Brian Ford
Thu May 13 20:40:00 GMT 2004

We seem to be seeing some significant thread contention issues in
malloc/free; presumably due to muto contention.  As such, I thought I'd
feel out the possibility of contributing a port of ptmalloc2.

Since it is used in glibc and is derived from the Cygwin's current Doug
Lea version, it shouldn't be too drastic a change.

The license for it, when obtained outside of the glibc distribution,
appears to be compatible with a contribution.

I assume there will be some issues to iron out for integration like:

1.) What needs to be done with the broken debug malloc support.
2.) Do we use pthread mutexes as it expects or muto's.
3.) How to deal with it's thread specific data requirement.
4.) It uses environment variables for tuning by default but this seems to
introduce a circular dependency.
5.) etc.

I'm sure there are more issues I haven't found yet, but I just wanted an
opinion before undergoing too much more effort.  Thanks.

Brian Ford
Senior Realtime Software Engineer
VITAL - Visual Simulation Systems
FlightSafety International
the best safety device in any aircraft is a well-trained pilot...

More information about the Cygwin-developers mailing list