>2017 >still no 128bit computers

>>58882251
Let's start with what is the bit-ness of a processor.

If nothing else, the bit-ness of a processor refers to its ALU width. This is the width of the binary number that it can handle natively in a single operation/instruction, and/or fit within a general purpose register. In this regard, a _32-bit_ processor can operate directly on values which are 32 bits wide, in a single instruction. Your _128-bit_ processor would therefore have a large ALU capable of performing addition/subtraction/logical ops/etc. on 128 bit numbers in single instructions. Do we need that? What problem makes worth having a large and expensive ALU, and paying for? Do we have the need for executing loops with more than 2^64 iterations?


We have to pay for the wide ALU whether we use it, or not, or only a small fraction of its capabilities. To justify a 128 bit ALU, we have to find a problem, significant enough, that can benefit from manipulating 128 bit words in single instructions. We can probably come up a few examples, however these are already served with special purpose instruction sets (like SSE) or highly targeted processors for that purpose (e.g. DSP), and there aren't enough of such problems that make manufacturers feel they will get a return on the investment required to produce such a chip. For instance, we wouldn't put one of those chips in a fridge, a car, etc. In terms of cost, consider the wide ALU doesn't only make it more expensive, but other parts of the chip as well. A 128 bit wide ALU also means there have to be 128 bit wide data paths: that takes a lot of silicon area. Also, that data has to come from somewhere and go somewhere, so there would need to be similarly large registers, cache, other memory, etc, for the wide ALU to be used effectively.

>>58882270
My final point is that we can already perform arbitrary width arithmetic on any processor. We can add 32 bit memory words on a PIC 18 in 8 instructions, whereas you could do it on the same architecture scaled to 32 bits in only 2 instructions. We can perform arithmetic on arbitrarily large numbers using software libraries which can be compiled on current processors. A narrow ALU doesn't keep us from performing big computations, it only limits us in certain aspects. Your question is therefore a question of speed (mostly), and not capability. If you look at the spectrum of applications that need to use particular width numbers, I trust you will see very few that require 128 bit words. The expense of accelerating those few applications in hardware, that won't help the other set of applications, just isn't a good investment at this time.

>>58882251
> 2017
> Still no quantum computing

>>58882251
If you're talking addressing, we aren't exhausting 64bit addresses anytime soon, in fact, Chrome's abuses that assumption, addressing hundreds of gigabytes of memory to sandbox NaCl applications. It doesn't use them (that would trigger OOM instantly) but it reserves them to prevent any sort of data and/or code from getting too close to the NaCl binary (which, if malware, jump to it and potentially cause mayhem).

If you're talking registers, any x86 with at least SSE (Pentium III, 1999) has several 128bit registers if not 256bit registers for SIMD operations. They can't be used for doing 128bit math (instead parallel 64/32/16/8 bit operations) but if you need 128bit operations you can probably deal with the overhead of doing twice as many 64bit ones.

>>58882270
>>58882281
Nice pasta

>>58882420
>nice pasta
Would you rather have me post a link to reddit cancer? Clearly op is faggot and didn't google his question so I did and pasted the info I found.

>>58883138
those were appreciation words, you nigger

>>58882251
For $100k I'll build you a 128 bit computer. Can't promise it being particularly fast or efficient for anything or to solve any particular problem except lack of 128 bit computers.

>>58882251
When will 32 bit be deprecated?

>>58883956
When every single task you need to do needs 64 bit to be efficient enough, i.e. never ever.

>>58882308
This

File: riscv-blog-logo[3].png (10KB, 366x289px) Image search: [Google]

10KB, 366x289px

>A 128bit CPU doesn't exi-

>>58884237
I don't think there's any 128-bit processors under development though.

>>58882251
>128-bit float
>i still can't check if something is equal with error less than 10^-26

decimal type floats would make more sense

>>58885110

The instruction set exists.
The need for CPUs with did instruction set does not exist.

>>58888553

With said instruction set*

Autocorrect sucks.

File: moon.jpg (141KB, 1920x1080px) Image search: [Google]

141KB, 1920x1080px

>>58882251
>2017
>still no trinary computers

>>58889231
You can try that:
http://trinary.ru/projects/setunws/