Using RISC-V in Heterogeneous Solutions to Solve Compute Challenges Pres...
summary
Automotive compute moves from many ECUs to zonal central brains and requires tens of GOPS to tens of TOPS for ADAS and thousands of TOPS for Level 4 autonomy so heterogeneous fabrics combine RISC V CPUs/DSPs for sensing, AIML accelerators for 70% of perception workloads, GPUs for scene fusion, and deterministic RISC V control for actuation. The stack demands toolchains, low and high level libraries, hardware features like vector extension ratified and support for IOMMU and PMP plus FP16/BFloat16 and safety attestation.
AI-written summary · about
chapters (10)
- 0:00Introduction to talk
- 1:04Automotive industry challenges
- 3:14Compute demands in automotive
- 5:26Heterogeneous compute explained
- 6:51Heterogeneous architecture example
- 9:22RISC-V enabling compute elements
- 10:04Software ecosystem for RISC-V
- 11:33Future areas for RISC-V
- 13:39Automotive SIG call to action
- 15:26Conclusion and future outlook
related talks
13mKeynote: Taking RISC-V Intelligence Everywhere
40mKeynote Panel: The Future of High Perfor...
45mThe Benefits of Building New AI Accelerators with RISC-V
30mRISC-V: Enabling Open Physical AI
17mRISC-V is Ready for Powering the Era of Intelligent General Computing
19mAccelerating Edge AI with the Help of RISC-V
15mKeynote: Reimagining the Future of High Performance Computing Catalysed by RISC-V
19mWhy RISC-V?