Little-endian YDL?

[Chris Gehlker]

On Nov 20, 2003, at 7:32 PM, Robert Sanders wrote:

> Chris,
>   I knew the G5 was having trouble with Virtual PC regarding what I 
> remember being
> called 'pseudo little-endian' mode.  I just wasn't sure exactly what 
> was meant here.  Was
> virtual PC actual shifting modes on the fly for the emulation?  I 
> could understand removing
> a dynamic on-the-fly endian shift, but I thought one of the core 
> points of the POWER
> architecture (including PPC) was bi-endianess.

It's been awhile since I've actually messed with mode shifting in the 
PPC architecture. The hottest chip back then was the 604. But the spec 
does call for PPC to be bi-endian depending on a bit in the MSR. And 
VPC does set that bit.

I don't for sure know how the term 'pseudo little-endian mode' came to 
be  applied to what the PPC spec calls simply 'little-endian mode'  or 
what is 'pseudo' about it. I suspect that some Microsoft engineer who 
was familiar with the 860i and/or 960i family of chips just applied it 
to PPC without really understanding it. The 860i does indeed have a 
pseudo little-endian mode.

Is the G5 'not a PPC' because it only operates big-endian? Too some 
extent, I'd have to say yes. There is nothing in the PPC specification 
that says the endian bit is optional.