At 2:00 PM -0700 7/25/03, Jerry Schwarz wrote:
>Personally, I always thought the no out of thin air requirement was
>unreasonable. Among other things it constrains optimizers in ways
>that have nothing to do with threading and synchronization. That
>is, I've always taken the goal to be to allow the optimizer to treat
>straight line code without any thread related operation (however
>that is defined) as if no threading were present. But consider
>
>
>Thread 1
> c = ...
> x = 3 * c - 1
>
>
>Thread 2
> y = x
>
>The not out of thin air requirement says that the compiler can't
>transform Thread 1 to
>
>
>c = ...
> x = 3 * c
> x = x - 1
>
Jerry,
You seem to be suggesting a model where any incorrectly
synchronized read can return absolutely any value, with no
constraints from the memory model.
I think this is a really bad idea.
Data races should not become the buffer overflow of the 21st
century. The biggest problem with buffer overflows is that they allow
completely arbitrary behavior. Data races in Java programs are going
to be at least as common as buffer overflows in C code. Data races
should not allow arbitrary values to be returned.
Bill
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:47 EDT