Re: [Xen-devel] [PATCH v8 02/17] Add cmpxchg16b support for x86-64

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 14.10.15 at 12:03, <feng.wu@xxxxxxxxx> wrote:

> 
>> -----Original Message-----
>> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
>> Sent: Wednesday, October 14, 2015 5:46 PM
>> To: Wu, Feng <feng.wu@xxxxxxxxx>
>> Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>; xen-devel@xxxxxxxxxxxxx;
>> Keir Fraser <keir@xxxxxxx>
>> Subject: RE: [PATCH v8 02/17] Add cmpxchg16b support for x86-64
>> 
>> >>> On 14.10.15 at 11:29, <feng.wu@xxxxxxxxx> wrote:
>> > I understand your comments above. My understanding about the gcc
>> > online doc is we don't need to add memory clobber when the variable
>> > is either an input _or_ output one. However, seems your options is
>> > memory clobber can be removed only when it is an input _and_ output
>> > one. It would be highly appreciated if you can correct me if I understand
>> > the doc incorrectly.
>> 
>> Yes, I'm pretty convinced you read it wrong: "performs memory
>> reads or writes to items other than those listed in the input and
>> output operands" means _any_ operation not expressed by the
>> ams operands. The two sentence following that actually explain
>> this in more detail. IOW - unless your operands correctly express
>> _everything_ the assembly instruction(s) do(es), you need a
>> clobber (and this is not limited to memory).
> 
> Thanks for the clarification. So do you think it is better to send only
> this single patch after fixing the issue in your comments, or wait
> for the next version of this whole patch set?

I don't know - I didn't get around to look at the rest of the series yet.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel