Reducing the chore of multicore....


Earlier this week my colleague, Sven Brehmer, CEO and Founder of Polycore Software (TI’s multicore software tools partner), and I presented a webinar in conjunction with the Multicore Association titled: Optimizing the performance and portability of multicore DSP platforms with a scalable programming model supporting the Multicore Association’s MCAPI.  

We discussed the state of multicore programming and its requirements and challenges, and how state of the art silicon, tools and standards are working together to reduce the complexity of multicore programming and improve productivity and time to market.   If you were unable to attend the webinar live, fortunately it can be found for your convenience at: Optimizing the performance and portability of multicore DSP platforms..

 

 

 

 

We had quite a few engaging questions during the Q&A portion and were actually unable to answer all of them in the allotted time.  So, Sven has joined me today to answer the remaining webinar questions.

  1. Why would I use an existing multicore communications API (MCAPI) implementation instead of building the communications myself?
    1. Multicore communications can be challenging already in a two core architecture.  As soon as you go beyond a single unidirectional communication, management and synchronizing events gets complicated and requires mastery of communications.  Using an existing implementation will save significant development time, that you can spend on your value add, and provides application portability.  A multicore programming platform providing MCAPI support through a combination of runtime and tools, provides an abstraction and the programming effort is simplified to a configuration task.
  2. How do you compare MCAPI messaging with message passing interface (MPI) messaging?
    1. MCAPI and MPI are both message passing paradigms. MCAPI targets closely distributed computing using a thread of execution.  MPI is used in widely distributed computing which assumes the process model.  MCAPI is simple and designed to allow small footprint implementations.  MPI on the other hand is more feature rich, which comes with a degree of complexity and is used in environments where execution footprint is of little concern.  So while both are message passing models, they are targeting different areas of multicore computing, and could therefore potentially both be used in a system that involves both widely and closely distributed computing.
  3. I am using TI’s Code Composer Studio™ (CCS) integrated development environment (IDE) now; can I use Poly-Platform with TI’s CCS IDE?
    1. Yes, Poly-Platform is integrated via Eclipse with TI’s CCS IDE programming and test tool suite.
  4. How does the collaboration between TI and Polycore help developers?
    1. It significantly improves time to market and portability for multicore solutions on TI’s multicore platforms.
  5. How do I get started with evaluation of TI’s processor family as well as Poly-Platform?
    1. Obtain a TI TMS320C6678 evaluation module (EVM) from the TI eStore and get Poly-Platform evaluation version. TI’s Wiki site has a  getting started guide (http://processors.wiki.ti.com/index.php/BIOS_MCSDK_2.0_User_Guide#Poly-Platform_from_PolyCore_Software) and provides appropriate links to get started.

We invite you to “attend” the webinar at your convenience and check out the PolyCore Software multicore programming platform at: http://polycoresoftware.com/solutions/products

We would love to hear any further feedback and questions that you may have so go ahead; ask away and Do More with MULTICORE today!