Asynchronous I/O

["Cheng3" <Cheng3@email.msn.com>]

	Hello everyone, I have a question regarding hurd running on systems that
has a Pentium Pro or higher chip.  In the Intel PII manual, it says that
processors higher or equal to Pentium Pro has two pipelined buses, which
feeds off the Level I cache.  This allows for some parallelism on a single
processor.  While this improves the performances of most software, but it
does not guarantee a synchronous flow of program.  The document also warns
OS programmers to disable caching or put in instructions that will force the
processors to execute program in synchronous order (such as cpuid instr.),
while making I/O calls on ports.  I wonder if this is implemented on hurd or
Mach kernel.  If not, will it cause some unexpected problems, such as
erroneous I/O.