• Jason House (1/1) Jan 21 2009 I was just looking at Atomic.d in the D2 port of tango and I see that al...
• Jarrett Billingsley (9/10) Jan 21 2009 the volatile keywords are commented out. My understanding is that the ...
• Sean Kelly (6/7) Jan 21 2009 Walter has said that compilers should not optimize across asm blocks,

Jason House <jason.james.house gmail.com> writes:

I was just looking at Atomic.d in the D2 port of tango and I see that all the
volatile keywords are commented out.  My understanding is that the ordering of
the assembly in those blocks is incredibly important to preserve as-is.  What
is the way to ensure that in D2?

Jarrett Billingsley <jarrett.billingsley gmail.com> writes:

On Wed, Jan 21, 2009 at 9:49 PM, Jason House
<jason.james.house gmail.com> wrote:
 I was just looking at Atomic.d in the D2 port of tango and I see that all=

 the volatile keywords are commented out.  My understanding is that the ord=
ering of the assembly in those blocks is incredibly important to preserve a=
s-is.  What is the way to ensure that in D2?

olol, use inline assembly or link to external functions that do that.

That's what Walter told me, anyway.

http://www.digitalmars.com/d/archives/digitalmars/D/announce/DMD_1.029_and_=
2.013_releases_11949.html#N11974

Sean Kelly <sean invisibleduck.org> writes:

Jason House wrote:
 I was just looking at Atomic.d in the D2 port of tango and I see that all the
volatile keywords are commented out.  My understanding is that the ordering of
the assembly in those blocks is incredibly important to preserve as-is.  What
is the way to ensure that in D2?

Walter has said that compilers should not optimize across asm blocks, 
and so a correct D compiler will not do so, and therefore the volatile 
label should be unnecessary.  But I hope that this will be revisited 
once multiprogramming in D is a bit more fleshed out.


Sean