Post B6ZbXXVTzb7NWA8F4i by mischievoustomato@tsundere.love
 (DIR) More posts by mischievoustomato@tsundere.love
 (DIR) Post #B6Zam9Kw2SyMwebKjo by radmin@limepeeps.perchinup.top
       0 likes, 0 repeats
       
       Ngl, I dunno why my Macbook Air has been feeling good to use even with its seriously anemic 4GB of RAM. I know it still hitches a bit every now and then, and that I mainly use it for portable web browsing and note-taking (aka fairly light), but I thought that 8GB was the new minimum for the longest time. Yet, it's carried along with both that and much heavier multitasking and even Blender 3D modeling. Maybe 4GB is really not enough with the much more bloated Windows 11, tho ​:drgn_think:​
       
 (DIR) Post #B6Zam9Y3FhSrbKjoBM by icedquinn@blob.cat
       0 likes, 0 repeats
       
       @radmin apple is always anemic with RAM, especially with their frameworks being on the hungrier side (smalltalk-style stuff with message passing, compositors running all the time etc), but the unices also tended toward better page management and apple standardized on ram compression.you can get linux to do it too but you have to go ask the kernel to, and then there's tuning as to how many pages it runs zstd on and such. i think you can choose lz4 or zstd?although rolz would be better than zstd, not up to me, so meh.
       
 (DIR) Post #B6Zasab1bLvDE6CJiC by mischievoustomato@tsundere.love
       0 likes, 0 repeats
       
       @icedquinn @radmin zswap + zstd + syscl tuning does a lot of good for machine performance fr
       
 (DIR) Post #B6Zavib4JNECuPi5iK by mischievoustomato@tsundere.love
       0 likes, 0 repeats
       
       @radmin 1gb Apple ram ≈ 2gb windows/linux
       
 (DIR) Post #B6ZbIFMVAkeItmdgSe by icedquinn@blob.cat
       0 likes, 0 repeats
       
       @mischievoustomato @radmin i've no idea what apple sets it to. just that a couple years ago they enabled it by default.8gb is legitimately still a lot of memory if you aren't outright wasting it. the only reason 16gb on the travel machine feels bad is AI is ram hungry (which is also just an argument to wasting it) and everything runs in electron (which is also just wasting it)zswap you have to pick how many pages go in to the compressor. which then has interplay with the kernel's "swappiness" setting because zswap means it now is obliged to shuttle a fixed amount of pages instead of as-needed, and last i remember is its basically a fight between 8 pages compresses way better but 1 page is what the swap manager actually wants to do
       
 (DIR) Post #B6ZbXXVTzb7NWA8F4i by mischievoustomato@tsundere.love
       0 likes, 0 repeats
       
       @icedquinn @radmin apple uses wkdm which is 1:2. with zstd+zsmalloc on linux, you can reach 1:3-5 and even higher
       
 (DIR) Post #B6ZbecHqcQ3cN083Gq by icedquinn@blob.cat
       0 likes, 0 repeats
       
       @mischievoustomato @radmin right more pages compresses better, but then it has to actually pack the load so it loses flexibility on paging. afaik
       
 (DIR) Post #B6ZbpS158zw9hbS9R2 by mischievoustomato@tsundere.love
       0 likes, 0 repeats
       
       @icedquinn @radmin I'm not very knowledgeable, the metrics i know are compression speed and latency, zstd is ok speed wise and latency can be handled with vm.page-cluster (syscl knob)there's a page by a kernel dev that eas recently posted... I'll link it in a bit