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.

STK-MED with the C6454

Other Parts Discussed in Thread: MIDAS

TI supplies examples with the stk-med 03.00.00 for two dsp platforms, the C6472 and C6678.  I would like to use the stk-med toolkit and am having difficulty targeting our platform, the C6454.  I'm thinking I could dearly use either a) an example for the C6454 using the stk-med, or b) any example using the C6454 platform. 

rfix

  • Hi Richard,

    The libraries are specific to C64x+ and C66x and should be compatible with any platform with those cores. However, as you pointed out, the test projects (unit-test benches for the libraries) have been tested only on the newer 6472 and 6678 devices as examples that users can leverage to take forward to the C64x+ and C66x based platforms in their use case. 

    I understand from your previous post (http://e2e.ti.com/support/embedded/bios/f/355/t/146105.aspx#527782) that you have STK-MED v2.0 too, where the test projects were tested on 6455. Since that was based on CCSv3.3 and older software components we had recommended moving to STK-MEDv3.0. However, you could always use that code as a reference in your porting effort.

    If you face any specific issues in your porting effort, we can try to help you resolve it. Were you able to get past http://e2e.ti.com/support/embedded/bios/f/355/t/146105.aspx#527782 based on the recommendations there?
    Please do let us know and if there's something specific, and do include the complete build logs so we can help you better. 

    Regards,
    Uday 

  • Is it possible for TI to provide a similar project example for the scu_ires based on the C6454 (or C6455) platform?

    I believe that would help me greatly.

    rfix

     

  • Following is a more detailed description of my current issue.

    My starting baseline is ccsProj_C6472.  I created a platform according to my understanding of the C6454 memory space.  I am able to compile and simulate.

    During simulation, I observe that the RMAN_assignResources call returns a value of IRES_ENORESOURCE.  I'm finding it difficult to identify where to proceed to troubleshoot theroot cause.  I believe its related to either my platform, or the ti.sdo.fc.edma3.settings.regionConfig file (which I have left unmodified from the c6472).

    I have exhausted my knowledge level on this problem and need assistance to get to the next step.

     

    rfix

  • Richard,

    As per your request, I've created a 6455-based 'scu_ires' test project as a reference to help with your 6454 porting effort. This successfully runs on a 6455 DSK and the outputs seen are bit exact with the reference outputs. The software dependencies are the same as the other STK-MED-v3.0 modules.

    Please unzip the package 'stk-med-scu_ires_6455-test-project.zip' 1768.stk-med-scu_ires_6455-test-project.zip and copy the folder 'ccsProj_C6455' to the location 'stk-med-03.00.00\miAlgos\scu_ires\test\.'

    Regards,
    Uday 

  • I got your test project and am much appreciated for your efforts.  I've only had 15 min with it, but I wanted to update you before I have to leave work.

    1.  I noticed the code generation tools were v7.2.1 and when i changed to 7.0.3 it complained but continued on when i closed the drop-down.

    2.  In the settings.xdctools.package repositories, i had to add "FRAMEWORK_COMPONENTS_CG_ROOT/fctools/packages" for it to find ti/xdais.

    3.  I simulated to the RMAN_assignResources line and got the same error, IRES_ENORESOURCES.

    Don't know quite what to make of these observations, but I'll continue to play around with it in the morning.

     

    rfix

  • Hi Uday,

    The error that Richard is seeing sounds like it is related to the the EDMA3 Driver. Are you using the EDMA3 driver from the MCDSK for the C6472 (02.10.03.04)? Can you verify the versions of the tools that you used in this build? Richard is following the readme file from the MED STK to the letter (for example, his reason for switching to CodeGen 7.0.3 that the SCU example asks for CodeGen 7.03 and the FFT is 7.2.1). Thanks for all your help on this issue.

  • I inadvertently mixed up the version numbers; was looking at version numbers for MIDAS (our demo package) instead of STK-MED. I apologize for the mix up.

    I'll fix this and try and get back to you today with a new zip package that is consistent with STK-MED v3.0 version numbers.

