Unable to run Epiphany tests

[ach1000]

Just gotten around to playing with my Kickstater 16-core Parallella gen1 board. The Linux side works just fine, however the board locks up when attempting any of the Epiphany tests; unresponsive to network/keyboard/mouse input and requires a hard-reset.

This is using the standard Ubuntu 14.04 release, with the HDMI kernel and FPGA binary.

Board: Kickstarter 16-core, ZYNQ 7020
PSU: 5v 2A (part no. 993-1247-ND from Digikey)
Cooling: 30x30x6.9 fan (part no. 259-1572-ND from Digikey)
Temperature: ~60degC according to xtemp
Kernel: kernel-hdmi-default.tgz (MD5 = 8eab4b6228e32a0432279f76de51300f)
Bitstream: parallella_e16_hdmi_gpiose_7020.bit.bin (MD5 = d920ca30ce6c1a9b1da841fa18e5dda3)

I suspect I'm seeing the same problem that was reported here: viewtopic.php?f=50&t=1401&p=8780

Any suggestions?

[aolofsson]

Sorry to hear that! What tests did you run and when did you get your board?

[ach1000]

[ralphmcardell]

@ach1000, @aolofsson:

This sound very similar to problems myself and mtimms2 have experienced, see topic:

viewtopic.php?f=50&t=1438

Regards

Ralph

[mtimms2]

Yes this is the same as my board which arrived at similar time in the UK.

SN 0001717
SKU A101040
Zync 7020 device.

I have tried various Linux images and both HDMI and headless FPGA versions and the epiphany tests always fail with a lockup which requires a power cycle.

[paulb]

I have a new Parallela-16 Micro-server with a zynq-7010 chip and the big heat
sink covering both the Xilinx chip and the Epiphany chip. I have experienced this problem.
When I try to run any test on the Epiphany III, the entire system locks up. Or, at least all
communication via the Ethernet quits. I am currently focused on the e-reset
test found in the directory /home/linaro/epiphany-examples/test/e-reset because
this appears to be the most basic test. Traces show that the program
successfully calls the e_init library routine; but, that does not seem to do
much other than read the configuration file. When the e_reset library routine
is entered, the file /dev/mem is opened and a memory map call completes and
then ... nothing. Has anyone resolved this issue?

In an act of desperation, I tried booting with the kernel that contains the
HDMI. That failed to boot and I pursued it no further.

Are the source codes for the e_* library routines available? I have not yet
tracked them down.

[9600]