Re: [PATCH] Go closures for s390[x]

["Ulrich Weigand" <uweigand@de.ibm.com>]

Richard Henderson wrote:
> On 12/19/2014 07:13 AM, Ulrich Weigand wrote:
> > Actually, it sort-of is; the code does:
> >         la      %r14,0(%r13,%r9)                # Set return address
> >         br      %r7                             # ... and call function
> > 
> > i.e. sets the return address register to point to one of the return stubs
> > and then jumps to the target function instead of calling it; so from the
> > point of view of an unwinder, it looks like the target function was called
> > from the instruction immediately preceding the return stub.
> 
> Ah, good point.  That's the sort of verbage that should be in the comment then.
> 
> Is that optimization really worth it?  Is there no call/return prediction stack
> to get confused?  I know I replicated it in the code that I wrote, but really
> only now do I start to question it.

There's no call/return stack as such on current processors; we don't have
hard-coded call/return instructions, and the various OSes on the platform
use registers in quite different ways as part of their calling conventions.

However, I agree that in general it's probably best to avoid tricks like that.


-- 
  Dr. Ulrich Weigand
  GNU/Linux compilers and toolchain
  Ulrich.Weigand@de.ibm.com