RISC-V Implementation


Not much more to say really, a RISC-V implementation of Haiku would be a great way to get a head start on that platform.


The majority systems are either x86 or ARM. I read a lot of tech news and I haven’t heard anything about RISC-V. Could you explain why you think this platform is important to Haiku?


My reasoning is that RISC-V is supposed to be an open instruction set architecture that is intended to be replicable in hardware without some of the the associated costs of x86 or ARM.

I know we’ve seen this kind of thing before but supposedly there is interest from manufacturers. If there were ever enough interest to produce desktop class CPUs based on RISC-V, Haiku could have passed Beta 1 by then and gained more interest. If the two coincide, it might be possible to build Haiku workstations using RISC-V CPUs as a sort of “PC Lite” or modern-day BeBox.

There’s not necessarily anything wrong with x86 or ARM, but RISC-V claims to be “open” and presumably free of microcode blobs.


RIRSC -V suppose to be an intel power-like proccessor, and is starting to call attention, but is not so popular like a raspberry pi with arm right now.


Is there a platform? Because we can’t do much if there is just hte CPU instruction set and a compiler. What would we use as a bootloader? What is the physical memory map? How do we get a framebuffer?

We are already strugling with getting x86_64 and x86 builds up to date and working properly. The ARM and PPC ports have been “work in progress” for years and never booted to the desktop. I’m not sure adding yet another platform is the best way to use our developers resources.

That being said… patches welcome, still!


Sure, I mean - I know how this works, I’m not asking anyone to do it :slight_smile:

Just making people aware of RISC-V that might not already be. As you say Adrien, there probably isn’t any hardware out there yet…


“V” means vaporware. I understand, the question was asked just to jabber jabber about something, but seriously, why not yet a stable and fully usable OS should bother itself with a fantom CPU architecture? to pull out farther from the reality?)
this “openness” is just a modern snake oil marketing trick.
there are a lot of mini-desktop grade arm boards of any kind. they are all slow and weak compared to x86, but they are real and still have its niches of use.
if someone wants so badly a different arcitecture, here you go - arm. Or even mips. but this risc-v is not yet really present.



May 23rd; RISC-V Linux Port Pursuing Mainlining in the Kernel

Again, I know that “chasing Linux” isn’t a goal, but I’m just throwing a bit more wood onto the fire.


I’ve run Linux in the simulator they supply with the lowrisc project. Not a big thrill, but the platform has promise. It’s way too early for the devs to be wondering about memory maps and frame buffers, but it’s definitely a platform to keep in mind as time progresses.

Right now you can run a RiscV machine in FPGA on a Parallella board, and there is an Arduino compatible RiscV MCU board (done in hard silicon) available for sale. That board is the SiFive one mentioned by someone already. They say they’re working on bigger scale stuff though. They have a dev board available with a bigger FPGA, that runs a RiscV machine and Linux OS variant. We’ll be watching them to see how it goes. The sites for info on the instruction set are at riscvfoundation.org and riscv.org.

Been thinking about getting one of the little MCU boards just to plunk around with it.


With actual RISC-V hardware produced and companies as large as Nvidia and Google becoming members of RISC-V it’s probably time to revisit this topic.

Personally, I think this is a great idea and if it manages to overtake x86 it may be critical to Haiku’s survival in the future.


Does RISC-V’s more open architecture make it easier (at least theoretically) to port Haiku over to it? Are there certain aspects of it that would make it easier or harder to get Haiku running on it in the future? No matter what the answers are, I think that it should be explored as a task for the future.


RISC-V isn’t mature enough to start a port yet…

If they start making PC hardware powered by it…then it will merit a look. As it stands most of the hardware is just very fast microcontrollers and embedded boards which change too offten (same problem as with arm).

We’ve brought it up on IRC power9 is a viable plaform, and the Blackhawk board to be launched soon would make a really powerful haiku system 4-8 Power9 cores under 1k usd. Sure the CPU is not open like RISC-V but every bit of the firmware is open and documented.


RISC-V is horsesh*t. Good in theory but will never take off. Basically, it’s an expensive toy.


That’s uncalled for. RISC-V has done what other similar architectures have not… it actually has ASIC implementations in the wild and you can buy it runs about 1.5Ghz etc… and is 64bit.

https://www.crowdsupply.com/sifive/hifive-unleashed $1k for a dev board like that is relatively cheap honestly especially considering how much of it is open source.

Unfortunately no PCIE… if it had that, it might have been worth a look if they had implemented it in a manner that makes it a supportable platform instead of just a single dev board eg future boards would be compatible with it… without PCIE it can’t run any addon cards to actually make it useful.

Progress on hardware like this is usually pretty slow, the Power8 and 9 TALOS workstations for instance have taken years to get where they are now.


Exactly. $1k for a Raspberry Pi-level board with little to no software support, made by some hobbyists. Do you know something I don’t? Are prices about to drop? I like the idea, and I’d buy one if I could afford, but… that’s never going to happen.


Except the RaspPI actually is a a closed source piece of crap… with an even worse GPU. Whereas this is fully documented, and probably that is why they don’t have PCIE yet… no open source implementation yet.

The point is the board isn’t meant for end users … its meant for developers to start work on the Linux port and other similar uses, perhaps someone building a product around thier SoC like a router etc… .Once that is done, and they have PCIE, they’ll probably mass produce a cheaper board for people to buy.

A 35$ Raspi is a toy with no real potential to branch into a separate product… this is a development tool to prime the software for upcoming boards.


But why develop for this when everything else already uses ARM or x86? It seems like too risky an investment. Sure, let’s pour cash and development time into a shiny, idealist architecture that may never gain any traction, instead of continuing to work towards the realistic goal of a widely-usable, easily-supported product compatible with what people actually can (and will be able to) buy.

RISC-V is a good idea, but it’s too little, too late.


Ever the pessimist huh?

ARM has licensing fees, you can’t even license X86… this should be obvious to anyone pretending to know anything about CPUs.

It already has traction anyway… NVidia uses it in every GPU they make as the embedded controller. I wouldn’t be supprised if it even replaced ARM TrustZone at some point in AMD processors… there is already an implementation of that with Hex Five’s MultiZone Security.

Who knows if it will ever be a desktop class CPU… but frankly it’s already the most successful new CPU architecture bar none. It certainly has the potential to scale to desktop useage… without the drawbacks that past architectures like OpenSparc have (Sparc is wonky… due to it’s crufty windowed register file).


So then I guess you did know something I didn’t… I still feel like it’s going to be a niche CPU, never really much more than a fast microcontroller, mainly due to the fact that they get way too expensive once you scale beyond that. Let me know when there’s a Linux-capable model I can actually afford.

And why would AMD replace TrustZone with RISC-V? They profit off our private information, that’s why the chips are so cheap - the customer is the product (if you believe the conspiracy theories - I don’t, but I’m still wary of any closed-source firmware of that nature. If TZ really were leaking our data, you’d see it on Wireshark). Once they replace it with open-source hardware, they can’t do that anymore, so prices go up.