3 ms·

It easily beats compressed riscV, which was a bit of surprise. Given the riscV hype, I was assuming it would at least match thumb2 but in this benchmark it is

It easily beats compressed riscV, which was a bit of surprise. Given the riscV hype, I was assuming it would at least match thumb2 but in this benchmark it is far far behind. Here is an interesting question: riscv is assumed to have better instructions per clock than ARM. but what if it has 5% better IPC but 20% worse code density??

snvzz 4y ago
In 64bit, RISC-V does easily beat ARM and everything else. In 32bit, it is about as good as m68k and x86, which are quite dense. A good result. Just not as small as ARM thumb/thumb2. However, RISC-V is not done yet; it will see further improvements, coming from B extension (already ratified, significant for size, and yet wasn't in the test) and from the ongoing Zc size-reduction work[0], which does already achieve a sizable reduction[1]. 0. https://github.com/riscv/riscv-code-size-reduction https://github.com/riscv/riscv-code-size-reduction 1. https://docs.google.com/spreadsheets/d/1bFMyGkuuulBXuIaMsjBINoCWoLwObr1l9h5TAWN8s7k/edit#gid=1837831327 https://docs.google.com/spreadsheets/d/1bFMyGkuuulBXuIaMsjBI...