Raptor Computing Systems Is Working On Bringing Up Chrome's POWER Support

[Mr.Radar]

The developer of TenFourFox, the Firefox port to PowerPC OS X, is currently working on a ppc64le JIT backend for Firefox "inspired by" the (32-bit big-endian) PowerPC JIT in TenFourFox.

[edwaleni]

https://wiki.mozilla.org/PowerPC

1 current maintainer. 4 people total.

Same issue, need help with getting the JIT cleaned up.

Before anyone jumps to conclusions, let me clarify that we understand that PowerPC is a tier-3 platform and there is no plan from this team to change that any time soon. If that ever happens, it will happen naturally as the result of our success maintaining the platform.

We want to spread the word to developers who want their code running properly on the PowerPC architecture. If you need help testing, debugging, verifying patches, or need access to PowerPC hardware, just let us know on #ppc64 or email Gustavo directly and we can arrange whatever you need.

When I checked the Chromium PPC site, it was built and working on Talos hardware, but had issues with VP9 support due to lack of of VMX/VSX (Altivec) support.

[tildearrow]

Typo:

POWER build-bot coverage for Chrone,

[carewolf]

Yeah. One of the platforms lost with QtWebKit

[shawnanastasio]

I'm also curious about that, given that it would seem more fitting to have the free-to-the-firmware machines running the browser developed by a public-interest non-profit, rather than a publicly-traded ad-tech corporation.

I'm one of the community members that worked on the chromium port. We initially targeted chromium because its Javascript JIT engine v8 already has mature ppc64 support. As others have stated, Mozilla doesn't seem interested in supporting ppc64 as a tier 1 architecture for any part of their browser.

[jabl]

As for technical stuff, since Rust uses LLVM as a back end, I am assuming this shouldn't be too hard to get ported to POWER? If it is, then this is yet another reason why keeping with C as a language has many benefits to portability.

LLVM already supports the ppc64le target, thus providing support for Rust, Julia, clang, and other languages leveraging LLVM. So this isn't the problem here. Presumably the issue here is support for the Javascript JIT compiler (an entirely custom affair not leveraging LLVM, GCC or any other compiler stack with existing ppc64le support), which is necessary for competitive performance on the modern web.

[freespirit]

I'm one of the community members that worked on the chromium port. We initially targeted chromium because its Javascript JIT engine v8 already has mature ppc64 support. As others have stated, Mozilla doesn't seem interested in supporting ppc64 as a tier 1 architecture for any part of their browser.

Is not interested because how many ppc there is out there? not too much, more software avalaible mean more selling than mean more numbers, and they could reconsider their tier, they have arm on tier1, arm is relatively a new platform, but is on tier1 just because of popularity
like qarium no firefox no party and there are a lot of people that need it, if i wanna use google i will not spend thousand of dollars for a pc when with the same money i can buy a pc an android phone and even a chromebook, and have all the software i need, if i buy something like a ppc to have a blob free system i wanna get free to google too

[King InuYasha]

LLVM already supports the ppc64le target, thus providing support for Rust, Julia, clang, and other languages leveraging LLVM. So this isn't the problem here. Presumably the issue here is support for the Javascript JIT compiler (an entirely custom affair not leveraging LLVM, GCC or any other compiler stack with existing ppc64le support), which is necessary for competitive performance on the modern web.

The front-end (clang, rustc, etc.) needs to also support the architecture, because they need to map the language concepts to primitives for the architecture.

[shawnanastasio]

Is not interested because how many ppc there is out there? not too much, more software avalaible mean more selling than mean more numbers, and they could reconsider their tier, they have arm on tier1, arm is relatively a new platform, but is on tier1 just because of popularity
like qarium no firefox no party and there are a lot of people that need it, if i wanna use google i will not spend thousand of dollars for a pc when with the same money i can buy a pc an android phone and even a chromebook, and have all the software i need, if i buy something like a ppc to have a blob free system i wanna get free to google too

Sure, Mozilla has their reasons for not supporting ppc64 at the same level as other architectures, but it's a catch-22. Mozilla won't provide support unless they see more users and more users won't switch unless they have a working browser (among other things). Porting a popular browser like chromium is one way to break the cycle, because it will allow more users to switch to the platform without sacrificing their browsing experience.

In the meantime, there are also projects like ungoogled-chromium which allow privacy conscious users to disintegrate the browser from google services.

[carewolf]

I'm one of the community members that worked on the chromium port. We initially targeted chromium because its Javascript JIT engine v8 already has mature ppc64 support. As others have stated, Mozilla doesn't seem interested in supporting ppc64 as a tier 1 architecture for any part of their browser.

I don't think you can clasify anything in Chromium as tier 1 unless it is in a product Google ships, everything else is always officially unmaintained and usually in permanent broken bitrotten state due to the frequent pointless refactorings they do which breaks everything not updated, and they only update what they build themselves.

But yes, not having to write a new JIT backend from stratch is a win.