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.

TMS320DM8168: Adjusted Peak Performance (APP) on DM8168

Part Number: TMS320DM8168
Other Parts Discussed in Thread: AM3874

Hi all,

May I know the Adjusted Peak Performance (APP) on DM8168?

Thanks in advance.

B.R.

OC

  • Hi Omin,

    We had similar question regarding AM387x device. You can re-use the info for DM816x device:

    AM3874 @600MHz    APP = 1.8 x 10^-4 WT

    AM3874 @800MHz    APP = 2.4 x 10^-4 WT

    Calculation for ARM Cortex-A8 with VFPv3:

    Frequency (clock speed), F = 600 MHz

    Floating-point operations, FPO = 1

    Architecture adjustment factor, W = 0.3 (for non-vector)

    APP = 1 (Cores) * 0.6 (GHz) * 1 (FPO) * 0.3 (W) * 10-3 = 0.00018WT

     

     

    From ARM Tech Support;

    -----------------------------------------------------------------------------------------------------

    The calculation depends on FPOi (the number of 64b floating point operations per cycle) Fi (processor frequency) Wi (architecture adjustment factor) Where Wi = 0.9 for vector processors or Wi = 0.3 for non-vector processors

    "Note: In determining FPO, include only 64-bit or larger floating point additions and/or multiplications. All floating point operations must be expressed in operations per processor cycle; operations requiring multiple cycles may be expressed in fractional results per cycle.

    For processors not capable of performing calculations on floating-point operands of 64-bits or more the effective calculating rate R is zero."

    So we can consider FPO=0 for the ARM core since it does not operate on 64-bit values.

    The VFP can operate on 64-bit double-precision values.

    The VFP can issue one instruction per cycle and can complete one 64-bit ADD per cycle, so FPO=1.

    "Note 7: A "vector processor" is defined as a processor with built-in instructions that perform multiple calculations on floating-point vectors (one-dimensional arrays of 64-bit or larger numbers) simultaneously, having at least 2 vector functional units and at least 8 vector registers of at least 64 elements each."

    The VFP performs short vector operations but it does so serially - it does not perform parallel operation on multiple 64-bit values in a single cycle. So we should consider it as a non-vector processor, Wi = 0.3.

    You need to specify the cpu frequency. If we assume 600MHz then

    APP = FPO x F x W = 1 x 600 x 10^-6 x 0.3 = 1.8 x 10^-4 WT (weighted teraflops)

    ----------------------------------------------------------------------------------------------------------

    Regarding the parameters of CA8 , FPOi is 1 if your product has VFP, and W is 0.3.

     

    Regards,

    Pavel

  • In below e2e threads we have info also regarding AM437x and AM572x devices:

    e2e.ti.com/.../576485
    e2e.ti.com/.../591278

    Regards,
    Pavel
  • Hi Pavel,

    Thanks so much for your reply.
    That is what I want.

    B.R.
    OC