I've read that 32 bit will be obsolete by 2038. 128 bit os when?

Maybe after we get real 64 bit cpus

>>61159509
this tbqh famalampai

>>61159498
>128 bit os when?
When we need to do 128 bit math in any large volume or need to address more memory than currently possible.

Probably never, we'll have hit the hard limit of what we can do with computers before then. As it turns out, two to the power of 64 is a really large number.

>>61159498
Already using a 128 bit filesystem.

>>61159570
We can already do 128 bit math operations, the 'bitness' of modern cpus is based on addressable memory, not register size
Modern CPUs have 256 bit registers, and some have 512

>>61159570
>Probably never, we'll have hit the hard limit of what we can do with computers before then. As it turns out, two to the power of 64 is a really large number.
How can you tell? Just a few decades ago we thought 640K ought to be enough for anybody

Only the amount of RAM depends on that

>16 bit = 65 KB
>32 bit = 4 GB
>64 bit = 16 PB

Not any time soon.

>>61159633
You are forget two things: Firefox and Chrome.

>>61159633
B- but doesn't the CPU's ALU also work with more bits, making it two times as fast?

>>61160381
Like I said earlier, bit size on CPUs only refers to addressable memory space, not values placed into registers
The only thing a 128 bit CPU can do more than a 64 bit CPU is address 128 bits worth of memory instead of 64

>>61159633

x86-64 and AArch64 have more differences over their 32-bit counterparts than just the amount of memory they can address, and they make up the vast majority of consumer 64-bit CPUs.

>>61159752
>>61159752
Kek'd

Yes ppl will need over 16PB of memory in 2040 so they can have 4000 BLACKED tabs open in firefox for over 48 hours without it crashing

>>61160455
'64bit' solely implies addressable memory, architecture differences weren't brought up in this because we can't predict the future, I don't think it's fair to bring up architectural extensions in this

>>61160504

Considering that the only 64-bit architectures in significant use are both massive extensions over their 32-bit versions, it's perfectly fair to bring up architectural extensions. Name one relevant 64-bit architecture other than these two. Hell, name one relevant architecture at all. POWER is dead outside of some large enterprise servers, and x86 is creeping in there. MIPS is dead outside of embedded, and ARM is fucking cannibalizing it like no tomorrow. SPARC and Alpha have been dead for ages with the exception of some old servers on life support. RISC-V, last time I checked has no standalone physical CPU, just some FPGA shit that has to be bootstrapped by an ARM CPU.

>we can't predict the future
It would be highly improbable for ARM and x86 to lose their throne. Too many legacy applications exist that won't work on other CPUs.

>>61160684
>Too many legacy applications exist that won't work on other CPUs.
Not him but you're forgetting the thing that put those two architectures into the position their in now; they're cheap.

>>61160684
There aren't many difference between base x86 and x86-64 though, registers + calling convention and the 0x48 prefix opcode are the main differences
The xmm/avx/whatever extensions could have technically been done on x86

>>61159615
>Just a few decades ago we thought 640K ought to be enough for anybody
lmao no.

>>61159509
what did he mean with this?

>>61159615
why do people keep saying this quote? Gates wasn't saying 640k would be enough indefinitely, only for that specific machine at that specific time.

>>61160684

Itanium is the future!

>there will be websites that load 16PB of nonfree javascript
>it will happen in our lifetime

It's an abstract type of feel.

>>61160905
Current 64 bit cpus can only address 48 bits, the others are fixed to always be the same as the furthest left bit

>>61159509

Itanium...

Thanks AMD for this shit AMD64.

>>61159498
When we have 128 bit processors

>>61159633
Just imagine, in 10 years your browser will take up 16PB of RAM

>>61160851

Being cheap is how you get your position. Legacy applications are how you keep it.

>>61161596
If only someone would tell women that.

even a low-specs browser like netsurf now only provides amd64 packages for debian.

>>61159498
developer will just stop doing anything for 32-Bit. like they're already started doing (Krita.org). How many Distros already stopped providing x86 Images?

The developers forced KDE4+, GNOME3+, systemd. They allowed blobs in the Linux Kernel. They still didn't give the GPL release of plan9 any love.

This clearly shows where things are going. Don't trust them, learn to develop software yourself.

there is nothing 32-bit anymore.

cant see a valid argument for using 128 bit over 64 bit architecture anytime soon.

2^64 is more than anyone could possibly put to work at the time. Not even google or the NSA

>>61162281
there are valid reasons why one would want more data in less cpu cycles calculated. the main problem are developers that lost the ability to use it right. this is why there is still a distinction between x86 and x86-64 and while multiarch does exist, it's implemention is pretty much retarded.

