On Thu, 22 Jan 2009, Mark Taylor wrote:
> On Thu, 22 Jan 2009, Peter W. Draper wrote:
>
>> I think the reason this should be OK for us is just that we synchronise all
>> calls to AST, so only one thread can be running AST code at any time. Since
>
> I think we should be able to do better than that - it ought to be possible
> for JNIAST to do locking on a per-object basis now that AST can cope with
> concurrency, rather than using one big mutex for the whole library which is
> what we had before. Unless there's some gotcha which I haven't spotted yet,
> that is.
What about the methods that use more than one object? You'd need to
acquire a lock on them all before proceeding.
|