RISC Five Thread

Those were both op trying to bump

no, don't you know how the new IP counter works? educate yourself at least a little bit before posting here, reddit

Attached: 1524978182658.png (562x175, 45K)

>More specifically, how much does ARM make from those little chips?
Not much, but you'll likely see similar license costs for risc-v because the people designing the implementation won't be the SoC makers in most cases

We're approaching a point where we can no longer improve hardware by just shrinking it and adding speedhacks. We're going to hit quantum tunneling soon, and Intel's hardware security nightmare should tell you speed hacks in the hardware are becoming too dangerous. The only way to improve hardware is going to be parallelizing it.
At the same time, we've hit the point where it's impossible to improve software by just piling on new layers of abstraction. We should rather look into optimizing by mathematical proofs and smarter compilers, but the hardware we run on isn't very fitting for software speedhacks either.
As it happens, software is currently compiled to be executable on the x86 platform, and the x86 platform it's running on is implemented in software on a much more optimized and capable processor. This setup speaks of our current problems in computing:
Our processors are advanced, but it's all used to implement a primitive processor. Our software is advanced, but it's all compiled into primitive executables.
We must get rid of the general purpose platforms in favor of specialized and optimized hardware, and we need to get rid of primitive languages in favor of advanced languages that can be optimized for the specialized hardware.

I'm reminded of the lisp machines of the past. Their architecture had hardware type information and type checking, eliminating runtime type checking and type errors. They treated memory as a storage for objects, eliminting memory corruption bugs and the ability to interprete memory contents as something it wasn't. They had hardware garbage collection, eliminating the need to implement it in software, and eliminating the risks of memory leaks. And they had special setups for optimizing operations on linked lists.
The true next generation computers will need these sorts of specializations, in addition to ones for modern problems. The hardware and languages used need to be able to parallelize tasks seamlessly, for instance.

>quantum tunneling soon
transistors leak since they exist

Something comparable to the $5 Pi Zero?
Around $999 USD

How well does qemu emulate risc-v? Id like to get started creating my own toy os for it for learning purposes.

>thinking that OP couldn't have changed his IP or posted from his phone
ishygddt

I agree compilers need to be smarter, but that’s about it. I would like to see a declarative language that automatically implements a class of algorithms to their current known asymptotic speeds. That way you could seperate implementation details even further from the programmer. Something like

sort list :: returns sorted where sorted is permutation list and {xi < xi+1 for all xi, xi+1 in list}

Although I think atempts at declarative languages died awhile ago. Or just the semantics of modern programming are inherently wierd so you necessarily need imperative level control.

>the x86 platform it's running on is implemented in software on a much more optimized and capable processor
No its interpreted in hardware. So it just makes decode slower but throughput remains the same. You sound like you need to learn the lesson intel learned 40 years ago. Its the software that matters not the hardware. Who cares if the software is a little slow or unsafe. All that matters is backwards compatibility.

Careful user, this is how TempleOS started

haha, exactly

RISC is a meme.All old computers were RISC. Theyve evolved beyond that.

>Implying the 6502 is not the peak of computing

ARM is RISC, you drooling retard

Think you mean CISC.

>imagine not having any floating point instructions
>imagine not even having multiply or divide in hardware
>let alone anything like hardware video decoding, encryption, etc
RISC is a joke

>freedom is stupid!
typical sour grapes europeon

I wonder where this will lead us.
Intel didn't add all these instructions just for the lulz - they serve a purpose. SIMD instructions and things like hardware accelerated AES actually make certain operations much faster, much faster than you could with mov shit.

So I guess RISC-V's solution to this is... having extensions? That's just wonderful. So it's going to probably end just like x86 where certain extensions are just mandatory for every RISC-V based CPU because the programs just expect them to be there and because not having them would make the CPU run terribly slow (if at all, which would require at least two code paths, thus bloating executable size) or with tons of incompatible, proprietary extensions from everyone.
Oh, I also can't wait for the manufacturer's to be forced by American's freedumz protection forces to include Intel ME-type backdoors into their CPUs unless they want to be mossaded.

>At the same time, we've hit the point where it's impossible to improve software by just piling on new layers of abstraction
This is why languages with smaller runtimes, AOT instead of JIT compilation and zero cost abstractions are now suddenly becoming more and more popular.

hardware video decoding can be added to a computer but really shouldn't be in the cpu in my opinion

>proprietary extensions from everyone
that's exactly what will happen.
Another 10 years more without being able to install linux on your phone.

Not really, which extensions counts as a standard general computing RISC V processor is specified. All RISC-V extentions are free(libre). Simd instructions are being worked on at the moment

>which extensions counts as a standard general computing RISC V processor is specified.
Who had a stroke here, you or I?

>It also runs Quake 2 now!
This is like bragging about being at the top of a special eds class. Wake me when it's capable of running modern software, not 25+ year old games and software that was originally compiled in the 80's..

I prefer MIPS

It also runs Debian. Like, almost everything in it.
youtube.com/watch?v=L8jqGOgCy5M @ 7:25

It went from theoretical to in a physical chip running a classic FPS in like 5 months

Who works on this?

One company, Linus did a tour of their offices

The instruction set and standards? Researchers at Berkeley.
The hardware? A variety of companies. The main one is Sifive, but there are others. Hopefully LowRISC makes more progress and releases their stuff too.
Also, there are many members of the foundation who are interested in the technology. Names such as Google, Western Digital, Microsemi, NVIDIA, Qualcomm, and so on.

link: youtube.com/watch?v=jNnCok1H3-g

Cool, thanks

A limux distro isn't exactly cutting edge tech.
Dev time is exponential you idiot. Don't expect to be running Crysis on it for years and years, and by then AMD/Intel will already have even better processors. They're trying to swim upstream with this, and I cannot see it getting popular outside of hobbyist circles.

>outside of hobbyist circles
You mean like the PC ?

I mean Jow Forums circlejerks. ;^)

define "cutting edge".
>muh steam and adobe

It's such a missed opportunity. Tagged memory and hardware-assisted type/bounds checking should have been part of the specification. Well, enjoy another decade of malware and buffer overflows, I guess.

>and by then AMD/Intel will already have even better processors
>he still believes in moore's law
we've been having diminishing returns for a long while, sweaty

>I cannot see it getting popular outside of hobbyist circles
It really wouldn't take much. The arch itself is sound. If some big company decides to make a RISC-V chip on 7nm with 8 cores then it'll be fine for mobile use and light laptop use. Some proof of concept RISC-V CPU on 48nm with one core is obviously not going to be more than a research / hobbyist toy.

Its RISC. Adding extra crap on top defeats the whole point.

If it can run linux it can run anything. Its just a matter of compiling it.

They could release a Chromebook with a RISC-V CPU.

They still need a low power GPU though

Chromebooks are fukken ded

>number of users increases every month
>"dead"
ok kid

Pure RISC isn’t viable in 2018 anymore, there’s a reason ARM and company became bloated monstrosities over time when they could no longer just out-clock their more complex competitors.

I mean I agree and think RISC is a meme. But if you want to try cool new CPU architectures and extra features you might as well just give up RISC.

>Numbers of users increase
>According to Google
>Who is only one who can access the original data
Good to know.

Cheap Asus chromebook running Coreboot?

Way easier to get a non-blobbed bios than libreboot. Not dead at all.