[plug] [link] Linus and the fire-hose

[Simon Scott]

> Bdales observations were that the other effect was improved robustness of
> the code because the need to port to other architectures "discovered" all
> of the non-portable programmer tweaks; endian issues, bashing specific
> I/O addresses (DOS anyone ?) etc etc.
> 

The thing I find most amazing is that much of the architecture dependent code
has been replaced (as you say) with generic C, and yet Linux still flies along
at a great rate of knots.

Anyone who has done bit-bashing can tell you that architecture specific
optimizations can be one of the best ways to squeeze performance. I guess
these days some of this is handled by the compiler, but still certain
optimizations would be unknown to the compiler. Im guessing it would be like
the difference between using the blitter and not using the blitter on an Amiga
:)

But somehow, Linux matches and even out-performs other totally architecture
specific OS's. How much x86 specific stuff do you reckon is in windows? Based
on the quality/stability of their Alpha NT distro, Id reckon quite a lot :)

Guessing aside, does anyone have much of an idea about the realities of this?
Is Linux simply 'coded better'? Or perhaps I am wrong and windows is also
platform independent?