not an ms engineer, which means it is just sad

> .NET Runtime community contributor & triager.

somehow the part i'm annoyed about is lkml replacing the art with some giant generic icons

on this topic, risc-v vectors can do mmio...
loading and storing to vector regs are not accessing "big words" in memory, they are a bunch of normal-word-sized accesses by definition, just with less ordering, and optimizable under the as-if rule on normal memory

this is supposed to be a post about "volatile doesn't let you pick instructions so why use it for mmio, just use inline asm"
i'm now convinced that yeah if you set up the memory attributes correctly it will send the correct bus request... even on say sse as long as it's a correct memory reference

i have now been informed that in addition to MMIO, volatile globals can be used to communicate with signal handlers (and presumably by extension, interrupt handlers) and volatile locals are preserved across upon longjmp-ing

i haven take a look at the docs for core::ptr::write_volatile. it told me to look at the C standard. i looked.

I tried to gcc -O3 my assembly program today. I don't know what I expected.

i can confirm this happens automatically with any shoes that look even a bit height-enhancing

reskeet with the trans-iest thing you did pre transition

make friends with all the trans people i see so i can make sure i'm not trans

at this point, i don't know how much of the pain in my life is self-inflicted... let alone figure out which parts are necessary

i'm fortunate enough to have like a 'my computer' because my birth father gave me a corp-issued laptop that depreciated to zero value. i put ubuntu on it ~start of high school because well it worked well enough for most stuff needed anyway, office suite even the flash games

the main reason i use glibc/linux as the os on my personal machines is just i'm used to it, like that's the root reason and other bits like 'i can go check the source code if something breaks' or 'the command line works for me' is secondary to that

day ? of waiting for Linux io_uring to add FUTEX_WAKE_OP for... 'totally normal' performance improvement use cases