[Haskell-cafe] Concurrency questions

Don Stewart dons at galois.com
Mon Jan 7 16:43:26 EST 2008


andrewcoppin:
> Spencer Janssen wrote:
> >On Sun, Jan 06, 2008 at 11:30:53AM +0000, Andrew Coppin wrote:
> >  
> >>Just a couple of things I was wondering about...
> >>
> >>1. Is there some way to assign a "priority" to Haskell threads? (The 
> >>behaviour I'd like is that high priority threads always run first, and 
> >>low priority threads potentially never run at all unless there's an 
> >>available processor which is completely idle.)
> >>    
> >
> >Not in the current system.  It is not clear that thread priorities are so 
> >nice
> >anyway (see 'priority inversion' on Wikipedia, for example).
> >  
> 
> Well, I was thinking more of using them for two things. One is for 
> speculative work (i.e., doing work which we might need later - but don't 
> bother unless there's cores going spare). The other is for working on 
> entirely independent tasks - I don't see how inversion can happen when 
> one thread is solving problem X and another is solving problem Y. But 
> sure, it certainly adds more complexity to have priority levels.
> 
> [I can also imagine situations where you might want to assign 80% CPU to 
> one thing, and 20% CPU to the other - but that really does sound hard to 
> implement...]
> 
> >>2. I have a situation where I have a thread generating some data and 
> >>putting it into a mutable array, and another thread trying to read that 
> >>data. Is there a way I can make the reader thread block if it tries to 
> >>read a cell that hasn't been computed yet, but not introduce too much 
> >>overhead for cells that have been filled in?
> >>    
> >
> >I'd probably use an Array of TMVars, they should be faster than MVars when
> >multiple threads are reading simultaneously.
> >  
> 
> Mmm, OK. I'll try that. (I wasn't actually aware that STM is working 
> yet...)

It works, and has done so for a couple of years now. It's used in
production systems. Where'd you hear otherwise?

-- Don


More information about the Haskell-Cafe mailing list