This is the mail archive of the mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] new mutex implementation

On Thu, Aug 15, 2002 at 09:19:52PM +0200, Thomas Pfaff wrote:
>This patch contains a new mutex implementation.
>The advantages are:
>- Same code on Win9x and NT. Actual are critical sections used on NT and
>kernel mutexes on 9x.
>- Posix compliant error codes.
>- State is preserved after fork as it should.
>- Supports both errorchecking and recursive mutexes.
>- Should be at least as fast as critical sections.
>- Will make us all rich and happy.
>Unfortunately the pthread_mutex_trylock call requires
>InterlockedCompareExchange that is not available on Win95.
>See my next patch for a workaround.
>Just like critical sections it will use a counter and a semaphore to block
>other threads. The semaphore is only used when at least one thread is
>waiting, otherwise a kernel transition is not needed.

This sounds suspiciously like a 'muto'.

I mentioned a while ago that the muto concept might be useful for
generic use in pthreads but that I didn't want to add any overhead to
the existing muto definition, preferring (essentially) a copy + paste
+ augment into the pthread call.

IIRC, Robert then posted a change to the muto implementation which added
existing overhead to the existing implementation, i.e., the opposite of
what I'd asked for.

I may be misremembering this but you should be able to find the discussion
in the cygwin-developers/cygwin-patches archives.  I doubt that the word 'muto'
shows up too often there.

However, I'd be interested in looking at your implementation.  I don't think
you attached the actual patch to your message.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]