The hits keep on coming. I'm not sure whether to laugh or cry. https://nibblestew.blogspot.com/2021/07/looking-at-performance-of-refterm.html …
-
-
Replying to @cmuratori
Most of this is bullshit, the high memory consumption looks interesting though, what's the performance like if you scale the glyph cache down until you hit 'normal' terminal levels?
1 reply 0 retweets 0 likes -
Replying to @t_cheng_1 @cmuratori
Also, why not use cstdlib memset/memcpy? seems like you're just doing extra work to kill performance, SIMD in memcpy makes a huge difference
1 reply 0 retweets 0 likes -
Replying to @t_cheng_1 @cmuratori
As Casey points out in another reply, clang actually does optimize that "naive" memcpy to use SIMD. But anybody could always use the CRT if they want, it's a reference renderer, not a shipping product, after all.
1 reply 0 retweets 2 likes -
Replying to @sandvich_maker @cmuratori
Just a weird choice to do extra work just to hope the compiler optimises back it to the less-work version In my experience, any hand-rolled memcpy is going to shit on an impromptu compiled one, although maybe compilers are good now (?)
1 reply 0 retweets 0 likes
That's not why it's done. It's done because this is a clean compile that has no libraries. That's usually how I do anything made for benchmarking, because I don't want there to be any unforeseen complications from things the CRT might do on startup, etc.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.