No abstract

Application-specific systems with FPGA accelerators are often designed using high-level synthesis or hardware construction tools. Nowadays, there are many frameworks available, both open-source and commercial. In this work, we attempt to fairly compare several existing solutions (languages and tools), including Verilog (our baseline), Chisel, Bluespec SystemVerilog (Bluespec Compiler), DSLX (XLS), MaxJ (MaxCompiler), and C (Bambu and Vivado HLS). Our analysis has been carried out using a representative example of 8×8 inverse discrete cosine transform (IDCT), a widely used algorithm engaged in, among others, JPEG and MPEG decoders. The metrics under consideration include: (a) the degree of automation (how much less code is required compared to Verilog), (b) the controllability (possibility to achieve given design characteristics, namely a given ratio of the performance and area), and (c) the flexibility (ease of design modification to achieve certain characteristics). Rather than focusing on computational kernels only, we have developed AXI-Stream wrappers for the synthesized implementations, which allows adequately evaluating characteristics of the designs when they are used as parts of real computer systems. Our study shows clear examples of what impact specific optimizations (tool settings and source code modifications) have on the overall system performance and area. It emphasizes how important is to be able to control the balance between the communication interface utilization and the computational kernel performance and delivers clear guidelines for the next generation tools for designing FPGA accelerator based systems.

Application-specific systems with FPGA accelerators are often designed using high-level synthesis or hardware construction tools. Nowadays, there are many frameworks available, both open-source and commercial. In this work, we aim at a fair comparison of several languages (and tools), including Verilog (our baseline), Chisel, Bluespec SystemVerilog (Bluespec Compiler), DSLX (XLS), MaxJ (MaxCompiler), and C (Bambu and Vivado HLS). Our analysis has been carried out using a representative example of 8×8 inverse discrete cosine transform (IDCT), a widely used algorithm engaged in JPEG and MPEG decoders. The metrics under consideration include: (a) the degree of automation (how much less code is required compared to Verilog), (b) the controllability (possibility to achieve given design characteristics, namely a given ratio of the performance and area), and (c) the flexibility (ease of design modifications to achieve certain characteristics). Rather than focusing on computational kernels only, we use AXI-Stream wrappers for the synthesized implementations, which allows adequately evaluating characteristics of the designs when they are used as parts of real systems. Our study shows clear examples of what impact specific optimizations (tool settings and source code modifications) have on the overall system performance and area. It emphasizes how important is to be able to control the balance between the communication interface utilization and the computational kernel performance and delivers clear guidelines for the next generation tools for designing FPGA-accelerator-based systems.

No abstract

Data access conflicts may arise in hardware designs. One of the ways of detecting such conflicts is static analysis of hardware descriptions in HDL. We propose a static analysis-based approach to data conflicts extraction from HDL descriptions. This approach has been implemented in the Retrascope tool. The following types of conflicts are considered: simultaneous reads and writes, simultaneous writes, reading of uninitialized data, no reads between two writes. Conflict assertions are formulated as conditions on variables. HDL descriptions are automatically translated into formal models suitable for the nuXmv model checker. The translation process consists of the following steps: 1) preliminary processing; 2) Control Flow Graph (CFG) building; 3) CFG transformation into a Guarded Actions Decision Diagram (GADD); 4) GADD translation into a nuXmv format. Conflict assertions are automatically built using static analysis of the GADD model and passed to the nuXmv model checker. Bounded model checking is used to check whether these assertions are satisfiable. If true, counterexamples are generated and then translated to HDL testbenches by the Retrascope tool. The proposed approach was applied to several open source HDL benchmarks like Texas-97, Verilog2SMV, VCEGAR and mips16 modules. Potential conflicts have been detected for all of these benchmarks. Future work includes propagation of conflict assertions to the interface level (thus getting assertions on modules' communication protocols) and generation of built-in HDL checkers.