Code Density Improvements Beyond The C Standard Extension
summary
Baseline RISC‑V C extension replaces about 53% of 32/64‑bit ops with 16‑bit, yielding ~27% code‑size reduction. Adding store/load byte/word with full register set and several immediate/branch encodings yields another ~10% aggregate saving in tests (memory ops +3%, extra branches +1%, ABI tweaks up to +5%). LLVM sensitivity to register vs immediate limits some gains. New encodings add little hardware cost (mostly decoder), and customer workloads see larger wins than standard benchmarks.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:29Company Introduction
- 1:43BK Family
- 3:25Studio
- 4:50SDK
- 7:25Why Code Density
- 9:50Test Conditions
- 18:24Cost Savings
related talks
19mCode Size Reduction Deep Dive: Zc Extensions and Dictionary Compression Custom Instruction
20mAnalysis for Code Size Opportunities in RISC V
17mUnlocking 15% More Performance: A Case Study in LLVM Optimization for RISC-V
16mImprovements to RISC-V Vector code generation in LLVM
18mPreview RISC-V GCC Features and Optimization Improvements for GCC-14
28mLLVM for RISCV
17mRecent Developments in Optimizing Compilers for RISC-V
01h 10mWhy RISC-V Through ISA Customization Demonstration