mcelog is a daemon to handle machine check events (hardware errors) on x86-64 machines running an x86 Linux kernel. It accounts and logs CPU and memory errors, supports triggers on error thresholds, and can predictively offline memory pages and CPUs based on error trends. This daemon should run on all x86 Linux systems that want to handle hardware errors.
autoboot is a job scheduler/watchdog to automatically compile/boot and run test suites with experimental Linux kernels. It runs from a central server and a pool of clients. The central server builds various kernels, then automatically boots a subset of them on the clients and runs test suites (like autotest). The server is very careful to watch the clients for hangs and power switch them as needed, and will also automatically fetch serial logs from a console server. All the resulting information is stored in a unique output directory for each for easy post processing. autoboot is a collection of bash shell scripts. It will need some adaption for local infrastructure.
gccwrap 32bit is a set of simple wrapper scripts to make it easy to compile 32-bit only programs with uncooperative Makefiles on 64-bit systems. You can just use "32bit make" to compile them and the wrappers will transparently add any -m32 options needed and also do some other transparent fixups. This package requires a Linux distribution with proper FHS compliant -m32 / multilib support like SUSE, Fedora, or Mandriva.
ftracer is a simple user space implementation of a Linux kernel style function tracer. It allows you to trace every call in instrumented user applications. It is useful for debugging and performance analysis due to its fine grained time stamp. This allows you to do control flow oriented debugging without any special instrumentation. So if the program does something unexpected, it's easily possible to look at the function calls before that, and use that to deduce the cause of the problem. ftracer relies on gcc generating a call on top of every function call. The tracing slows every function call down (about 3x). The tracing is per thread and does not create a global bottleneck. It supports a dump function in C, directly callable by the program or on exit, and a gdb function to dump from gdb.
pmu tools is a collection of tools for profile collection and performance analysis on Intel CPUs on top of Linux perf. It has a wrapper to "perf" that provides a full core event list for common Intel CPUs. This allows you to use all the Intel events, not just the builtin events of perf. Support for Intel "offcore" events on older systems that do not have support for this in the Intel. Offcore events allow you to profile the location of a memory access outside the CPU's caches. It implements a workaround for some issues with offcore events on Sandy Bridge EP (Intel Xeon E5 first generation). This is automatically enabled for the respective events, and also available as a standalone program. Some utility programs to access pci space or msrs on the command line. A utility program to program the PMU directly from user space (pmumon.py) for counting. This is mainly useful for testing and experimental purposes. A library for self profiling with Linux since Linux 3.3 (for self-profiling on older kernels, you can use simple-pmu. An example program for address profiling on Nehalem and later Intel CPUs (addr). A program to print the currently running events (event-rmap).
tsx-tools provides headers and utilities for development with TSX (HLE/RTM) hardware transactional memory development on recent Intel CPUs. It includes compat headers that provide C inline functions support for RTM and HLE in various formats for older gcc compatible compilers (before gcc 4.8). These headers are useful to enable lock elision in a lock library or existing program. It also has various TSX related utilities, including a tool that detects the presence of TSX, a TSX aware assert, and various debugging utilities.
snappy-c is a C port of the google snappy compressor (http://code.google.com/p/snappy/). The compressor is very fast with a reasonable compression ratio. It is mainly useful for projects that cannot integrate C++ code, but want snappy. It also contains a command line tool, a benchmark, random test code, and a fuzz tester. The compression code supports scather-gather and linear buffers. The scather gather code is ifdefed (-DSG) and can be removed with unifdef.
spooky-c is a C version of Bob Jenkin's spooky hash. The only advantage over Bob's original version is that it is in C, not C++,'and comes with some test and benchmark code. This is a very competitive hash function, but is somewhat unportable (64-bit little endian only). It's more portable than some of the contenders like CityHash.