On Fri, Apr 01, 2011 at 02:13:53PM -0700, Matthew Dempsky wrote:
> On Fri, Apr 1, 2011 at 1:39 PM, Stephen Wilson <[email protected]> wrote:
> > So I think the main question is:  does having the full 64 bit pointer as
> > an ID give useful info to the developers using user-space threads?
> 
> Is there a reason not to just make the thread ID a 64-bit value and be
> done with it?  Seems like doing anything else is over-engineering the
> problem.

I was not trying to bike-shed.  Just understand the problem.

If there is "good information" in the upper 32-bits then we should
display it in our logs to save developer/human time when debugging.

On linux we currently have a ProcessMessage class (might replace it in
time, but no matter).  I actually do care about its size (it has a tid_t
in it) since it is passed around by value.  When processing internal
events we could be generating thousands of these guys per second so
efficiency does matter -- not by much, but perhaps a little.

So these types of things are worth thinking about.


-- 
steve

_______________________________________________
lldb-dev mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev

Reply via email to