Automatic Heterogeneous Compilers allows blended hardware-software solutions to be explored without the cost of a full-fledged design team, but limited research exists on current partitioning algorithms responsible for separating hardware and software. The purpose of this thesis is to implement various partitioning algorithms onto the same automatic heterogeneous compiler platform to create an apples to apples comparison for AHC partitioning algorithms. Both estimated outcomes and actual outcomes for the solutions generated are studied and scored. The platform used to implement the algorithms is Cal Poly’s own Twill compiler, created by Doug Gallatin last year. Twill’s original partitioning algorithm is chosen along with two other partitioning algorithms: Tabu Search + Simulated Annealing (TSSA) and Genetic Search (GS). These algorithms are implemented inside Twill and test bench input code from the CHStone HLS Benchmark tests is used as stimulus. Along with the algorithms cost models, one key attribute of interest is queue counts generated, as the more cuts between hardware and software requires queues to pass the data between partition crossings. These high communication costs can end up damaging the heterogeneous solution’s performance. The Genetic, TSSA, and Twill’s original partitioning algorithm are all scored against each other’s cost models as well, combining the fitness and performance cost models with queue counts to evaluate each partitioning algorithm. The solutions generated by TSSA are rated as better by both the cost model for the TSSA algorithm and the cost model for the Genetic algorithm while producing low queue counts.

The DC House project at Cal Poly State University faces a challenge of supplying DC voltage to household appliances. Each appliance in the DC House constitutes a DC load that has a unique voltage and power rating, hence the need to develop a smart DC wall plug that will automatically adjust to the operating voltage required by any DC load. This thesis entails a proof of concept design of the smart DC wall plug which can automatically detect an appliance’s voltage rating. The design employs a dc-dc converter in conjunction with a microcontroller to sense load current to properly adjust the required load voltage. Hardware implementation to demonstrate the functionality of the smart wall plug was developed. Results performed on several dc loads show that the smart wall plug is able to adjust to the required load voltage within an acceptable range. An algorithm to improve the accuracy was attempted and presented along with the results. Further recommendations to improve the current design will also be discussed.

FPGAs are used in a wide variety of digital systems. Due to their ability to support parallelism and specialization, these devices are becoming more commonplace in fields such as machine learning. One of the biggest benefits of FPGAs, logic specialization, can lead to security risks. Prior research has shown that a large variety of malicious circuits can snoop on sensitive user data, induce circuit faults, or physically damage the FPGA. These Trojan circuits can easily be crafted and embedded in FPGA designs. Often, these Trojans are small, consume little power in comparison to the target circuit, and are hard to detect via simulation or physical inspection. Computer-aided design (CAD) software in FPGAs has been the subject of extensive research and development of FPGAs for the past thirty-five years. The current FPGA software landscape includes vendors that provide widely used software flows to convert behavioral and register-transfer level (RTL) descriptions to bitstreams needed to program an FPGA device. Given the complexity of the algorithms needed to perform this translation, these CAD tool flows are generally structured as black boxes with limited transparency regarding design conversion steps or the logical equivalence of the generated design and initial design specification. vi This work explores the enhancement of open-source FPGA software, SymbiFlow, that focuses on FPGA RTL synthesis, place and route and bitstream generation. SymbiFlow uses Yosys for synthesis, VPR for place and route, and Project X-Ray for bitstream generation. We focus on synthesis using Yosys and formal verification using the Cadence Conformal Logic Equivalence Checker (LEC) for Xilinx Artix-7 FPGAs. Yosys is used to synthesize 160 benchmarks written in Verilog. We implement required code modifications to Yosys for designs to pass the equivalence checker. For Conformal, this work involves processing 160 benchmark designs with the equivalence checker. Parameters can be toggled on or off to obtain results that indicates if a design has passed formal verification when comparing RTL and synthesized netlists.

The open-source community is a valuable resource for many hobbyists and researchers interested in collaborating and contributing towards publicly available tools. In the area of field programmable gate arrays (FPGAs) this is no exception. Contributors seek to reverse-engineer the functions of large proprietary FPGA devices. An interesting challenge for open-source FPGA engineers has been reverse-engineering the operation and bitstreams of digital signal processing (DSP) blocks located in FPGAs. SymbiFlow is an open-source FPGA toolchain designed as a free alternative to proprietary computer-aided design tools like Xilinx’s Vivado. For SymbiFlow, mapping logical multipliers to DSP blocks and generating DSP block bitstreams has been left unimplemented for the Artix-7 family of FPGAs. This research seeks to rectify this shortcoming by introducing DSP information for the place and route functions into SymbiFlow. By delving into the SymbiFlow architecture definitions and creating functioning FPGA assembly code (FASM) files for Project X-Ray, a bitstream generator for Artix-7, we have been able to determine the desired output of the open-source Versatile Place & Route tool that will generate a working DSP bitstream. We diagnose and implement changes needed throughout the SymbiFlow toolchain, allowing for DSP design bitstreams to be successfully generated with open-source tools.

<p>Energy efficient machine learning accelerator design</p>

No description available.

No description available.

No description available.

No description available.

No description available.