Hacker News Reader

Comment by throwaway167 | Hacker News Reader

We've come a long way when Google are peddling non-free software run on freedom hardware.

Commercially available RISC-V hardware is not more "free" than other types of CPUs, in any meaningful way.

The openness of the RISC-V specification is great for chip designers, manufacturers, and researchers. But it has basically no effect on end users, except indirectly (i.e. hopefully the chips are cheaper, because there's more competition and no license fees).

You could also say that open source has basically no (direct) effect on end users, because the majority of end users are not coders, and even those that are rarely look at the code of the apps they use.

The indirect effects are the point.

The indirect effect of open source software are really strong though. It only takes one coder to remove an anti-feature that the devs decided to include and publish.

The indirect effects of proprietary extensions, which means your software isn't portable anymore?

What are the indirect effects? You realize the least troubling part of making Android run on some proprietary hardware was the ARM processor? That part is perfectly fine!

It's the graphics card, the display hardware, the camera, the video encoding/decoding - these are all proprietary, undocumented blobs of IP with no upstream support. Your RISCV Android phone is still going to have all of those, it makes exactly zero difference.

Wouldn't the main indirect effect be that it's easier for new companies to enter the smartphone space due to reduced costs from the ISA being open? I'd expect that to take a while, but wouldn't that be the main benefit?

The cost of developing a competitive CPU is high, and then you have to package the whole thing into a SoC and optimize for power usage, support it for years, etc. The ISA is not the limiting factor there.

>The cost of developing a competitive CPU is high

It can be dodged by either licensing some of the hundreds of commercial core designs available, or by using one of the (several) available open source RISC-V implementations.

Other competitive ISAs also have a few -albeit not anywhere as many- open source implementations, but there's no dodging the ISA royalty fee on these.

How many of those open source implementations have competitive performance?

As another commenter already pointed out, open source tends to build on open source, so I'd expect more and more complex chips to be built on the platform over time in the open. Maybe there's a future world where a viable open source smartphone can exist even.

Actual designs of chips you’ll be able to buy will not be open source, with a couple of irrelevant exceptions by a foundation or an educational institution or something. It won’t do you any good, though: the feedback loop to get the open source gradual but exponential improvement engine working is too long and too expensive for hobbyists.

It possibly could work with a community of companies, but my imagination stops there, can’t see how mbas could ever agree to giving away differentiators when fixed costs are so high, it’d destroy margins.

For the majority of open source projects the indirect effects are the potential for more eyes, more bugfixes, and more knowledge to spread.

RISC-V CPUs don't have any of that as there's no requirement that actual implementation be open source, only the ISA. And it's not like you can change it. You can't buy a RISC-V CPU and fiddle with the design, you don't get the source nor the toolchain. All you get are the man pages, and guess what? You get those with x86 and ARM, too!

> RISC-V CPUs don't have any of that as there's no requirement that actual implementation be open source, only the ISA.

There's a growing list of RISC-V cores that are themselves open source (that's how RISC-V ecosystem got started, btw). And unlike in x86 or ARM world, no-one would hassle you for licensing or royalties if you bring those designs to a foundry & mass-produce.

Now as for say, GPUs to stick onto those, that is another matter.

Not into open hardware, but recently bumped into this FPGA-based approach - the MNT RKX7 Open Hardware FPGA CPU Module [0] - that is a good example of how to experiment with a RISC-V design before spending the money to produce the actual chip.

[0] https://mntre.com/media/reform_md/2022-09-29-rkx7-showcase.h...

The blog post is by the open source office in Google, not specifically the android team

And so far entirely about AOSP and getting that running.

GApps (the non-free part) isn't mentioned at all. And like someone else said, there's use cases besides phones that use android. And some people Run de-googled android on phones, it's just degraded :/

- googler, but far from anything android

Freedom hardware?

Assuming US-market-first hardware (As opposed to EU or China). This is my first time seeing such a term too

More likely, it allures to RISC-V being a royalty-free open standard, unlike ARM. As teraflop said, though, in practice this doesn't mean much. Chip designers will throw in proprietary IP just for patentability and differentiation.

RISC-V isn't freedom hardware, there's no equivalent to copyleft.

Uncopylefted software can still be free in every sense.

Free to proprietarize it and free to lock it down and free to spy on users. So much freedom!

They are peddling closed software that spies on you on closed hardware.

Reply on news.ycombinator.com