For evaluation of the code generated by GCC a number of benchmarks are available. Some people run these benchmarks on a regular basis and therefore allow to monitor how GCC's optimizations evolve over time and whether optimizations really make a difference on the tested benchmarks.
AMD Athlon, AMD Bulldozer, Intel Haswell and Intel Itanium SPEC CPU 2000 and SPEC CPU 2006 results are available at http://gcc.opensuse.org.
SPEC CPU 2000 results on Intel Core2 are available at https://vmakarov.fedorapeople.org/spec/index.html. The scripts used are based on original scripts by Diego Novillo.
SPEC CPU 2006 results on Intel Core2 are available at the same place as SPEC2000 ones: https://vmakarov.fedorapeople.org/spec/index.html.
Statistics about GCC code size for several targets are available from the GCC Code-Size Benchmark Environment (CSiBE), along with the testbed and measurement scripts.
SUSE runs various other C++ benchmarks and Polyhedron. Results can be found at https://gcc.opensuse.org.
Tobias Burnus also runs the Polyhedron benchmark with mainline GCC. Results can be found at http://users.physik.fu-berlin.de/~tburnus/gcc-trunk/benchmark/.
Diego Novillo compiled a collection of relevant scripts at https://gcc.gnu.org/wiki/PerformanceTesting
Openbench is a free benchmarking suite similar to SPEC.