    -- Uday 

  • Today I'm divided between two tasks and will not get back to the dsp until later. I wanted to report a finding though.

    There are evidently two places to specify the edma3 low level driver. One is for the rtsc and the other for the xdctools.

    For proj6472 (working from TI), in properties.ccsBuild.rtsc, the edma3 lld product lists 02.10.02 (unchecked).  In properties.c/c++ build.settings.xdctools.package repositories, the path is  EDMA3_LLD_INSTALL_DIR=edma3_lld_02_10_03_04 (windows environmental variable).

    For the proj6455 received yesterday, when the checkbox IS checked the error manifests, and when unchecked - NO error.

    I tried the same with my original project without success.   I will attempt further investigation later and report.

    Thanks for assistance.

    rfix

  • Richard,

    I corrected the 6455 project so that the version numbers for software components are consistent with the other STK-MEDv3.0 ultrasound modules, and was able to run the testscu.out without errors on the 6455. However, when I switched to the 6454 simulator, and recompiled, I do see the same 'RMAN Assign Resource Failed [7]' error as you've observed.

    I will step-through the code and investigate; will need to get back to you on it. In the meantime, if there's someone else on the forum who has some thoughts on this, please do chime in. 

    -- Uday

    PS: A note on the RTSC platform. As you might have noticed, we use the custom 'plat6455' RTSC platform package for the project (located at 'ccsProj_C6455/platformRepo/plat6455'). I compared the 6454 and 6455 memory maps from the datasheets and it seems like the only thing different in the memory map is a higher address range for L2 in 6455 (since it has 1MB more). However, since 'plat6455' doesn't try to access that memory range anyways, the same customization of the address table as defined in 'plat6455' can be used for a similar 'plat6454' RTSC platform. 

  • Hi Richard,

    I've attached a new zip package (ccsProj_C6454.zip) which contains the scu_ires test project that runs successfully on the 6454 simulator (little endian). Please unzip this as miAlgos/scu_ires/test/ccsProj_C6454/

    6864.ccsProj_C6454.zip

    Here's an explanation on why we were observing the runtime issue, and the solution:

    I spoke to a member of the 'framework components' team at TI (Gunjan Dang), and identified that the global EDMA configuration for 6455 is defined within framework components (see code after line 4265: "#ifdef C6455" in file 'framework_components_3_20_01_26\packages\ti\sdo\fc\edma3\edma3lld_cfg.h'). However, the global configuration for 6454 is missing in this file (you'll see that there is no "#ifdef C6454"). The RMAN/IRES interface (part of framework componets) are used to access and allocate EDMA resources. Since the global configuration for 6454 was missing, RMAN/IRES were unable to allocate any EDMA resources. 

    The solution was to manually define a global configuration for the 6454 device in our 'testscu' project. In the attached zip you'll see that there is now a new file called edma3_6454_globalConfig.c, which defines the global configuration for 6454. This global configuration is then manually referred to in the "bios6.cfg" RTSC configuration file (part of the testscu project) as "EDMA3_PARAMS.globalConfig = 'ti_sdo_fc_edma3_lld_globalConfigParams';"

    Please do let us know once you have verified that it works for the 6454 simulator in your environment too or click on 'verify answer'.
    If you face any further issues, please do post back.

    -- Uday

  • This looks awesome, but I am unable to confirm operation because the edma3_c6454_globalConfig.c file has an invalid path.   Can you provide that file to me?

  • I should have looked in the directory first, the config file is there, I just need to point to it.  I'll try that and get back with results.

  • Yup that should work. The right way should have been for me to include a relative path to the file (just like other source files in the project) instead of an absolute path.

    -- Uday 

  • Hi Uday :

    Firstly I am very glad to see this article.

    The STK-MED v2.0 is my searching for these days because of the demand of the Project  which I was involved in.

    But I have found the STK-MED v3.0, so could you please tell me where I can download the  STK-MED v2.0 which is so important for me now ?

    Thanks !

    Regards,

    kris