Historically, many architectures have attempted to run ILP32 software on 64-bit ISAs, such as x86-X32, mips-N32, and arm64-ILP32. However, only arm64ilp32 achieved commercial success on Apple Watch OS in a closed-source manner.

Today, we present the commercial deployment of RV64-ILP32 based on Allwinner v861 AI Glasses chips (Dual-Core RISC-V XuanTie C907) 1. This demo showcases AI Glasses running the ILP32 Linux kernel on RVA22S64. Compared to traditional RV32, performance improves significantly: iperf throughput reaches 1.5×, and lmbench shows 1.1–1.2× gains across most tests. Furthermore, another demo runs LP64 applications on an RV64-ILP32 Linux kernel within a 2GB address space for the first time, highlighting this ABI's compatibility, flexibility, and potential. This demo achievement marks a milestone in bringing 64-bit RISC-V architectural benefits to resource-constrained embedded AI devices while maintaining ILP32 memory efficiency based on an open-source software stack.

This demo illustrates ILP32 on RVA (22/23) S64.
Next, call for sponsors for ILP32 on RVA (22/23) U64!