On Wed, 27 Jan 2016, Will Deacon wrote:
> > Overall I think it should be safe after all to use SYNC_RELEASE and other
> > modern lightweight barriers uncondtionally under the assumption that
> > architecture was meant to remain backward compatible. Even though it
> > might be possible someone would implement unusual semantics for the then
> > undefined `stype' values, I highly doubt it as it would be extra effort
> > and hardware logic space for no gain. We could try and reach architecture
> > overseers to double-check whether the `stype' encodings, somewhat
> > irregularly distributed, were indeed defined in a manner so as not to
> > clash with values implementers chose to use before rev. 2.61 of the
> > architecture specification.
>
> Do you know whether a SYNC 18 (RELEASE) followed in program order by a
> SYNC 17 (ACQUIRE) creates a full barrier (i.e. something like SYNC 16)?
By my reading of architecture specifications it does. Specifically
SYNC_RELEASE (18) applies to older loads and stores, and newer stores, and
SYNC_ACQUIRE (17) applies to older loads, and newer loads and stores. So
the two combined ought to be the equivalent to SYNC_MB (16), which applies
to both older and newer loads and stores. Of course care has to be taken
about what happens between SYNC_RELEASE and SYNC_ACQUIRE. This is still
more lightweight than classic SYNC (0). See the architecture documents,
e.g. the MIPS32 one[1] for details.
References:
[1] "MIPS Architecture For Programmers, Volume II-A: The MIPS32
Instruction Set", MIPS Technologies, Inc., Document Number: MD00086,
Revision 5.04, December 11, 2013, Table 4.7 "Encodings of the
Bits[10:6] of the SYNC instruction; the SType Field", p. 305
HTH,
Maciej
