Some more info: I have added zero-copy buffers (1000) as advised by libusb. While on USB3 odroidxu no error occurs, on usb2 I get the following error:
 libusb: error [op_dev_mem_alloc] alloc dev mem failed errno 19 

C error #19 means "no such device" google advises. 
What could it be?

On Sat, 29 Dec 2018 at 07:07, Ioannis Makris <makrisj@gmail.com> wrote:
Hello and thanks for the amazing piece of software. 
Despite having managed to do much stuff with it under x64 linux, I failed miserably to make it work under ARM architecture.
To be more specific, I tried the Odroid-XU usb3.0 capable board and the Odroid-C1 Usb2.0 board to no avail.
Odroid XU kernel: 3.4.y Odroid-C1 kernel: 3.10.107

I used all reasonable ranges of sampling rates and here are the results:
Program seems to run normally. No error codes and.. no buffer underruns at all. In my i3 laptop I get some buffer underruns but to arm I have never spotted such a message no matter what the sampling rate is. 

When the sampling rate is a fraction of what fl2k_test reports as maximum, I see the sampling rate spike on my spectrum analyzer (Advantest TR4132). But when I opt for fl2k_fm, I see no response whatsoever other than the enabling of the DAC visible as an increase of a few dB's of noise and a hint (3dB C/N)  at the sampling frequency.

I tried the binary repositories of libusb for my distribution, and I even compiled locally libusb from source - nothing changed. 

What could be making such a mess? I have spent over 3 days on efforts to no avail. I just want to explore upconversion possibilities but I need to encapsulate both the host and the FL2000 into a shield in order to get meaningful experimental results. 

Please help if you can
Thanks 
Ioannis (John) Makris
SV9OFO