GAPuino is an Arduino Uno form-factor board that integrates GAP8 and all peripheral interfaces necessary to prototype GAP8 applications. It is compatible with most Arduino shields and features a camera connector, a range of Arduino compatible I/Os, a USB to UART/JTAG converter and external Flash and RAM over a HyperBus interface. The GAP8 SDK, that is built in part on the PULP SDK, allows users to develop, compile and execute applications on GAP8. The SDK includes a full toolchain which targets GAP8’s RISC-V cores, including support for GAP8’s extended instructions and a tool for remote debugging, flashing and control of the GAP8. The SDK provides two real time operating systems supporting GAP8: PULP OS and Arm Mbed OS.
In addition to the core toolchain, the GAP8 SDK also integrates higher level utilities for developing and porting algorithms to GAP8. These include a code generation tool, the GAP8 AutoTiler, which simplifies the use of the cluster, implementing a wide range of algorithms such as CNN layer, HoG, FIR Filters, FFT and others and TF2GAP8, which provides an automated workflow from a simple TensorFlow graph all the way through to a functional GAP8 application.
For more information on how to download the GAP8 SDK and information on the GAPuino development board, please visit: https://greenwaves-technologies.com/en/sdk/
The GAP8 IoT Application Processor, announced in February of this year, enables the cost-effective development, deployment and autonomous operation of intelligent sensing devices that capture, analyze, classify and act on the fusion of rich data sources such as images, sounds or vibrations. GAP8 is uniquely optimized to execute a large spectrum of image and audio algorithms including convolutional neural network (CNN) inference, with an energy efficiency compatible with operation on batteries for years. GAP8 integrates a communications and control processor with an eight-core computational cluster combined with a convolution hardware engine.
- Up to 250 MHz (FC) 175 MHz (Cluster) internal clock
- ~8 GOPS at a few tens of mWs
- 5x5 convolution 16 bit-fixed point in one cycle
- FC delivers 200 MOPS at
- 10mW @1.2V/250MHz
- 4mW @1.0/175MHz
- 2μA deep sleep current
- 8 μA to retain each of the four 128kB banks of L2 memory
- 1.2 V down to 1V core VDD supply
- 1.8 V to 3.3 V for I/Os
- 0.5 ms cold boot time
- 200μs to start 10μs to stop cluster