Making the Case for a Keccak Instruction
summary
Proposes a dedicated Keccak/SHA3 instruction that packs the 1600-bit Keccak state across vector registers and an operand collector, showing the permutation costs ~2,000 software cycles vs ~24 cycles in hardware. Benchmarks on Kyber and Dilithium find 60–80% of cycles spent in Keccak; handwritten intrinsics give ~3× speedup, but a single instruction yields an order-of-magnitude (10×+) improvement, saving significant server TLS CPU and energy and working on prototype silicon with 256-bit collectors.
AI-written summary · about
chapters (7)
- 0:00Introduction to Keccak instruction
- 1:50Current RISC-V crypto standards
- 3:36Post-quantum cryptography standards
- 6:47Performance impact on post-quantum crypto
- 9:16Keccak’s software bottleneck
- 12:06Proposal for Keccak instruction
- 17:22Conclusion and next steps
related talks
20mBenchmarking RISC-V Post-Quantum
19mPQCP Support for RISC-V Vector, Future Keccak Extensions
19mThe RISC-V Vector Cryptographic Extensions in the Real World
50mProgress in Standardizing Cryptography Extensions for RISC-V Processors
21mThe RISC-V Vector Cryptography Extensions
40mImplementation
31mPost-Quantum Crypto: Challenges for Embedded Applications
41mFault Attacks & Side Channels
