CAS?

[Johnny Billquist <bqt%softjar.se@localhost>]

I was looking a bit at the VAX code now, since performance seems to have 
dropped seriously lately, and I noticed that mutex, lock, and possibly 
other operations to handle locking out of threads, interrupts, and 
multiprocessors are implemented via a software implementation of CAS on 
the VAX.
This can hardly be efficient, and I wonder if there is any specific 
reason for this, or if it might be worth some work to rewrite these to 
as far as possible use VAX primitive instructions for the needs instead.

Anyone have any thoughts or ideas? Opinions? Maybe someone who was 
involved in the current implementation can share some thoughts on why it 
was done the current way? Maybe I'm missing something?

        Johnny