On Tue, Mar 21, 2023 at 01:37:46PM +0100, Linux User #330250 wrote: > Lionel Élie Mamane wrote on 03/21/23: >> On Tue, Mar 21, 2023 at 02:43:44AM +0000, Edward Robbins wrote:
>>> this project is making slow but steady progress. Looks like it >>> may be crazy expensive in the end though >>> https://www.powerpc-notebook.org/en/ >> Thanks for the link, interesting and I didn't know about this one >> indeed. (...)I see the one-but least FAQ >> https://www.powerpc-notebook.org/faq/ >> says that it won't run a "modern distro" in little-endian mode, as >> "although it does support LE, modern distros require some >> functionality that are not available to this CPU". >> And Debian's only 64 bit Power port seems to be... little endian? >> Big endian is not even listed on https://www.debian.org/ports/ as >> being in progress, it is not there at all. > With the one exception of the Raptor POWER10 systems maybe, but they > are quite expensive as well. You mean POWER9; apparently POWER10 has freedom problems so they can't upgrade to it, but they hope to "negotiate" the freedom problems away... > Also, especially with Laptops, it's also much about taste... and > style... and "feeling"... I like my Lenovo Legion more, and > traditionally always had something like a ThinkPad, because I like > its style the most (the keyboard, the three mouse buttons -- which > sadly the Legion doesn't have). Ah yes, I had forgotten about that... I'm taking three buttons so much as an obvious given that I didn't think some might not have them :-| That's the strong reason I was always buying one Thinkpad after the other back when I was actually buying computers more than once in a decade, plus I like their Trackpoint, like a lot. And I hate touchpads, like a lot. Oh well, I suppose I can live with the MNT Reform's trackball. > The Raptor however was always way above my margin of what I can > afford. I'm discovering hardware/driver compatibility is ... an issue. https://wiki.raptorcs.com/wiki/POWER9_Hardware_Compatibility_List/PCIe_Devices while "often" the comment is the hardware/driver bugs or "cutting corners" to shave a few cents per device, that's... not helpful to the user in practice. > One thing though, and maybe someone can clarify for me: Why is it > software-wise not possible to emulate an LE system on top of a BE > system? The (Linux) kernel should be able to emulate being LE on BE > hardware, shouldn't it? In the same way that one can emulate anything on anything, yes. However, that won't run a native speed... it is emulation. > (E.g. I know of, but have never used, patched Mac OS X kernels, XNU, > with SSSE3 emulation, i.e. the kernel will provide SSSE3 support > even though the CPU running on doesn't have SSSE3.) I think that works by intercepting the the CPU 'invalid opcode' fault and then running the effect that the instruction would have had through non-SSE3 instructions, and/or dynamically changing the program's code to replace the SSE3 instructions by non-SSE3 instructions. I think the key thing is that "not many" instructions need to be emulated in this way, and the rest runs at native speed. > Would a live BE<->LE translation be so different? I thin so, yes. > I'd rather have a slower but working emulated LE system than a in > theory faster BE system with constant problems, like the one > mentioned in Firefox.[5] The one you link to in Firefox is not linked to little-endian vs big-endian. In my understanding it happens on Power in little endian mode, too.