The Most Affordable & Open-Source POWER9 System To Date Can Now Be Pre-Ordered

[mulenmar]

Perhaps Firefox is slower? (much slower at running javascript, fast at everything else). The one thing I can imagine is there's no JIT compiler for PPC64le in Firefox (is there one?). If you're concerned by firmware philosophy there's a good chance you might be running NoScript anyway. Or trying LibreJS would fit in there.

Some quick research shows that there were some issues recently, both with simply starting the browser (https://bugzilla.mozilla.org/show_bug.cgi?id=1462566) and some webRTC tests failing (ppc and ppcle, https://bugzilla.mozilla.org/show_bug.cgi?id=1408504), but they are already fixed. I didn't find anything quickly wrt the JIT, but frankly I'd be surprised if Firefox even still includes a non-JIT implementation anymore, what with their rewrite-in-Rust. Regardless, web developers unfortunately think the entire world is x86 and little-endian ARM and use Emscripten to generate "fast" Javascript asm.js blobs that only run on little-endian systems, so even if the browser itself works, that doesn't mean websites will work in big-endian mode.

POWER9 supports both modes. I don't know when IBM put that support in, or if there's a performance loss; iirc, some older versions had multimedia extension instructions that only worked in big-endian, even though the rest of the instruction set worked in bi-endian mode. I doubt many mainline-Firefox developers have such machines to work on or to test in everyday-usage anyway, so the only way anything's going to be checked and fixed is for somebody to get the hardware and try.


The developer of TenFourFox, a Firefox build for Mac OSX 10.4 PPC, apparently has a Talos II and has been messing around with it. http://tenfourfox.blogspot.com/2018/...mputer-or.html, so perhaps something will end up being done in that department.

[drSeehas]

... a commodity ATX board, ...

Are you sure?
Isn't it a SSI EEB/Extended ATX (EATX) board?