People


Pete Wilson, CTO and Founder:



I am an experienced computer architect, specializing in the embedded space with demonstrated capability to both create and to synthesize concepts and apply them to improving a product; to demonstrate through modeling and simulation the value of the changes - including the construction of novel tools as needed - and to assess and suggest appropriate changes to the software toolchain to make use of the innovations, including the proposal and demonstration of new language features. My communication skills allow me to communicate the rationale, details and value of the changes to management and customers. 

My broad work history includes the initiation of RapidIO; the definition of the PowerPC Book E architecture; the creation of the e200 family of synthesizable PowerPC cores, giving access to a ~$2B lifetime TAM; the architecture of the Freescale e500 processor including the DSP-oriented SPE extensions; the VLE extensions to PowerPC architecture which added $100Ms to e200 lifetime TAM; the creation of a concurrent extension to C/C++ and of an Architecture Description Language.

Most recently, I initiated and drove the quarq Architecture Research Project in Freescale/NXP’s Discovery Labs. The project defined a highly efficient embedded-friendly architecture which deployed many (from one to hundreds or thousands) multicontext processor cores, low-cost inter-context message-passing through a wormhole mesh network, and a static homing cache coherent memory. We collaborated with the internal Freescale compiler team to get an llvm compiler built. We exercised the architecture in simulation across a wide range of application domains including network packet processing, file compression and expansion, vision systems based on Histograms of Oriented Gradients (HoG) and most recently in collaboration with Professor Omer Khan and his students at U. Connecticut looking at Convolutional Neural Networks and Graph-based problems. With reasonable estimates for power and area, we were able to demonstrate significant performance advantages over the GPUs of the time - while offering a much simpler-to-understand programming model.

As a result of the purchase of Freescale by NXP, I retired from NXP in December 2016.

I'm now working, slowly, on a simple toolkit for investigating computer and systems architecture. It's hoped that the first version of the teq toolkit will become available in 2020.