Behavior of setjmp/longjmp and registers

[d.a.morano]

In article <9467 at ihlpb.ATT.COM>, gregg at ihlpb.ATT.COM (Wonderly) writes:
> It strikes me that setjmp() could save all GP registers.  I know of at
> least one implementation were a jmp_buf is enough space for the entire
> register set.  longjmp() of course just reloads the registers grabs the
> return value passed, and jumps to the return address given to it (on the
> stack or elsewhere).  The expense is a consideration but the results are
> guaranteed in the cases I can think of.

I have implemented setjmp/longjmp assembly routines for both the 
MC680x0 and WE32x00 processors and this is exactly what I did.
The inefficiency of saving all of the registers is worth it from a
safety point of view in order to allow the compiler to perform its best 
register optimizations.  It is not clear to me why there seems to be 
so much interest in the performance of setjmp.  It is only a few 
assembly instructions as it is.  I am not even sure why ANSI made 
setjmp and longjmp macros.  I know that some computer vendors (in 
Europe ??) complained that their architectures did not support these 
as subroutines, but an implementation that saves all of the registers 
in a subroutine seems to me to be both safe and relatively straight 
forward, on any architecture that I am familiar !  I can only assume 
that I do not fully understand the issues involved.

Dave Morano	AT&T Bell Laboratories, MT	mtgzz!dam