This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Linux/DRA725: omapconf DDR Bandwidth test

Part Number: DRA725

Tool/software: Linux

Hi,

   My customer use omapconf v1.72 and below command on DRA72x to test the DDR Bandwidth.

omapconf  trace bw --tr r+w --m0 dss --m1 gpu_p1 --m2 gpu_p2 --m3 ipu2 --m4 sdma_rd --m5 alldmm

 

SW Build Details:

  Build:

    Version:  _____                    _____           _         _  

  Kernel:

    Version: 3.14.57

    Author: ubuntu@ubuntu-desktop

    Toolchain: gcc version 4.7.3 20130226 (prerelease) (crosstool-NG linaro-1.13.1-4.7-2013.03-20130313 - Linaro GCC 2013.03)

    Type: #1 SMP PREEMPT

    Date: Thu Jan 5 22:56:26 CST 2017

 

Counter: 0  Master:             dss Transaction: r+w Probe: emif1

Counter: 1  Master:          gpu_p1 Transaction: r+w Probe: emif1

Counter: 2  Master:          gpu_p2 Transaction: r+w Probe: emif1

Counter: 3  Master:            ipu2 Transaction: r+w Probe: emif1

Counter: 4  Master:         sdma_rd Transaction: r+w Probe: emif1

Counter: 5  Master:          alldmm Transaction: r+w Probe: emif1

delay in us: 1000000

overflow delay in us: 1000000 (iterations=1)

accumulation type: 2

iterations (0=infinite): 0

Overflow counter index: DISABLED (overflow delay used)

Overflow threshold: DISABLED (overflow delay used)

              Time Stamp(32KHz ticks) ->   Throughput(MB/s)

time:         End       Start   Delta ->   Counter0   Counter1   Counter2   Counter3   Counter4   Counter5

time:    47593272    47560479   32793 ->     221.00     809.71     976.73       3.59       0.52    2011.99

time:    47626066    47593272   32794 ->     221.00     809.54     976.33       3.59       0.52    2011.41

time:    47658858    47626066   32792 ->     221.00     806.11     972.76       3.59       0.52    2004.43

Question is the DRA72x should have a single core GPU, why both gpu_1&gpu_2 have the measurement number?

Best Regards.

   Marvin

  • Hi Marvin,
    There are two 128bit initiator ports for L3 interconnect for GPU.
    Other IPs which have two ports are GC320, VIP, VPE and EVE.

    The data transfers here involve both the ports, hence we see _p1 and _p2 as master initiators for these IPs with omapconf
    Please refer TRM Chapter 14 for more details.
  • Prasad,
    Thanks for your reply. So actually 'omapconf trace bw' only trace the initiator's bandwidth allocation of L3 BUS, right?
    Would I know the L3 total theory bandwidth?
    Best Regards.
    Marvin