Session
Look Both Ways: New Directions in High-Level Synthesis and Approximate Computing
DescriptionThis session presents 6 exciting papers describing recent advances in high-level synthesis (HLS) and approximate computing. On the HLS front, four papers describe new contributions to HLS-generated asynchronous dataflow circuits; the generation of circuits that perform speculative execution efficiently; the use of multiple clock domains to reduce power; and, automated selection of kernels for hardware acceleration. The last two papers are related to approximate computing. The first of these automates design-space exploration in HLS to identify design points with high potential for area/power savings from approximation. The next introduces an efficient approach for binary-to-unary number conversion.
Event Type
Research Manuscript
TimeWednesday, June 251:30pm - 3:00pm PDT
Location3004, Level 3
EDA
EDA5: RTL/Logic Level and High-level Synthesis
Presentations
| 1:30pm - 1:45pm PDT | PipeLink: a pipelined resource sharing system for dataflow high-level synthesis | |
| 1:45pm - 2:00pm PDT | Optimizing Recovery Logic in Speculative HLS | |
| 2:00pm - 2:15pm PDT | AutoClock: Automated Clock Management for Power-Efficient HLS Designs on FPGAs | |
| 2:15pm - 2:30pm PDT | Cayman: Custom Accelerator Generation with Control Flow and Data Access Optimization | |
| 2:30pm - 2:45pm PDT | ADVISOR: Approximate Computing-frienDly High-LeVel Synthesis DesIgn Space ExplORer | |
| 2:45pm - 3:00pm PDT | Comparison-Free Bit-Stream Generation for Cost-Efficient Unary Computing |