>>61162481
'128 bit cpus' and '64 bit cpus' don't denote the register size this makes no difference

>>61162546
>numbers don't have meaning
>logic is racist
>everything is equal
why are you people everywhere? just gulag yourself already!

>>61162606
64 bit cpus have 256 bit registers, the point you originally made is invalid because you don't understand how computers work

>>61162786
>something , something, it's your fault
just stop talking. you're obviously retarded.

>>61162937
Average /g/ user, no actual computer knowledge but at least you figured out how to install Ubuntu right?

>>61159615
640k is just a somewhat large number, 2^64 is closer to grains of sand on the earth. 2^128 is more than all the quarks in our whole universe and perhaps like 50 other universes put together. At some point you can't back that address space by memory, there isn't enough anything in our universe for that.

>>61163061
>there isn't enough anything in our universe for that.
ur fuken WOKE

>>61159498
>2038
That's referring to Unix timestamps.

>>61159633
With some messy hacks a 32bit system can support more ram and address 4GB to each program

>>61159498
It already is. Apple is killing it completely next year.

128bit OS doesn't make sense any time soon.

>>61160939
>fug, my Internet is only 7Mbps.. a PB download would take years.

>>61162027
>They still didn't give the GPL release of plan9 any love.
Releasing Plan9 under that disease of a license was a travesty anyway.

>>61162786
as someone with a phd in computer engineering i want to know how you came to this conclusion

>>61164558
what conclusion? that cpus have 256 bit registers???https://en.wikipedia.org/wiki/Advanced_Vector_Extensions#Advanced_Vector_Extensions_2

>>61164783

Those are specialized registers for certain processors m8

>>61165101
most modern CPUs have at AVX2, it came out like 5 years ago

>>61165132

Yeah but those are used for number crunching exclusively, its not like you would use them like you use R/EAX plus the instructions arent standard by any means.

>>61165259
ya ofc, it isn't addressable memory, there's no reason for it to be, we're not even close to the 48 bits we already have addressable
your compiler will emit them for math operations needing values that large, but there otherwise isn't any reason to use them

>>61161015
>The Itanic Disaster
Did you even read the halloween papers?

>>61164783
>>61165132
if by "most modern cpus" you mean "intel and possibly amd x86 cpus", then yes.
these really aren't 256-bit registers at all though, they're more of a control scheme for using multiple registers at once. they support math operations as a group, not on single 256-bit numbers. plus, they can do 512-bit groupings as well (with avx-512), which would theoretically mean you have 2 total registers on modern x86_64.
>>61165331
"math * that large" isn't really the best way of putting it, since your compiler definitely can't decide to use hardware-assisted multiprecision from out under you in any language standards that i know of, and honestly even if you do it manually and your compiler realizes it, it's a coin toss if it's gonna vectorize something even as well-optimized as gmp.
i also feel like it should be noted very heavily that vectorization is *NOT* easy and definitely has to be tiptoed around by compilers, because the simple act of maintaining the data flow of vectors may often forego far more important optimizations that carry weights of hundreds of cycles, not 3-6 (per operation)

>>61165707
>they support math operations as a group, not on single 256-bit numbers
that's kind of disingenuous, that's how normal registers work too, unless you mean that they're not sitting together on the actual die, because i have no idea about shit that low

I think we're using it for different applications, I'm speaking from the crypto side, where use of AVX is common on supported systems

>>61161015
Was this a serious post?

>>61165707
Unless you had meant that it literally just fills 2 smaller registers, and the top 128 is usable at the same time the bottom 128 is

>>61159498
There is no need for a 128 bit os atm.
If the need ever rises, then maybe we'll see one. We already have 128, 256 and 512 bits extensions though.

>>61160684
Just a heads up, there is a arduino-like board with a Risc-V MCU. It's growing slowly, but it's growing.

>>61160992
That's still over 200tb

>>61160992
>>61166515
The AMD64 instruction set was still designed to work with a full 64-bit address space.
They only implemented 48-bits in actual hardware, meaning current software will be forwards compatible with true 64-bit x86 CPUs when they actually make them.

>>61165818
he probably meant that AMD64 is still 64-bit extensions to x86, that there's nothing in the Intel world that's "pure" 64-bit from the ground up as you would see in an ultrasparc, a digital alpha, or a powerpc chip. when you boot from AMD64, you always start in real mode, switch to protected mode, then you pass to long mode.

>>61161918
>be whore
>get preggers
>trap male into marrige
>????
>PROFIT!