The Hoard Scalable Memory Allocator

This release incorporates a number of fixes and improvements: better per-thread allocation, improving speed for Unix platforms that do not support __thread; interception of the Solaris threads API (thr_*) to work with older Solaris programs; a fix for a possible race condition; and increased checking to protect against heap corruption or other errors. Hoard now uses GNU-supported hooks on platforms with glibc (especially Linux), allowing it to work better with legacy programs.

This release fixes a serious bug for Windows in thread reclamation. It fixes a small memory leak in thread creation. It substantially increases performance for small objects on 64-bit platforms. It works on 64-bit Windows.

The fast path for allocation was streamlined,
substantially accelerating both 32-bit and 64-bit
code. pthread_* based thread-specific data is now
used for x86-64, where it's faster. These combine
to improve benchmark performance on x86-32 by up
to 13%, and on x86-64 by up to 32%.

This release has fixes for building on Mac OS X (Darwin) and to support some older versions of the Sun compilers. It adds FreeBSD as a supported target.

An x86-64 target was added for Linux/GCC along with support for posix_memalign (for Unix systems), and a static library build option for Linux.