Messages in this thread | | | Date | Tue, 10 Dec 2013 09:37:38 -0800 | From | "Paul E. McKenney" <> | Subject | Re: [PATCH tip/core/locking 4/4] Documentation/memory-barriers.txt: Document ACCESS_ONCE() |
| |
On Tue, Dec 10, 2013 at 04:10:50PM +0100, Ingo Molnar wrote: > > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote: > > > On Thu, Dec 05, 2013 at 10:05:47AM -0800, Paul E. McKenney wrote: > > > On Thu, Dec 05, 2013 at 10:50:42AM +0100, Ingo Molnar wrote: > > > > > > > > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote: > > > > > > > > > + (*) The compiler is within its rights to reload a variable, for example, > > > > > + in cases where high register pressure prevents the compiler from > > > > > + keeping all data of interest in registers. The compiler might > > > > > + therefore optimize the variable tmp out of our previous example: > > > > > + > > > > > + while (tmp = a) > > > > > + do_something_with(tmp); > > > > > + > > > > > + This could result in the following code, which is perfectly safe in > > > > > + single-threaded code, but can be fatal in concurrent code: > > > > > + > > > > > + while (a) > > > > > + do_something_with(a); > > > > > + > > > > > + For example, the optimized version of this code could result in > > > > > + passing a zero to do_something_with() in the case where the variable > > > > > + a was modified by some other CPU between the "while" statement and > > > > > + the call to do_something_with(). > > > > > > > > Nit: I guess references to variable names such as 'a' should be quoted > > > > (same for 'tmp', 'b', etc): > > > > > > > > For example, the optimized version of this code could result in > > > > passing a zero to do_something_with() in the case where the variable > > > > 'a' was modified by some other CPU between the "while" statement and > > > > the call to do_something_with(). > > > > > > > > which makes reading it less ambiguous and more fluid IMO. This > > > > observation applies to the whole document as 'a' is used in many > > > > places. > > > > > > Good point, fixed. > > > > Which reminds me -- the thing that makes me most nervous about > > prohibiting speculative stores is the bit about having to anticipate > > all compiler optimizations that might get rid of the needed > > conditionals. > > > > Thoughts? > > As long as current compiler versions behave I don't the potential of > future problems is a problem that can (or should) be solved via > documentation - there will always be a colorful tension between > specification and reality, both at the hardware, the code and the > compiler level ;-)
There certainly has been in the past. ;-)
> It doesn't hurt to outline our expectations in any case, agreed?
Fair enough, I will leave it as is.
Thanx, Paul
| |