Comment by juancn
1 day ago
First time I heard about this, for those like me:
https://en.wikipedia.org/wiki/Loongson
TLDR: it's a MIPS-compatible CPU architecture
1 day ago
First time I heard about this, for those like me:
https://en.wikipedia.org/wiki/Loongson
TLDR: it's a MIPS-compatible CPU architecture
Loongson started with MIPS CPUs but current CPUs are not MIPS-compatible. LoongArch, while being very similar to MIPS, uses a different encoding. And some other details have changed. Better to say, MIPS-inspired.
What are LoongArch's technical advantages over RISC-V? In other words, why should a company develop their own architecture (which then they need to push support for) rather than use an existing, free one?
Back when LoongArch was announced, RISC-V did not yet have enough (ratified) extensions to achieve feature-parity.
Even if it had, LoongArch is much more similar to MIPS. LoongSon would have had to make more microarchitectural changes before being able to tape out their first non-MIPS CPU.
I don't know about advantages, but lead times in the chip business are long and you're not turning around on a dime without very pressing reasons. Loongson has probably had many things in the pipeline as RISC-V started gaining steam. Their current processors are more advanced designs than the best known RISC-Vs.
> LoongArch is now a fully legitimate, upstreamed ISA—with mainline Linux, Debian, and Rust support—ensuring it will be maintainable and usable at scale, but its momentum is institutional and domestic rather than market-driven: it exists to guarantee China a sovereign, unblockable CPU stack on a short timeline, not to attract global vendors, startups, or cloud ecosystems. In contrast, RISC-V is accumulating multi-vendor, cross-border adoption and economic gravity that define a global ISA; LoongArch’s success criteria are different, narrower, and largely already met.