By "real error" I mean java.lang.Error or something worse.
For some reason I thought I remembered one of Doug's past papers
alluding to this possibility in multi-processor environments.
Bob
victor.luchangco@sun.com wrote:
>Hi Bob,
>
>
>
>>In the current memory model, what are the implications of not
>>synchronizing? If you have multiple processors, can it cause a real
>>error or will the value just be incosistent?
>>
>>
>
>I'm not sure what you mean by "a real error". If you don't synchronize
>access, you may get inconsistent results. If your program can handle
>getting inconsistent results (i.e., it knows how to make sense of
>them, or else to ignore them), then you should be fine. However, many
>programs seeing an inconsistent result will proceed to do operations
>that they should not do, which typically results in an error in your
>program. Such an error may be noticeable to the program's user, but it
>may also cause the program to silently return the wrong result, which is
>even worse in most cases.
>
> Victor
>-------------------------------
>JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
>
>
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:52 EDT