[llvm-commits] [PATCH] Thread-safe ManagedStatic

[Chris Lattner]

On May 18, 2009, at 6:25 PM, Owen Anderson wrote:

> So, now that we have the basic infrastructure more or less in place,  
> here's another attempt at making a thread-safe version of  
> ManagedStatic based on double-checked locking with explicit fences.   
> It also proved necessary to provide atomic update facilities for the  
> global StaticList as well.
>
> Please see the included block comment for more details on exactly  
> what safety guarantees this code provides.

The basic algorithm looks fine, however:

1. Please move the meat of the code for LazyInit out of line somehow.   
Only the case when tmp != 0 should be inline ideally.
2. Again, we only need a single global lock, we do not need a mutex  
per ManagedStatic.
3. Why do you create your own mutex out of atomic operations instead  
of just using a pthreads lock?  spinning is not always an efficient  
solution to contention, particularly if your machine really only has  
one cpu! (in which case you end up spinning away the rest of your  
timeslice)
4. Instead of going through the trouble to make the locking more  
efficient, I think it would be good to special case the scenario when  
there is only one LLVM thread.  Why not make clients explicitly opt-in  
to multithreaded llvm, by making an explicit llvm_multithread call?   
If multithreading is disabled, this call should return an error.  If  
enabled, it would set a global.  All the various "locking" clients  
could just check the global before taking potential heavy-weight locks.

-Chris