From Markus Levy

July 2017

The Multicore Association®



  • New Version of EMB2 and MTAPI
  • SHIM Working Group Reconvenes to Enhance Specification
  • Benchmarks for Heterogeneous Compute

New Version of EMB2 and MTAPI

EMB2 or the Embedded Multicore Building Blocks is available now for download at GitHub. Various frameworks are available for parallel programming, mainly targeting desktop and server applications or high performance computing, but unlike EMB2 these don’t take into account typical requirements for embedded systems such as predictable memory consumption, task priorities or support for heterogeneous systems-on-a-chip. The new features in EMB2 include a) fully compliant MTAPI reference implementation plus C++ wrappers for convenient task management; b) support for heterogeneous systems at all levels (MTAPI and high-level algorithms / patterns); c) ready-to-use plugins for OpenCL, CUDA, and distributed systems communicating over a network; d) - wider platform support and easier portability to other hardware architectures (which means you should start using it today). More information

SHIM Working Group Reconvenes to Enhance Specification

This week, the SHIM working group met for the first time to begin finalizing the details for SHIM 2.0. This specification is starting to get some real traction and use cases. One example use case for SHIM 2.0 would be to model an SoC (present and/or future) used in an autonomous driving processing platform. Actually, it will work for any SoC in any complex (or simple) application. Contact MCA for more information about SHIM and joining the working group.

Benchmarks for Heterogeneous Compute

In the previous newsletter, I mentioned the benchmark in development by EEMBC for analyzing heterogeneous processor architectures for automotive vision (it’s called HetMark). The benchmark helps determine optimal load balancing of the compute tasks and distribution of data across multiple compute resources. Current version, almost in beta, includes a vision pipeline with neural net for detecting and recognizing objects in an input stream. More information.

If you do not wish to receive e-mail from The Multicore Association®, you can un-subscribe. MCA sends no more than one e-mail per month to registered users at Continuing your subscription ensures you'll be notified when events and other important announcements are available.