• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 99 subscribers
  • Views 802 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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/TDA2: TDA2H and TDA3x ：DSP benchmarks

user4643736
Intellectual 920 points
Part Number: TDA2


Tool/software: Linux

Dear TI experts，

      When i run below test code on dsp，i found out the TDA3x's DSP is more efficient.

Test Code:

 

#define TSC_getDelay(preTSC) ((unsigned int)((0xFFFFFFFFl+TSCL)-(unsigned long long)preTSC)+ 1)

short *T0;

 

T0 = (short*) Utils_memAlloc(UTILS_HEAPID_DDR_CACHED_SR, 10240, 8U); //DDR

 

{

         int t=0;

         int t0_time;

         int pclk;

         TSCL=0;

         pclk=TSCL;

         for(t =0;t<10240;t++)

{

         T0[t] = t;

}

         t0_time= TSC_getDelay(pclk);  

Vps_printf("DEBUG:---DDR  = %d Result = %d ",t0_time,T0[1024]);

}

Result:

      [DSP1  ]     38.649484 s: DEBUG:---DDR  = 12412 Result = 1024       //TDA3X with 709M DSP ;VSDK2.12

   [HOST] [DSP2  ]     58.513542 s: DEBUG:---DDR  = 43946 Result = 1024     //TDA2H with 750M DSP; VSDK2.12 linux

Any ideas what is going on? Thank you.
Best Regards

Sid

 

  • 0
    TI__Expert 5575 points

    Have you checked if they are both built in release mode?  You can double check the asm file generated for this file and see that the ASM code for this code section is the same for both.

  • 0 in reply to Jesse Villarreal
    TI__Genius 13395 points
    also make sure no task preemption during the execution of this code, it can be assured by interrupt disable and enable before and after the code
    as below

    UInt hwiKey = Hwi_disable();
    ::::::::::::::::::::::::::::::::::::
    code
    :::::::::::::::::::::::::::::::::::
    Hwi_restore(hwiKey);


    regards, Shiju
  • 0 in reply to Shiju Sivasankaran
    Intellectual 920 points
    Hi Jesse & Shiju,
    Thanks .
    I found out when run the test code before my usecase started on tda2x, the result will be close to tda3x. The tda3x is only run a simple chains for test.
    when my usecase start on tda2x, there will have other three alg plugin links also run on dsp2, therefore the test code will consume more times to run.
    I think i need Optimization my alg code.

    Best Regards
    Sid