Million-point FFT

Hi all,

I am interested in performing million-point FFT in FPGA. I've read some
papers and concluded that the best way of performing million-point FFT
would be having multiple smaller size FFT in parallel, for example 8x
16K-point FFT, with twiddle factor multipliers (I think it will be
radix-2^2 streaming FFT with multiple inputs)
But, since I am a newbie in DSP, I want to check if there is any easier
way. I am especially interested in a way where I can avoid parallel input.

Could you give me your idea/comment?
Thanks!

Thomas