RISC-V

>>58827231
Unfortunately no, catering technology to tech illiterates was a mistake

Intel won't allow a breakthrough until they put out their new architecture in 2021.

>>58827231
>Well, now that you can use GCC 7 in RISC-V software development that means we will see more and more companies building RISC-V hardware?

I don't think it's that easy. You need to give big incentive to somebody to switch architectures and you can probably only succeed with RISC-V if they already planned a switch. And it will only be embedded/microcontroller design wins, so nothing important for normal users.

However, with the current maturity of the ecosystem, most will likely still go for some other design, even if there are some costs with that.

BTW with RISC-V there will be one big problem - what GPU to use for those chips, assuming you aren't fine with headless devices only.

Graphic IP and in extension drivers are already problem in ARM SoCs.

>>58827231
what is the difference between this and OpenRISC?

>>58829119
talos had a 2d accelerator, we can start with that

>>58829741
>we

>>58829640
I think it's only the licenses. OpenRISC is LGPL while RISC-V is BSD.

>>58829640
>>58831498
i think RISC-V might have had more work done though, i never, EVER hear about openRISC, except when we're comparing it like this

>>58831553
For some companies GPL and derivatives are shitty licenses for software, let alone for hardware.
You have to release all your projects, in which you invested millions of dollars and hours of manpower, for free and with the risk of your rival could just copy it and market it better, thus outselling your product (even when it could be exactly the same). That's why nobody is backing OpenRISC.

>>58829640
Risc-V has a future.

OpenRISC does not.

>>58829119
Someone needs to buy Imagination Technologies and release the PowerVR IPs, or at least release a GPU with more than 5 years of support.

>>58827231
>Well, now that you can use GCC 7 in RISC-V software development that means we will see more and more companies building RISC-V hardware?
Probably not, "lack of compilers" is hardly what was holding people back from spending millions of dollars retooling and developing their own RISC-V chips to cater to a niche market that won't want to pay for it anyway.

>>58831710
So basically RISC-V is just going to be proprietary garbage.

Good to know

>>58832738
there may be a proprietary variant

then it'll encourage people who believe in free software to compete with it

>>58829119
PowerVR is a perfect choice desu

>>58832738
>actually believing it would turn into anything else
RISC jerkers shilling these "open" architectures like RISC-V and ARM for desktop use are hilariously deluded in this way
>>58832762
>then it'll encourage people who believe in free software to compete with it
because that's totally worked out for freedom before

>>58832769
>because that's totally worked out for freedom before
worked for the BSDs, find me a proprietary version of openssh that's actually better

>>58827231
>tip for ip graphic cores
>tip for ip southbridge
>tips for various codecs
>ugh guys whe did my money go? why the fuck does it cost more than intel xeon?

>>58832738
>>58832769
ISA is all the matters you autists

Who cares about the backend being proprietary

>>58827231
I hope either Risc-V or OpenRISC prevails. We need some open architectures...

>>58832779
sure it's worked for individual software components, GNU has pretty much supplanted the old SysV ecosystem

but hardware's a different story, that's a shit ton of investment far out of the reach of the typical small "community" efforts that are staples of free software development, and no buzzword-laden kickstarter is going to produce viable competitive hardware that can take on a big-name OEM

>>58832784
one of RISC-V's biggest selling points is its openness, not its shitty ISA, the RISC vs. CISC wars ended years ago and you should stop living in the past

File: (JPEG Image, 225 × 225 pixels).jpg (7KB, 225x225px) Image search: [Google]

7KB, 225x225px

>>58832835
>one of RISC-V's biggest selling points is its openness, not its shitty ISA, the RISC vs. CISC wars ended years ago and you should stop living in the past
t. completely uninformed

>>58833036
inform us then, what interesting feature does the RISC-V ISA have that gives it a discernible advantage over its more established competition?

>>58827671
Google, Oracle, and a bunch of tech megacorps are backing RISCV

>>58832738
Basically it's an open design that allows proprietary implementations, exactly the way things should be.

>>58827231
You could use llvm long ago, and it's a better compiler than gcc, so no.

>>58829640
OpenRISC was an implementation and a ISA. LGPL meant you had to open source if you wanted to put it in your hardware.

RISC-V is just an ISA. You get the toolchain for free and you can put your "special sauce" in your chips

>>58833672
Megacorps "back" a lot of irrelevant things, doesn't make them worth a shit on their own.

Remember how the Itanium worked out with the weight of practically every high-end vendor worth a shit behind it?

>>58832780
But if it took off instead of for example ARM, the chip would be some cents cheaper because of the absence of that royalty.

That's why companies are likely to adopt it for their embedded controller usage and such roles over time, like Nvidia did.

>>58834161
There are RISC-V core designs too.

>>58833634
You can make your own RISC-V processor without getting sued by some shitty corporation like ARM.

>>58836426
The biggest draw is the fact that the designs are actually useful in this decade.

You can make processors compatible with a proprietary ISA as long as the relevant patents have expired, which takes like 15-20 years.

>>58836520
Then why nobody has made a x86 compatible CPU outside Intel, AMD and VIA in the last 10 years?

>>58837589
They can make a compatible chip but if the manufacturer doesn't have a licensing deal with Intel and/or AMD they cannot use instruction set extensions like SSE1-4, AVX1/2, FMA.

AMD might license out x86-64, they've already done it once before with VIA but I think it requires SSE

>>58834122
>You could use llvm long ago
What version?

File: 1359289932104.jpg (10KB, 205x200px) Image search: [Google]

10KB, 205x200px

>BSD
Into the trash it goes.

>>58831724
>Risc-V has a future.
Sure...

>>58834224
>Itanium
Almost nobody backed Itanium the moment AMD launched their first x86_64 CPUs. Everybody knew Itanium was a mess and they didn't want to deal with that shit.

>>58836520
>with a proprietary ISA as long as the relevant patents have expired, which takes like 15-20 years.
So at the point they're completely useless

GCC literally removed support for all ARM CPUs older than 20 years old (ARMv3 and older) which really hurts things like the amber core (ARMv2 soft core clone)

>>58842166
Usually they're so old that they're no longer relevant by the time they expire but there's at least one ISA that is being revived by the open source community because it still might have some use.

https://lwn.net/Articles/647636/

http://j-core.org/

GCC is under a nonfree licence, the GPL, so I suppose this support is only useful for GNU/Linux users rather than users of free operating systems.

>>58839530
>disregarding something because it's properly licensed

>>58843465
Please report to get your "biggest contrarian of the week" sticker.

>>58843465
Pretty sure LLVM supports it too

>>58832835

RISC-V's selling point is that the ISA is open. That is, you don't have to ask anyone's permission or pay licensing fees to make a RISC-V CPU. Most implementations will be proprietary if it takes off, however.

>>58836426
that was the "openness" I was referring to, that has nothing to do with the ISA itself or what it's actually good for accomplishing

>>58843908
The main advantages are: open design, royalty free and you don't have to share your modifications.

>>58831724

Risc-V is just other meme.

>>58845860
It's not a meme if they make progress.

>>58833036
He is right, though.