Microsoft .NET Runtime Lands Initial Code For RISC-V Support

[phoronix]

Phoronix: Microsoft .NET Runtime Lands Initial Code For RISC-V Support

A Phoronix reader pointed out that there are initial code that landed for adding RISC-V processor support to Microsoft's .NET runtime...

https://www.phoronix.com/news/Microsoft-dotNET-RISC-V

[rmfx]

If Microsoft is smart, they already started to port the OS to risc-v anyway…

[schmidtbag]

If Microsoft is smart, they already started to port the OS to risc-v anyway…

Agreed. Maybe not with the intention of running desktop Windows, but at least for embedded devices.

[zexelon]

If Microsoft is smart, they already started to port the OS to risc-v anyway…

Why? Because they did so well with Zune? RISC-V is a fun toy and very interesting research / community effort but it has some pretty serious challenges to getting accepted as mainstream. At best its a "fringe cpu" and will likely remain so for the next 20 - 30 years if it can hold out that long. I mean the very first ARM CPU was 1983... they didnt really develop into anything big until they JV'ed with Apple in the early 90s... and they didnt start really showing up until they made inroads into cell phones in the what 2010's? They have only in the last 5 - 10 years started showing up in usable laptops.

Microsoft will keep an eye on RISC-V no doubt, and so should a lot of companies... but its way to early to invest any substantial capital on the software dev side.

[ayumu]

Microsoft will keep an eye on RISC-V no doubt, and so should a lot of companies... but its way to early to invest any substantial capital on the software dev side.

In the talks by the foundation itself in RISC-V Summit, there was mention of decisions made to accommodate Windows, like "Windows needs this".

They couldn't be more obvious about knowledge Microsoft is working on a Windows port, without actually saying "we're working with Microsoft on a Windows port".

[PerformanceExpert]

RISC-V is a fun toy and very interesting research / community effort but it has some pretty serious challenges to getting accepted as mainstream. At best its a "fringe cpu" and will likely remain so for the next 20 - 30 years if it can hold out that long.

I'd say 5-10 years - the slow path Arm took from home computers to embedded to mobile to server is no longer relevant. RISC-V can do certain things quicker - for example software is far more portable today than it was decades ago, partly thanks to Arm's investments. Despite that, building up a new software ecosystem from scratch is a slow and painstaking process. RISC-V is only now adding ifunc support to the kernel and GLIBC! As for CPUs, fast implementations are still years away.

[zexelon]

I'd say 5-10 years - the slow path Arm took from home computers to embedded to mobile to server is no longer relevant. RISC-V can do certain things quicker - for example software is far more portable today than it was decades ago, partly thanks to Arm's investments. Despite that, building up a new software ecosystem from scratch is a slow and painstaking process. RISC-V is only now adding ifunc support to the kernel and GLIBC! As for CPUs, fast implementations are still years away.

For sure they could learn a lot from the way Arm did things and skip over some of the errors. It would be awesome if they could pull it off in 5 - 10 yrs!

[stormcrow]

I'd say 5-10 years - the slow path Arm took from home computers to embedded to mobile to server is no longer relevant. RISC-V can do certain things quicker - for example software is far more portable today than it was decades ago, partly thanks to Arm's investments. Despite that, building up a new software ecosystem from scratch is a slow and painstaking process. RISC-V is only now adding ifunc support to the kernel and GLIBC! As for CPUs, fast implementations are still years away.

The devil will be in the implementation details and any ISA extensions used by hardware implementers. Apple bought their way into the ARM ecosystem with PA Semi and it still took them around 10 years from the initial Apple silicon A series iPhones to the M series laptops. RISC-V is starting from scratch and right now all that's available are some not-great prototypes and early development boards of limited capability. Just because software is "more portable" as some have put it, doesn't mean it's easy to have a usable, workable port of an OS and all the software that you need to make it usable on top of that will be there in a decade. It might start being usable around a decade later.

Phoronix readers are often guilty of having tunnel vision. Their use case is somehow the only use case that matters. That's not how the computer industry works. To have a successful, usable implementation that's going to be a success companies like Microsoft have to make their products appeal to the customers with money. Major deep pockets money. That means they'd have to sell a RISC-V port of Windows to the most conservative customers on the planet: corporations and governments. If the value proposition isn't there because of compatibility, performance, and cost, that's not going to fly. It's why Windows on ARM never went anywhere till Apple showed that it was not only possible to have a well performing and energy efficient ARM CPU, but they could also do it and still maintain near full x86-64 compatibility at the hardware layer with as-good-as or better than native performance. There's an old trope in the Windows community that the best hardware for Windows is Apple Macs. That was really not all that true when Macs were Intel. But as far as Windows on ARM is concerned, it's true that the M series can wipe the floor with any of Qualcomm's current CPUs with both of them running Windows on ARM (via Parallels on the Mac). That has value to corporate buyers purchasing for their in house developers who may have to test software on Windows but prefer the Mac experience, or has to support both plus Linux via web apps. One device, much native software testing, plus being able to easily SSH to any given Unix server with native terminal support. It's actually cheaper (and potentially more time efficient) using a Mac with Parallels than having a native box for each OS your developers have to support plus mixes of production hardware for pre-deployment testing.

[Ironmask]

I mostly don't see RISC-V taking off because ARM already exists. The main reason ARM got popular at all is because x86/AMD64 is a monster. Slow, hot, unreasonably bogged down with legacy, held hostage by Intel despite AMD dragging the architecture kicking and screaming into the 21st century. ARM isn't open source, but it's so permissive that, for all intents and purposes, at least for the size of the corporation you'd need to manufacture them, who can handle the cost and legalese and the team to modify the architecture, it may as well be considered open. x86 is everywhere, why would we replace it? Because it's suffocating, and ARM is a breath of fresh air. Maybe not up to everyone's specs, but it's good enough. I'd certainly like to see RISC-V succeed, but it's pretty much redundant, and, like so many other architectures, is probably just going to die on the vine.

One lone Microsoft developer adding support for it in .NET doesn't really mean anything in regards to Microsoft's overall interest in RISC-V or the industry's interest as a whole, it's probably just a hobby/passion thing for him. As for corporations telling Microsoft RISC-V is the future, they've heard it before. NT has supported so many architectures in the past and they've all died out. DEC Alpha, MIPS, PowerPC, Itanium (although they knew that one was DoA and mostly did it for contractual reasons). That won't stop them from supporting it, but I don't doubt they're skeptical at this point.

[OneTimeShot]

I think that RISC-V will become big rather quickly.

Initially it’ll be the embedded market (where C code is just recompiled). Then probably supercomputers where custom ISAs are feasible.

Finally, Google will add Android support as a mandatory AppStore requirement.