[SunRescue] Q on "optimal" OS for Sun4c machines, now that Solaris 8 won't run

[James Lockwood]

On 14 Jul 2000 jwbirdsa at picarefy.picarefy.com wrote:

>    I don't know of any that "figure it out", but chips of switchable
> endianness are nothing new, even for Intel. The 80960 could do so circa
> 1991. My employer at the time was developing emulators for several versions
> of the 80960, and while I wasn't working on them myself, I know that that
> feature gave the developers a lot of hell. IIRC, the endianness could be
> flipped programmatically, so it could for example support per-process
> endianness with an appropriately written kernel. When you're trying to
> disassemble the instruction stream based on snooping memory fetches, the
> last thing you need is for the world to turn upside down.

One of the big pushes for bi-endianality for the Alpha was because it
simplified compatibility with x86 code.  Unfortunately the time taken to
flip endianness was nearly two orders of magnitude higher than the time
taken to do a "regular" context switch, so the idea of per-process
endianness (with some sort of endianness id bit in the process table)
went out of favor in that case.

-James