Continued work on neon 64-bit correctness. This is really dragging out now. I had hoped to have had it fixed by now, but subtle bugs are subtle.
The 16-bit opcodes patch is now committed both upstream and in Linaro GCC 4.7, however, so that's some progress at least.
Posted benchmark results for the the 64-bit shifts in core registers. The results are inconclusive: the benchmark runs are too noisy.