On Tuesday, 12/16/2003 at 02:14 CET, Phil Payne <[EMAIL PROTECTED]>
wrote:
> > I imagined that MVC was "atomic" in nature meaning that the second
processor
> > couldn't "garble" things until the MVC had run it's course.
>
> I think that's "implementation-dependent", as IBM would say.  Unwise to
rely on  it.

No, not implementation-dependent.  Look in the PoP under "Consistency
Specification".

Subject to the documented initial conditions, MVC and MVCL will have
consitency on each 4 or 8 bytes that are read and written.  That means
that the storage access to those locations is serialized and other CPUs
will observe storage changing in 4- or 8-byte increments.

As another example,  (assuming 390 architecture mode) if you load all 16
registers with Load Multiple, and the operand address is on a word
boundary, then registers will be loaded 2 at a time (doubleword) with
consistency.  Regs 0 & 1 will be consistent, 2 & 3, and so on.

Alan Altmark
Sr. Software Engineer
IBM z/VM Development

Reply via email to