[GSoC] Co-Processors Update #8

Hello all,

Logistical:

Progress:

  • Last week I successfully tested the turbo decoder using a DSP only
    program. I figured out how I can best share the code without breaching
    any
    license terms so I’ll have that up this weekend.
  • This week I was able to get shared memory working between the ARM and
    DSP. Basically I create a contiguous buffer using CMEM on the ARM and
    get
    its physical pointer. I pass the physical pointer to the DSP (no MMU)
    using
    MessageQ and it can directly manipulate the buffer I created on the ARM.
    Works well with a little test of assign value on ARM and read on DSP,
    assign on DSP, read on ARM. Lot of annoyances but finally tweaked the
    right
    knobs to get this working.
  • Next step is to now pass the physical pointer containing the
    log-likelihood ratios over to the DSP, and the DSP should be able to
    place
    the hard decisions in another physical pointer address that I pass to
    it. I
    can then test whether the resulting HDs match the expected HDs on the
    ARM.
    Should have this working early next week.
  • After that I’ll be able to play around with getting data in and out of
    GNU Radio so I can perform a full system integration test.

Things are moving!