Gnuradio and beagle board

Hi,

I’m testing gnuradio with the beagle board.
I’m using:

  • gnuradio 3.2.1 (with neon patch).
  • libusb-1.0-0

First, I’ve tried benchmark_dotprod_fff:

./benchmark_dotprod_fff

generic: taps: 256 input: 4e+07 cpu: 1119.820 taps/sec:
9.144e+06
cortex_a8: taps: 256 input: 4e+07 cpu: 43.391 taps/sec:
2.36e+08

(and I believe it is ok)

Then, I’ve tried usrp_benchmark_usb, with very low performance:

./arm-angstrom-linux-gnueabi-usrp_benchmark_usb.py
Testing 2MB/sec… [ 718.592163] usb 1-2: USB disconnect, address 2
[ 719.116760] usb 1-2: new high speed USB device using ehci-omap and
address 3
[ 719.283386] usb 1-2: configuration #1 chosen from 1 choice
gr_vmcircbuf_createfilemapping: createfilemapping is not available
gr_vmcircbuf_sysv_shm: shmat (2): Invalid argument
usb_throughput = 2M
ntotal = 1000000
nright = 994162
runlength = 994162
delta = 5838
OK
Testing 4MB/sec… usb_throughput = 4M
ntotal = 2000000
nright = 1982575
runlength = 1982575
delta = 17425
OK
Testing 8MB/sec… uUuUuUuUuOuOuOuOuOuOuOuOuOuOuOuOuOuOuOusb_throughput
= 8M
ntotal = 4000000
nright = 3720433
runlength = 22608
delta = 3977392
FAILED
Testing 16MB/sec…
uUuOuUuOuUuOuUuOuUuUuOuUuOuUuOuUuUuOuUuOuUuOuUuUuOuUuOuUuUuOuUuOuUuOuUM
ntotal = 8000000
nright = 1934045
runlength = 0
delta = 8000000
FAILED
Testing 32MB/sec…
uUuOuUuOuUuOuUuUuOuUuOuUuOuUuUuOuUuOuUuOuUuUuOuUuOuUuOuUuUuOuUuOuUuOuUM
ntotal = 16000000
nright = 167550
runlength = 0
delta = 16000000
FAILED
Max USB/USRP throughput = 4MB/sec

Do you have any suggestion?
thanks

andrea

On 01/15/2010 12:11 PM, andrea wrote:

cortex_a8: taps: 256 input: 4e+07 cpu: 43.391 taps/sec: 2.36e+08
(and I believe it is ok)

Then, I’ve tried usrp_benchmark_usb, with very low performance:

Which usb port did you attach the USRP to? EHCI or MUSB? You might get
better performance if you use EHCI (and you are lucky, or have a C4
board). I’ll double check what I get on my beagle and/or overo this
week. I use the EHCI port.

I do not expect an OMAP3 to run the USB bus at full wire speed. I would
expect to get enough data over usb to do some useful work. The real key
will be to do more processing in the FPGA though.

Philip