ENERGY, SIZE & PERFORMANCE OPTIMIZATION- SOMNIUM® DRT

Document created by Bill Krakar Employee on Jul 10, 2015Last modified by Bill Krakar Employee on Mar 15, 2016
Version 9Show Document
  • View in full screen mode

ENERGY, SIZE & PERFORMANCE OPTIMIZATION:

EEMBC® Autobench™ DEMO USING K64 and SOMNIUM® DRT NXP EDITION

 

 

This Demo Is Probably of Interest If You:

  • Want smaller, faster and more energy efficient software
  • Develop for Kinetis Devices

eembc_drt_2_1_versus_kds_3_0_newlib.png

 

Description

 

SOMNIUM are members of the EEMBC Automotive Subcommitee and use EEMBC Autobench as part our of product development and validation process.

 

AutoBench is a suite of benchmarks that allow users to predict MCU performance in automotive, industrial, and general-purpose applications. Its benchmark kernels include the following:

 

  • Generic Workload Tests

Bit manipulation, matrix multiplication, floating point arithmetic, cache busting, pointer chasing, PWM and other features typically used in encryption algorithms.

 

  • Signal Processing Algorithms

Including algorithms important for sensors - FFT, iFFT, FIR, iDCT and IIR.

 

  • Basic Automotive Algorithms

Include CAN, tooth-to-spark, angle-to-time conversion, road speed calculation and table lookup/interpolation

 

The results achieved clearly demonstrate the practical impact of DRT's benefits - over 50% energy reduction and ROM/RAM reductions of 30%/70% with performance increases of over 15%. When applied to IoT designs this enables usage of the wide range of Kinetis devices allowing you "do more with less", extracting more functionality, performance and compute from Kinetis devices.

Attachments

    Outcomes