Hoard Allocator

Is there any way to use the Hoard allocator (ww.hoard.org) for C++ with pgCC? I have been trying to get it to work but to no avail.

I read that using Hoard helps if you use the STL in multithreaded applications. Is there anything else I can do to speed up the STL implementation with threads? My application is currently twice as slow when computing with one thread per CPU (Opteron system) than it is with just one CPU doing all the computing. I thought maybe it could have something to do with it heavily relying on hash maps, vectors, etc.

Scott

Hi Scott,

I was able to free up some time today to see if I could get Hoard working, but like you, encountered problems. I had trouble determining how to build it since there isn’t a top level “./compile” script as the build instructions on the web site indicate. I tried compiling by hand using g++, but encountered many syntax errors. I got farther when I tried compiling on a Redhat 7.3 system, but still no luck. The web site seems to indicate that Redhat 6.0/6.1 are the only supported Linux systems? Please let us know if anyone has had better luck than we’ve had.

I personally have not done a lot of multi-threaded C++ programming so may not be the best of help here. However, I do know that the STL and libC are not protected for reentrant execution. Could the slow down be attributed to the use of mutexes in your code which are used to manage your shared data objects?

  • Mat