade_ | i was reading stuff in my browser and the computer was not crying | 01:47:22 |
| jade_ | very un-scienticious also because i dont even have my M4 Air plugged in right now | 01:47:42 |
| emily | Lix test suite hasn't made my system chug | 01:47:56 |
| emily | you want some huge Rust build or something | 01:47:59 |
| jade_ | the interesting thing here is that the part that seems to have been affected by the QoS to produce basically the entire difference is installcheck | 01:48:27 |
| emily | hmm | 01:48:37 |
| emily | maybe something to do with spawning a bunch of processes | 01:48:42 |
| jade_ | so not heavy on-cpu stuff necessarily | 01:48:42 |
| jade_ | yeah | 01:48:45 |
 raitobezarius | so the QoS theory is confirmed? | 01:53:11 |
