ARTICo³
ARTICo³ is an open-source run-time reconfigurable processing architecture to enable hardware-accelerated high-performance embedded computing. It is highly flexible, offering adaptive computing performance, energy efficiency, and fault tolerance on demand.
Framework Components
The ARTICo³ Framework provides three components to deploy reconfigurable and adaptive multi-accelerator systems:
- A hardware-based reconfigurable processing architecture – ARTICo³
- An automated toolchain to implement multi-accelerator systems
- A runtime to manage FPGA reconfiguration and accelerator execution
The ARTICo³ architecture operates in a processor-coprocessor approach, where computationally intensive tasks are offloaded to one or several accelerators to exploit both task- and data-level parallelism.
Hence, the Framework takes as inputs the host application code and the descriptions of the hardware accelerators. These descriptions can be provided as low-level HDL code – VHDL and Verilog are supported – or as high-level C/C++ code – High-Level Synthesis (HLS) is used in this scenario.
Then, the Framework automatically composes and builds the multi-accelerator system, providing as outputs the application binaries and the FPGA configuration files (i.e., bitstreams). It also manages all Dynamic and Partial Reconfiguration (DPR) processes, and schedules all data-parallel processing over the pool of instantiated hardware accelerators at run time.
Currently, the ARTICo³ Framework supports the following platforms:
- Xilinx Zynq-7000 SoC
- Xilinx Zynq UltraScale+ MPSoC
Contact
If you are interested in the ARTICo³ Framework or if you want to request further information, do not hesitate to contact us. Technical issues should be handled via GitHub Issues.