Pthat's with maximum optimisation
Mwhat the fuck
Ayep
Mthis shit is in firefox?? 
Ay e p
Athe best way to do this is genuinely to allocate an entirely new damn array
Pmarginally better https://godbolt.org/z/87vjddjY3
Lnow get the 2nd char in a C string
Awhich just is major pain
PMajor Pain 
Mdo what all c++ programmers do and roll your own string type
Asimple,
and if you have a japanese character then just
just
give up
perfect!!!
thing[1]and if you have a japanese character then just
just
give up
perfect!!!
Mthere are algorithms for this
Lcorrect
Punrelated but i found that someone made a SIMDified masked v-byte encoding
P(where one bit is a marker bit for the rest of the data)
Min english, four-eyes /ref
PMwhat
Ai'm convinced making simd work well is akin to magic
Mis vbyte a variable-length integer
Pyep
Moh okay
Aooh
Mthat explains a lot
Mlooked it up and just got bitcoin stuff
Mwhat is it.. for
Moh, storing packed arrays of ints and operating on them
Lin zig the compiler just gives you some nice vector types and functions and does all the hard work for you :^)
Lyou can do
@Vector(8, u32)
Trust has the same, and it's hit or miss
Lyeah im sure for optimal perf its not perfect, but in general it gives good results and is a million times easier
Lyou can always inline asm the raw instructions
Tban the asm
Li think the stdlib has some of them already done for you for x86 and arm
Ayeah
Athats cool
Lits basically free SIMD for a lot of stuff, since
@Vectorc-colour simd types???
Awhat are you doing with colours that requires simd
Lmultiplying
Tmeh, they can't put everything behind a do-it-all magic helper
Lalpha
Laddition
AOHFUCKIREGRETALTTABBINGOUTOFOSUNOWHELP
Lyeah not everything is behind the magic helper, but thats what
std.simd@Vectorinline asm more like death, straight up
Mimage operations
Lits not that bad in Zig