This is really very well done, thanks for putting this together!
Two questions from the blog:
(1) QEMU Xilinx ARM emulate NEON Instruction ?
(2) QEMU Xilinx ARM emulate double precision FPU ?
If you dive into the source code of qemu, specifically within the qemu git repo here:
within the target-arm folder there is a file called translate.c. Within this file there is a function called disas_neon_data_insn(). This handles the NEON instruction decoding.
As for the double precision, I see no reason why it would not, as the ARM emulation should be operationally equivalent.
I have a post about linpack performance on the Zedboard, I'll see if I can write a quick follow up about the performance in qemu - that will confirm the double precision float using the NEON core.
I have written a blog post detailing the QEMU environment available from Xilinx.