Optimizing RISC-V Processor Performance with Adaptive Execution Unit Lengths in Harsh Environment Conditio
- authored by
- Jan Szücs, Malte Hawich, Holger Christoph Blume
- Abstract
Modern deep drilling requires a vast array of sensors, actuators and controllers, which are deployed inside the bottom end of the drill string. These electronics are exposed to temperatures up to 175°C and 200MPa of pressure, a so called harsh environment. There is no fast communication to the surface, so in situ digital signal processing is required to operate autonomously. A high-performance RISC-V processor is developed that can operate reliably regardless of the harsh environment. To produce reliable electronic circuits that work under these conditions the XT018 180nm Silicon-on-Insulator (SOI) process from XFAB was chosen as it is one of the state-of-the-art technologies for this harsh environment [1]. Typical RISC-V designs follow the established five-stage pipeline design, but this leaves only one stage for the execution phase, severely limiting the required DSP capabilities. To overcome this limitation, the custom RISC-V processor is equipped with a heterogeneous execution pipeline, resulting in delay differences between units ranging from a single cycle to up to 28 cycles. With this new instruction-dependent pipeline length the RV32IMCF processor is capable of clocking up to 180MHz at 175°C. The synchronization of the asymmetric processor execution stages is accomplished through the use of handshakes, which adds more control logic than usual to each execution unit but offloads the control workload from the controller itself. Compared to an RV32IMFC processor in this technology, which only allows works with a single cycle execution stage, this new design can achieve a clock frequency up to 18 times higher.
- Organisation(s)
-
Architectures and Systems Section
- Type
- Conference contribution
- Pages
- 1-5
- No. of pages
- 5
- Publication date
- 2024
- Publication status
- Published
- Peer reviewed
- Yes
- ASJC Scopus subject areas
- Signal Processing, Energy Engineering and Power Technology, Instrumentation, Health Informatics, Electrical and Electronic Engineering, Computer Networks and Communications
- Electronic version(s)
-
https://doi.org/10.1109/PACET60398.2024.10497073 (Access:
Closed)